Gavin Baker on Orbital Compute, TSMC, and Frontier Models

Full Transcript

https://www.youtube.com/watch?v=Mmj_G9RlW-I

[00:00] What was happening in AI was I think the most extraordinary moment in the history of capitalism, the history of American business.
[00:08] Anthropic they added 11 billion of ARR.
[00:11] The three highest profile SAS companies founded in the last 10 12 years are Palunteer, Snowflake and Data Bricks.
[00:21] And these three companies spent 10 years building their businesses.
[00:23] Anthropic added their combined businesses in one month.
[00:30] That's just nothing like that has ever happened in the history of capitalism.
[00:35] Forget my career.
[00:35] Just the flatout history of capitalism, the history of business.
[00:53] All right.
[00:53] So, this is our uh sixth time doing this, if you can believe it, which puts you back into first place uh or at least tied for first place with Girly.
[01:01] Least tied for first place with Girly.
[01:03] And I think even since last time when we did this, which was so exciting and spectacular, I think we're in an even more interesting time now.
[01:10] Maybe just start by riffing on how it felt for you living through March and April of this year, which which felt to me just like a completely unique economic, technology, and market environment.
[01:22] And you're the biggest student of of of history and of these times.
[01:24] So what does it feel like?
[01:25] I would say broadly speaking there are two kinds of draw downs.
[01:27] They're drawowns where you're wrong, a company misestimates, your hypothesis was invalidated and you have to take your medicine and you crystallize that loss.
[01:41] And then there are draw downs or periods of underperformance where you you're underperforming because of companies you know really really well and where you profoundly disagree with the price action and you can lean in and instead of crystallizing uh negative performance you can kind of build pent up alpha pent up future performance and for me that is what
[02:03] performance and for me that is what March felt like.
[02:05] It felt like uh you March felt like.
[02:08] It felt like uh you know the NASDAQ was selling off and at the same time what was happening in AI was I think the most extraordinary moment in the history of capitalism the history of American business and what I just mean by that is that anthropic they added 11 billion of ARR and what is astonishing to me about this is that the SAS and cloud revolution it created we'll call it between 5 and 10 trillion dollars of value and I would say Arguably the three highest profile SAS companies to have kind of been founded in the last 10 12 years are Palunteer, Snowflake and Data Bricks.
[02:51] And these three companies have spent employ thousands of people, tens of thousands collectively.
[02:54] They've all spent 10 years building their businesses.
[02:57] And Anthropic added their combined businesses in one month.
[03:00] month.
[03:02] That's just nothing like that has ever
[03:04] That's just nothing like that has ever happened in the history of capitalism.
[03:06] happened in the history of capitalism.
[03:09] Forget my career. Just the flatout history of capitalism, the history of business.
[03:12] I wild. And then Krishna comes on this show and shares some stats.
[03:15] 500% in DR.
[03:19] Yeah. You do the math on that for three years. Insanity.
[03:22] We So there's just no precedent for this.
[03:25] And we, you know, tech tech investors, you hear a lot of discussions about S-curves and investing in exponentials.
[03:28] I've just never seen an exponential like this.
[03:34] It felt even more extreme than Deepseek, which was a very similar setup.
[03:36] If we go back to 25 and there's a huge sell-off on Deepseek,
[03:41] which was very strange because the paper gets published 7 days before Deepseek Monday.
[03:43] got published, I believe, on a Monday that was a holiday in America.
[03:54] And I read it, I thought, hm, you know, this this feels like it might not read that positively for um you know, the AI
[04:07] That positively for, um, you know, the AI trade.
[04:10] You I took action.
[04:10] We had DeepSeek Monday where AI really imploded a week later and that was really strange.
[04:18] Because by DeepSec Monday it was super clear that this was going to be the most positive thing that had ever happened to compute demand.
[04:25] Prices in the AWS available availability zones in Asia had already like doubled.
[04:31] You were seeing GPU availability go down.
[04:34] And this was just the first time we saw how much more compute-hungry reasoning models are during inference than non-reasoning models.
[04:46] And so that was a similar setup, but you had to do some work to see that.
[04:51] I mean, it's not that hard to say, oh wow, stocks are selling off.
[04:55] The price of DRAM is going vertical.
[04:57] The price of GPUs in Asia going vertical.
[05:00] GPU availability is going down.
[05:02] And then like two or three days later, you know, GPU prices in in America started going up, GPU rental prices.
[05:07] All you had to do
[05:09] Up, GPU rental prices.
[05:13] All you had to do in in March was just simply observe what was happening to Anthropic.
[05:14] There there's all these people who seem to regret,
[05:16] regret, you know, not buying during '22, not buying during COVID, not buying during Deep Seek.
[05:22] You had the same valuation setup at the beginning of April and an even clearer AI inflection.
[05:32] And so there have been all these chances to buy into AI and then of course what complicated it was the straight of.
[05:39] Foremost I became a believer and am a believer that I think maybe one thing that the market was mispricing.
[05:47] And I'm I'm no macro expert, I do do a lot of pro national security investing.
[05:53] And so I do have access to people who are experts and are excited to share their thoughts and opinions with me.
[06:02] That the straight of horm being closed is actually relatively awesome for America.
[06:05] Awesome for America.
[06:07] Why?
[06:08] Because particularly for the goals of
[06:10] Because particularly for the goals of the current administration.
[06:12] The current administration. So electricity is a very important industrial or manufacturing input.
[06:17] The key input into American electricity prices which feeds into AI is in G1 natural gas on Bloomberg that was down 20%.
[06:26] And natural gas in Asia, Europe, everywhere else doubled or tripled.
[06:35] So our relative manufacturing competitiveness improved overnight and for better or worse that is what the Trump administration seems to care about.
[06:43] They are very focused on America's relative position.
[06:48] And I think a lot of people had memories of the 1970s.
[06:53] And what made the 70s so traumatic was it wasn't just that prices went up, it's that there were actual gas shortages.
[07:01] And then you go through, okay, well the US economy is dramatically less energy intensive than it was.
[07:07] US econ the United States is now the world's largest producer of oil and gas and we've become
[07:12] producer of oil and gas and we've become now the world's largest exporter of oil.
[07:15] now the world's largest exporter of oil and gas and then on top of that there's.
[07:18] and gas and then on top of that there's this relative manufacturing advantage.
[07:21] this relative manufacturing advantage and so that made it I think easier to.
[07:26] and so that made it I think easier to stay focused on AI fundamentals stay.
[07:30] stay focused on AI fundamentals stay focused on what were historically.
[07:33] focused on what were historically attractive valuations I think on a.
[07:35] attractive valuations I think on a relative basis.
[07:37] relative basis tech essentially got as cheap as it's.
[07:38] tech essentially got as cheap as it's been versus the rest of the market has.
[07:41] been versus the rest of the market has at any point over the last 10 years and.
[07:44] at any point over the last 10 years and just think about that in the context of.
[07:45] just think about that in the context of market efficiency. We have the most.
[07:47] market efficiency. We have the most extraordinary moment in the history of.
[07:49] extraordinary moment in the history of capitalism that's wildly bullish for AI.
[07:53] capitalism that's wildly bullish for AI and you get a chance to buy AI.
[07:56] and you get a chance to buy AI at really attractive valuation. What do.
[07:59] at really attractive valuation. What do you make of the multiples that.
[08:01] you make of the multiples that specifically Anthropic and OpenAI, which.
[08:03] specifically Anthropic and OpenAI, which in my mind are like the reference assets.
[08:05] in my mind are like the reference assets that are the most pure play takes on.
[08:07] that are the most pure play takes on this trend really being not that crazy?
[08:10] this trend really being not that crazy? Like if you just look at the sales.
[08:11] Like if you just look at the sales multiple and compare it to maybe what.
[08:13] multiple and compare it to maybe what data bricks and snowflake and these.
[08:15] data bricks and snowflake and these companies traded at at their peak like how do you make sense of it?
[08:18] I do think OpenAI and Enthropic are pretty different animals from a capital efficiency perspective.
[08:19] OpenAI and Enthropic are pretty different animals from a capital efficiency perspective.
[08:21] And Enthropic clearly is has a dramatically lower cost per token than OpenAI.
[08:24] And Enthropic clearly is has a dramatically lower cost per token than OpenAI.
[08:26] They just do.
[08:30] And you can just see that in the amount of money that they have burned to get to a roughly similar revenue scale.
[08:31] you can just see that in the amount of money that they have burned to get to a roughly similar revenue scale.
[08:35] I think have have they burned maybe 80% less than OpenAI.
[08:37] I think have have they burned maybe 80% less than OpenAI.
[08:39] than OpenAI.
[08:41] So as businesses, they clearly have very different structural ROIC's.
[08:44] So as businesses, they clearly have very different structural ROIC's.
[08:46] I think OpenAI is doing a lot.
[08:48] I think Sarah Frier is one of the most exceptional CFOs.
[08:49] CFOs.
[08:51] I think they're doing a lot of things to try to improve this.
[08:53] and they've secured a lot of compute.
[08:55] and they've secured a lot of compute.
[08:55] more more than they've secured a lot of compute.
[08:57] That's another big difference.
[08:59] Um it turns out being aggressive really paid but yeah I just anthropic at 900 billion for 50 billion and you know ARR and you know I.
[09:02] being aggressive really paid but yeah I just anthropic at 900 billion for 50 billion and you know ARR and you know I.
[09:05] growing a thousand%.
[09:09] billion and you know ARR and you know I.
[09:10] Yeah, growing at ridiculous rates.
[09:13] Maybe a true statement is that if Anthropic
[09:15] A true statement is that if Anthropic had all the compute, they'd probably be doing well north of hundred billion dollars today, maybe 150.
[09:26] And I do, you know, they have clearly deprecated the intelligence of Claude.
[09:31] There's an analysis Claude is even on Opus is generating 70% less tokens for the exact same question.
[09:37] And you know, as we talked about last time, token quantity equals quality of answer and quality of thinking at some level.
[09:43] You know and there is an intelligence density per token that also matters you know I think I felt that as as a user so I think they would be doing materially more 100 150 maybe 200 billion so you might be buying it at more like five times unconstrained.
[09:59] I'm going to make up a new number urr unconstrained run rate revenue yes.
[10:09] Why do you think they don't raise $und00 billion at a $3 trillion valuation or something like this.
[10:14] Like if you were
[10:16] something like this.
[10:18] Like if you were the anthropic CFO, uh Krishna is the anthropic CFO, uh Krishna is awesome.
[10:19] We just had him on.
[10:21] Or if you're the open if you're Sarah, certainly if if the inbound I received following the Krishna episode is any indication, everyone I've ever met is trying to invest in in both these companies.
[10:24] following the Krishna episode is any indication, everyone I've ever met is trying to invest in in both these companies.
[10:27] indication, everyone I've ever met is trying to invest in in both these companies.
[10:28] trying to invest in in both these companies.
[10:29] companies.
[10:33] So I think it's wise it the future is uncertain.
[10:36] it the future is uncertain.
[10:38] you are clearly in a very capital intensive game even if you are you know Enthropic
[10:41] intensive game even if you are you know Enthropic
[10:43] Enthropic um I'm sure is at very positive gross margins on inference today I think probably starts generating cash this year if they are not already generating cash which I think is probably the case
[10:45] um I'm sure is at very positive gross margins on inference today I think probably starts generating cash this year if they are not already generating cash which I think is probably the case
[10:48] margins on inference today I think probably starts generating cash this year if they are not already generating cash which I think is probably the case
[10:49] probably starts generating cash this year if they are not already generating cash which I think is probably the case
[10:51] year if they are not already generating cash which I think is probably the case
[10:55] cash which I think is probably the case but still you probably want to be able to raise more capital access more compute the world is uncertain Ukraine
[10:57] but still you probably want to be able to raise more capital access more compute the world is uncertain Ukraine
[10:59] to raise more capital access more compute the world is uncertain Ukraine is starting to really really win how is Russia going to respond ond, you know, I think there's still a lot of uncertainty in Iran.
[11:01] is starting to really really win how is Russia going to respond ond, you know, I think there's still a lot of uncertainty in Iran.
[11:04] Russia going to respond ond, you know, I think there's still a lot of uncertainty in Iran.
[11:06] think there's still a lot of uncertainty in Iran.
[11:07] in Iran. All this uncertainty, I think, probably amplifies geopolitical uncertainty over Taiwan.
[11:10] All this uncertainty, I think, probably amplifies geopolitical uncertainty over Taiwan.
[11:12] probably amplifies geopolitical uncertainty over Taiwan.
[11:14] uncertainty over Taiwan. So, it's an uncertain world. If if I think about
[11:16] uncertain world. If if I think about Elon, Elon has always made investors money.
[11:19] Elon, Elon has always made investors money. He treats it like a sacred covenant.
[11:21] money. He treats it like a sacred covenant. And as a result, because he's made people money for now 20 years, he has a superpower.
[11:24] covenant. And as a result, because he's made people money for now 20 years, he has a superpower. And that is he can essentially raise as much capital as he wants, whenever he wants.
[11:27] made people money for now 20 years, he has a superpower. And that is he can essentially raise as much capital as he wants, whenever he wants.
[11:29] has a superpower. And that is he can essentially raise as much capital as he wants, whenever he wants. And I think it's wise that these companies are taking I don't know if that's how they think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
[11:32] essentially raise as much capital as he wants, whenever he wants. And I think it's wise that these companies are taking I don't know if that's how they think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
[11:35] wants, whenever he wants. And I think it's wise that these companies are taking I don't know if that's how they think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
[11:37] it's wise that these companies are taking I don't know if that's how they think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
[11:38] taking I don't know if that's how they think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
[11:41] think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
[11:45] focused on making investors money is wise and creates benefits that don't just last for like a year or two. They can last for the next 20 to 30 years.
[11:49] just last for like a year or two. They can last for the next 20 to 30 years.
[11:52] just last for like a year or two. They can last for the next 20 to 30 years.
[11:54] can last for the next 20 to 30 years. >> And the way Elon did this was sort of systematically underpricing SpaceX or whatever else.
[11:57] >> And the way Elon did this was sort of systematically underpricing SpaceX or whatever else. Like what is the actual method?
[11:59] systematically underpricing SpaceX or whatever else. Like what is the actual method? Just never being greedy on valuation, never pushing valuation.
[12:00] whatever else. Like what is the actual method? Just never being greedy on valuation, never pushing valuation.
[12:02] method? Just never being greedy on valuation, never pushing valuation. >> Just that simple.
[12:04] Just never being greedy on valuation, never pushing valuation. >> Just that simple.
[12:06] never pushing valuation. >> Just that simple. >> You know, my friend Antonio pointed out SpaceX compounded, you know, low 30% per year for whatever that was a decade.
[12:08] >> Just that simple. >> You know, my friend Antonio pointed out SpaceX compounded, you know, low 30% per year for whatever that was a decade.
[12:09] >> You know, my friend Antonio pointed out SpaceX compounded, you know, low 30% per year for whatever that was a decade. And and that was just because Elon was, I
[12:12] SpaceX compounded, you know, low 30% per year for whatever that was a decade. And and that was just because Elon was, I
[12:16] year for whatever that was a decade. And and that was just because Elon was, I
[12:18] And that was just because Elon was, I think, focused on preserving the superpower and having trying to strike a fair balance between investors and employees.
[12:26] But I I think it's wise.
[12:28] But could Anthropic raise money at probably at least a 100% premium to this rumored latest mark?
[12:38] Of course, most software companies try to maximize your time on their app to juice engagement.
[12:41] RAMP does the exact opposite.
[12:43] RAMP understands that no one wants to spend hours chasing receipts, reviewing expense reports, and checking for policy violations.
[12:51] So, they built their tools to give that time back using AI to automate 85% of expense reviews with 99% accuracy.
[12:59] And since Ramp saves companies 5%, it's no wonder that Shopify runs on RAM, Stripe runs on RAM, and my business does, too.
[13:05] To see what happens when you eliminate the busy work, check out ramp.com/invest.
[13:10] Felix by Rogo is a personal finance agent that turns a single prompt into finished client ready work using your firm's own templates, context, and standards.
[13:17] Send Felix an email like,
[13:19] Standards. Send Felix an email like, "Take these comments and turn them for me or update my tracker with the context of these emails."
[13:25] Or, "Run the ability to pay math on this buyer and Felix sends back finished PowerPoint decks, Excel models, and sourced research."
[13:31] Felix works the way your team already does, delivering work quickly and accurately around the clock.
[13:36] Learn more at rogo.ai/felix.
[13:40] OpenAI, Cursor, Enthropic, Perplexity, and Versell all have something in common.
[13:44] They all use WorkOS.
[13:47] And here's why.
[13:49] To achieve enterprise adoption at scale, you have to deliver on core capabilities like SSO, skim, arbback, and audit logs.
[13:55] That's where work OS comes in.
[13:57] Instead of spending months building these mission critical capabilities yourself, you can just use work OS APIs to gain all of them on day zero.
[14:02] That's why so many of the top AI teams you hear about already run on work OS.
[14:07] Work OS is the fastest way to become enterprise ready and stay focused on what matters most, your product.
[14:13] Visit works.com to get started.
[14:15] Let's get to the Watson wafers part of the discussion.
[14:19] Always my favorite thing to talk about with you.
[14:19] Uh the importance
[14:21] Talk about with you.
[14:21] Uh the importance of this infrastructure buildout.
[14:24] I feel of this infrastructure buildout.
[14:24] I feel like every time I feel like it's getting overheated and then the next time I talk to you, it seems like we should have done way more than we did.
[14:27] overheated and then the next time I talk to you, it seems like we should have done way more than we did.
[14:29] And you've studied S-curves and the steepness of those S-curves a lot.
[14:33] Uh and you know a lot about history.
[14:35] Talk us through how you're thinking about Watson wafers today as the key to inputs into this whole thing.
[14:42] Yeah, I would say I think capitalism is going to solve the Watts shortage absent big regulatory political blowback.
[14:49] which I think is a real possibility.
[14:53] the head of kind of data center infrain investing at one of the big PE firms.
[14:58] You know, I think Blackstone, Apollo, KKR said it used to be energy and chips were our biggest gating factors.
[15:04] Now it's zoning and approval much more important.
[15:10] And I think a lot of companies are waiting till after the midterms to take action in terms of maybe workforce reductions.
[15:17] Nobody wants to be, you know, piñata during the
[15:22] To be, you know, piñata during the midterms.
[15:24] Midterms, but you know, you've seen a lot of companies that make turbines significant.
[15:26] But you know, you've seen a lot of companies that make turbines significant announce of plans to significantly increase capacity.
[15:30] Announce of plans to significantly increase capacity.
[15:32] There's like two of these machines that can cast these big blades.
[15:34] These machines that can cast these big blades.
[15:37] We haven't made one in 80 years in the West.
[15:39] We don't know how to make them anymore, etc., etc., etc.
[15:42] All of that is true.
[15:44] And I and and by no means am I minimizing, you know, the industrial engineering, you know, magic and artistry that goes into those, but capitalism is very good at solving problems like these over time.
[15:49] And artistry that goes into those, but capitalism is very good at solving problems like these over time.
[15:52] There's other sources of energy besides these turbines with a longer time frame.
[15:54] So I think the watts shortage will probably begin to alleviate 27 28 and then I think orbital compute will really solve that.
[15:59] And I do I do want to like reframe orbital compute because I think when people hear data centers in space they which we discussed on our last episode they picture a pentagon sized building in space.
[16:09] They're like well we can't do
[16:23] in space.
[16:23] They're like well we can't do that.
[16:23] That's not what it is.
[16:28] A blackwell that.
[16:28] That's not what it is.
[16:28] A blackwell rack weighs you 3,000 lbs.
[16:31] It's 8 ft rack weighs you 3,000 lbs.
[16:31] It's 8 ft high.
[16:31] It's 4 feet deep.
[16:31] 3 feet wide.
[16:35] high.
[16:35] It's 4 feet deep.
[16:35] 3 feet wide.
[16:35] It's racks in space.
[16:38] And SpaceX has it's racks in space.
[16:38] And SpaceX has showed you an illustration and it's a
[16:41] showed you an illustration and it's a rack.
[16:41] That's the satellite.
[16:41] Uh but it's probably about the size of a blackwell
[16:44] rack.
[16:44] That's the satellite.
[16:44] Uh but it's probably about the size of a blackwell
[16:46] probably about the size of a blackwell rack.
[16:46] It has these solar wings that are
[16:48] rack.
[16:48] It has these solar wings that are probably 500 ft long on each side.
[16:51] probably 500 ft long on each side.
[16:51] You keep it in a sunsynchronous orbit.
[16:54] keep it in a sunsynchronous orbit.
[16:54] So those solar panels are always in the
[16:56] those solar panels are always in the sun.
[16:56] And then because it's in an exactly
[16:58] sun.
[16:58] And then because it's in an exactly sunsynchronous orbit, the radiator which
[17:01] sunsynchronous orbit, the radiator which extends behind it for hundreds of feet.
[17:05] extends behind it for hundreds of feet.
[17:05] >> This is a common criticism.
[17:05] Yeah.
[17:06] how
[17:06] you going to go.
[17:07] >> I've spent a lot of time at Starbase
[17:11] over the years and I've talked to a lot
[17:13] of SpaceX engineers and I do think it is
[17:16] the most talented group of engineers on
[17:19] planet Earth and they're very confident
[17:21] they have solved this and they're not
[17:23] they have solved this and they're not always confident like I think probably.
[17:27] always confident like I think probably you know there's some engineering that needs to happen to turn the starship into a Mars colonial transporter.
[17:30] needs to happen to turn the starship into a Mars colonial transporter.
[17:32] Will they do that? Absolutely.
[17:34] What are they more focused on? I would say probably, you know, the repair and maintenance.
[17:37] more focused on? I would say probably, you know, the repair and maintenance.
[17:39] These are the two big, you know, the two big responses, the radiator and and how do you repair the whatever issue goes wrong in the rack.
[17:40] big responses, the radiator and and how do you repair the whatever issue goes wrong in the rack.
[17:42] do you repair the whatever issue goes wrong in the rack.
[17:44] And the answer is like until you have probably an, you know, floating optimuses.
[17:45] wrong in the rack. And the answer is like until you have probably an, you know, floating optimuses.
[17:47] You don't. Now, I do think Starship is going to change the space economy in ways we cannot imagine.
[17:50] optimuses. You don't. Now, I do think Starship is going to change the space economy in ways we cannot imagine.
[17:52] Starship is going to change the space economy in ways we cannot imagine. Particularly if regulation becomes a constraint to data centers, none of it's going to matter.
[17:54] economy in ways we cannot imagine. Particularly if regulation becomes a constraint to data centers, none of it's going to matter.
[17:56] Particularly if regulation becomes a constraint to data centers, none of it's going to matter. you're going to sell as much orbital compute as you can make.
[17:57] constraint to data centers, none of it's going to matter. you're going to sell as much orbital compute as you can make.
[17:59] going to matter. you're going to sell as much orbital compute as you can make. And then obviously you link these racks using lasers traveling through vacuum which are already on every Starlink.
[18:01] much orbital compute as you can make. And then obviously you link these racks using lasers traveling through vacuum which are already on every Starlink.
[18:04] And then obviously you link these racks using lasers traveling through vacuum which are already on every Starlink. And it's just it's just mindblowing to me that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit.
[18:07] using lasers traveling through vacuum which are already on every Starlink. And it's just it's just mindblowing to me that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit.
[18:10] which are already on every Starlink. And it's just it's just mindblowing to me that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit.
[18:11] And it's just it's just mindblowing to me that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit.
[18:15] it's just it's just mindblowing to me that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit.
[18:17] that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit. Every
[18:24] of all satellites in orbit.
[18:24] Every Starlink they're cooling it today.
[18:28] Starlink they're cooling it today.
[18:28] And you know, I think Starlink V3 is going to operate at 20 kW.
[18:33] A Blackwell rack is only 100 kows.
[18:37] And people talk a lot about density.
[18:39] Well, if you're connecting the racks with lasers through vacuum, you know, you can make the rack bigger physically.
[18:45] You're focused on weight, not size.
[18:49] In a data center on Earth where you're trying to connect racks, ideally using copper, minimize lengths, etc., etc.
[18:53] Cabling is a big cost.
[18:56] um you do want that rack to be small because you know copper when you can, optics when you must.
[19:01] But in space, you know, there's all sorts of things that SpaceX can do that I think maybe some of these naysayers are not contemplating, but it's just they operate more satellites than you want.
[19:12] They have a 20 kow satellite today, so maybe you just scale that up to 60 kilowatts to start.
[19:18] They seem very confident they're going to go right to 100 to 120.
[19:23] And they also the same company now also operates the largest
[19:26] company now also operates the largest data center on earth.
[19:28] They have the data center on earth.
[19:31] They have the world's best hardware engineers and all sorts of people almost all of whom are not smart enough or practical enough to work at SpaceX are these armchair skeptics.
[19:37] not smart enough or practical enough to work at SpaceX are these armchair skeptics.
[19:40] work at SpaceX are these armchair skeptics.
[19:42] skeptics.
[19:43] You know, I don't want to quote Larry Ellison, but somebody was, you know, being skeptical and Larry and Larry was just like, "Listen, he's out there landing rockets.
[19:47] being skeptical and Larry and Larry was just like, "Listen, he's out there landing rockets.
[19:49] just like, "Listen, he's out there landing rockets.
[19:51] landing rockets.
[19:53] I don't see anybody else landing rockets.
[19:56] And the reality is is that 10 years later, no other company is consistently capable of landing and fully reusing an orbital rocket.
[19:59] is consistently capable of landing and fully reusing an orbital rocket.
[20:03] And fully reusing an orbital rocket.
[20:05] And none of this works makes sense without reusability.
[20:06] That means you have to land it.
[20:08] I would like to redefine orbital compute has racks in space, not giant floating pentagoniz data centers in space, which is just, you know, that's silly.
[20:14] floating pentagoniz data centers in space, which is just, you know, that's silly.
[20:17] data centers in space, which is just, you know, that's silly.
[20:19] But you can, you know, what makes a data center is you're connecting these racks with lasers.
[20:23] connecting these racks with lasers.
[20:24] So it'll be racks in space that are connected with lasers into a virtual data center.
[20:26] connected with lasers into a virtual data center.
[20:26] And and if you think about
[20:28] data center. And and if you think about that state of the world, let's say that
[20:31] that state of the world, let's say that all happens and we're really good at
[20:32] all happens and we're really good at getting these things up economically,
[20:34] getting these things up economically, running matrix multiplication all over
[20:36] running matrix multiplication all over space. What does that mean for
[20:37] space. What does that mean for terrestrial data centers? Someone once
[20:40] terrestrial data centers? Someone once said, um, you know, America was going to
[20:44] said, um, you know, America was going to suck as hard as it can on every energy
[20:47] suck as hard as it can on every energy source it can get. And I just think the
[20:49] source it can get. And I just think the same is true of compute.
[20:51] same is true of compute. >> It's why I'm probably less worried about
[20:53] >> It's why I'm probably less worried about like an edge AI barecase than I was.
[20:57] like an edge AI barecase than I was. >> We're going to consume as much compute
[21:02] >> We're going to consume as much compute as we can. And inference I think is very
[21:06] as we can. And inference I think is very sensible for orbital compute. Training
[21:09] sensible for orbital compute. Training will be done on Earth for a long time.
[21:12] will be done on Earth for a long time. So I don't think that this is super
[21:14] So I don't think that this is super bearish for terrestrial data centers. I
[21:16] bearish for terrestrial data centers. I think those are going to be valuable for
[21:18] think those are going to be valuable for my lifetime.
[21:21] my lifetime. But I do think if you are in this
[21:23] But I do think if you are in this ecosystem of power production and
[21:26] ecosystem of power production and cooling and you are massively ramping
[21:31] cooling and you are massively ramping capacity and you know a lot of these
[21:33] capacity and you know a lot of these capacity ramps are going to be hitting
[21:36] capacity ramps are going to be hitting just as I think you know all of the
[21:38] just as I think you know all of the silly skeptics start to understand that
[21:40] silly skeptics start to understand that orbital compute is very real like I
[21:42] orbital compute is very real like I think it's worth thinking long and hard
[21:44] think it's worth thinking long and hard about that if you're one of those
[21:46] about that if you're one of those companies and then all sorts of cool
[21:48] companies and then all sorts of cool stuff is happening in the interim you
[21:49] stuff is happening in the interim you know we're getting really good at
[21:51] know we're getting really good at repurposing jet engines. You know,
[21:53] repurposing jet engines. You know, there's that boom aerospace that is
[21:54] there's that boom aerospace that is doing this.
[21:55] doing this. >> So, there's a lot like capitalism is
[21:57] >> So, there's a lot like capitalism is hard at work
[21:59] hard at work >> on on watts. On wafers, though, it's
[22:03] >> on on watts. On wafers, though, it's just this group of, you know, flinty
[22:08] just this group of, you know, flinty older humans in Taiwan who are the most
[22:11] older humans in Taiwan who are the most important humans in Taiwan. They are the
[22:14] important humans in Taiwan. They are the overwhelming fraction of the country's
[22:15] overwhelming fraction of the country's GDP, water usage, electricity usage.
[22:18] GDP, water usage, electricity usage. They talk about the Silicon Shield. They
[22:21] They talk about the Silicon Shield. They all view themselves as inheritors of,
[22:25] all view themselves as inheritors of, you know, Morris Chain's sacred legacy.
[22:27] you know, Morris Chain's sacred legacy. I vividly remember like visiting Science
[22:29] I vividly remember like visiting Science Park more than 20 years ago and, you
[22:34] Park more than 20 years ago and, you know, talking to them. Do you think you
[22:35] know, talking to them. Do you think you could catch Intel? And they said, "This
[22:38] could catch Intel? And they said, "This is such a beautiful dream, but it's a
[22:40] is such a beautiful dream, but it's a dream for our grandchildren."
[22:43] dream for our grandchildren." >> And they did it. partly because of
[22:45] >> And they did it. partly because of Intel's self-inflicted wounds, but just
[22:48] Intel's self-inflicted wounds, but just they don't they think very differently.
[22:51] they don't they think very differently. You know, one reason, you know, Jensen
[22:53] You know, one reason, you know, Jensen flies over there so much is he wants
[22:55] flies over there so much is he wants them to expand capacity. I do think it's
[22:57] them to expand capacity. I do think it's wild that Jensen has never had a
[22:59] wild that Jensen has never had a contract with Taiwan Semi. They do
[23:01] contract with Taiwan Semi. They do business on what seems fair in
[23:03] business on what seems fair in handshakes. Just fascinating. No
[23:05] handshakes. Just fascinating. No contract. It's going to be fair over
[23:07] contract. It's going to be fair over time. We're partners. We're going to be
[23:09] time. We're partners. We're going to be fair to each other. And the truth is,
[23:11] fair to each other. And the truth is, you know, based on every every prior
[23:15] you know, based on every every prior market precedent for a foundational new
[23:17] market precedent for a foundational new technology like AI, you've always had a
[23:19] technology like AI, you've always had a bubble. You know, Carleta Perez wrote
[23:21] bubble. You know, Carleta Perez wrote this great book about this. And
[23:23] this great book about this. And basically, markets are efficient. They
[23:25] basically, markets are efficient. They correctly understand that this is a
[23:27] correctly understand that this is a foundational technology.
[23:29] foundational technology. There's what Mobison calls a breakdown
[23:31] There's what Mobison calls a breakdown in diversity.
[23:33] in diversity. Everyone becomes bullish on this new
[23:35] Everyone becomes bullish on this new technology. And I am beginning to worry
[23:38] technology. And I am beginning to worry a little bit about a diversity
[23:39] a little bit about a diversity breakdown. And then you get a bubble.
[23:43] breakdown. And then you get a bubble. That bubble funds the buildout of this
[23:45] That bubble funds the buildout of this new technology, but supply gets ahead of
[23:48] new technology, but supply gets ahead of demand. And you get a crash and it's a
[23:51] demand. And you get a crash and it's a particularly severe crash if it's a
[23:53] particularly severe crash if it's a debtfueled buildout like the year 2000.
[23:56] debtfueled buildout like the year 2000. And one thing really happy about really
[23:59] And one thing really happy about really good about the current buildout is it's
[24:00] good about the current buildout is it's still overwhelmingly funded out of
[24:02] still overwhelmingly funded out of operating cash flows which is a a really
[24:05] operating cash flows which is a a really important fundamental difference versus
[24:06] important fundamental difference versus the year 2000 has is valuation has is
[24:09] the year 2000 has is valuation has is the fact that every GPU is running at
[24:11] the fact that every GPU is running at 100% utilization when 99% of fiber was
[24:14] 100% utilization when 99% of fiber was unutilized. So there's all these
[24:15] unutilized. So there's all these fundamental differences, but we do have
[24:17] fundamental differences, but we do have to history doesn't repeat, but it rhymes
[24:19] to history doesn't repeat, but it rhymes and and as investor, we have to be very
[24:21] and and as investor, we have to be very cognizant of it
[24:23] cognizant of it and recognize that based on the last two
[24:26] and recognize that based on the last two or 30 hundred years, you know, forget
[24:28] or 30 hundred years, you know, forget the internet bubble. We had a railroad
[24:29] the internet bubble. We had a railroad bubble, a canal bubble, we should expect
[24:31] bubble, a canal bubble, we should expect a bubble. And that's terrifying. Like
[24:35] a bubble. And that's terrifying. Like nobody wants a bubble. A bubble is
[24:37] nobody wants a bubble. A bubble is terrible. reason it's terrible is if
[24:39] terrible. reason it's terrible is if you're valuation sensitive, you like
[24:41] you're valuation sensitive, you like massively underperform. You get fired by
[24:44] massively underperform. You get fired by probably all your clients. George
[24:46] probably all your clients. George Vanderhiden, who um is is is no longer
[24:49] Vanderhiden, who um is is is no longer with us, great port uh fidelity
[24:51] with us, great port uh fidelity portfolio manager, he fought the bubble
[24:54] portfolio manager, he fought the bubble in 99 and he retired in two in early
[24:58] in 99 and he retired in two in early 2000 because I think he just couldn't
[24:59] 2000 because I think he just couldn't couldn't take it.
[25:00] couldn't take it. >> He knew it was wrong and you know his
[25:03] >> He knew it was wrong and you know his his clients were deeply skeptical.
[25:05] his clients were deeply skeptical. George, you're out of step. you know, he
[25:07] George, you're out of step. you know, he had he had white hair. He's truly great
[25:09] had he had white hair. He's truly great man. I I only overlapped with him
[25:11] man. I I only overlapped with him briefly, but he was a very important
[25:13] briefly, but he was a very important mentor and friend to my good friend and
[25:16] mentor and friend to my good friend and mentor Jennifer Urick. So, I have a lot
[25:18] mentor Jennifer Urick. So, I have a lot of Vanderhiden DNA through her. He was
[25:21] of Vanderhiden DNA through her. He was the same person who said being early is
[25:23] the same person who said being early is the same thing as being wrong. George
[25:24] the same thing as being wrong. George retires because he can't take the
[25:27] retires because he can't take the underperformance and he can't take
[25:29] underperformance and he can't take clients saying what's wrong with you?
[25:31] clients saying what's wrong with you? You don't get it. and he has like 40% of
[25:34] You don't get it. and he has like 40% of his funded tobacco, 40% did homebuilders
[25:38] his funded tobacco, 40% did homebuilders and literally he underperfor he probably
[25:40] and literally he underperfor he probably outperformed the NASDAQ
[25:43] outperformed the NASDAQ by like 20 or 30x over the next three
[25:47] by like 20 or 30x over the next three years. Okay. And I have been optimistic
[25:50] years. Okay. And I have been optimistic that this fundamental shortage of wafers
[25:53] that this fundamental shortage of wafers which really today is controlled by
[25:56] which really today is controlled by Taiwan Semi will prevent one. If Taiwan
[25:58] Taiwan Semi will prevent one. If Taiwan Semi did what Jensen wanted, I think
[26:00] Semi did what Jensen wanted, I think Nvidia could sell two trillion dollars
[26:02] Nvidia could sell two trillion dollars of GPUs in 26 in 26 or 27, maybe two.5
[26:07] of GPUs in 26 in 26 or 27, maybe two.5 trillion, maybe three trillion, but
[26:09] trillion, maybe three trillion, but there is a limit where consumers would
[26:11] there is a limit where consumers would consume so much that you probably would
[26:14] consume so much that you probably would be in an overbuild. And so Taiwan Smi,
[26:17] be in an overbuild. And so Taiwan Smi, if we don't get a bubble, like we need
[26:18] if we don't get a bubble, like we need to throw a party for them because they
[26:20] to throw a party for them because they will have single-handedly prevented a
[26:22] will have single-handedly prevented a bubble. Okay, you are starting to see
[26:25] bubble. Okay, you are starting to see companies go to Intel
[26:28] companies go to Intel and Samsung.
[26:29] and Samsung. >> Let's just assume TSM stays super supply
[26:31] >> Let's just assume TSM stays super supply constraint versus you know the latent
[26:33] constraint versus you know the latent demand like what what happens?
[26:35] demand like what what happens? >> Well, one of you know the history
[26:38] >> Well, one of you know the history markets is I don't know who but one of
[26:40] markets is I don't know who but one of Intel and Samsung they're not going to
[26:42] Intel and Samsung they're not going to stay disciplined. They will break and
[26:44] stay disciplined. They will break and then at some level that will force
[26:47] then at some level that will force everyone else to break.
[26:50] everyone else to break. So like I think a lot of this may come
[26:52] So like I think a lot of this may come down to the degree to which Taiwan SIM
[26:55] down to the degree to which Taiwan SIM can maintain a lead over Intel and
[26:59] can maintain a lead over Intel and Samsung. You got to remember it's
[27:00] Samsung. You got to remember it's whatever it is it's 9 12 15 months.
[27:02] whatever it is it's 9 12 15 months. >> Sort of like the leading node edge. You
[27:04] >> Sort of like the leading node edge. You mean
[27:04] mean >> exactly you know the pace at which they
[27:07] >> exactly you know the pace at which they expand capacity. Like if I were to watch
[27:10] expand capacity. Like if I were to watch one thing to understand whe there's a
[27:11] one thing to understand whe there's a bubble it's Taiwan Simmy's capacity
[27:13] bubble it's Taiwan Simmy's capacity decisions. And I think there's a
[27:15] decisions. And I think there's a Goldilocks zone where they expand enough
[27:21] Goldilocks zone where they expand enough they make it hard for Intel or Samsung
[27:24] they make it hard for Intel or Samsung to really truly emerge as like a um at
[27:29] to really truly emerge as like a um at scale second source with something you
[27:31] scale second source with something you know well north of 30% market share. And
[27:35] know well north of 30% market share. And yet they also keep this fundamental
[27:38] yet they also keep this fundamental constraint on wafers
[27:41] constraint on wafers that you know helps us avoid a bubble.
[27:44] that you know helps us avoid a bubble. And then obviously I think the terapab
[27:47] And then obviously I think the terapab um is going to play into this too.
[27:48] um is going to play into this too. >> Say more about that for people that
[27:51] >> Say more about that for people that >> the turfab it's a SpaceX I believe
[27:53] >> the turfab it's a SpaceX I believe Tesla's involved as well um joint
[27:56] Tesla's involved as well um joint venture to build the world's largest fab
[27:59] venture to build the world's largest fab here in America and I'm I think they're
[28:03] here in America and I'm I think they're going to be successful. on they have a
[28:05] going to be successful. on they have a partnership with Intel which is very
[28:06] partnership with Intel which is very important um because they're getting
[28:08] important um because they're getting access to 50 years of institutional
[28:11] access to 50 years of institutional knowledge that's just you know a nine
[28:13] knowledge that's just you know a nine months a few quarters 12 months 3 to
[28:16] months a few quarters 12 months 3 to five quarters behind the front that's an
[28:17] five quarters behind the front that's an advantage it's also an advantage that I
[28:21] advantage it's also an advantage that I believe that terafab is going to get
[28:23] believe that terafab is going to get attention from the a teams at all the
[28:25] attention from the a teams at all the semicap equipment companies like one big
[28:27] semicap equipment companies like one big reason Taiwan semicought up is ASML and
[28:31] reason Taiwan semicought up is ASML and KA tenor and lamb research and applied
[28:33] KA tenor and lamb research and applied material materials. They wanted them to
[28:35] material materials. They wanted them to catch up. They didn't they don't like
[28:37] catch up. They didn't they don't like having a monopsiny and so the A teams
[28:40] having a monopsiny and so the A teams were in Taiwan working. Intel made some
[28:42] were in Taiwan working. Intel made some mistakes and presto. And so the A teams
[28:46] mistakes and presto. And so the A teams will will be here because of Elon's
[28:48] will will be here because of Elon's reputation in in hardware engineering.
[28:51] reputation in in hardware engineering. And then just to a degree that I think
[28:54] And then just to a degree that I think is u maybe hard for people to imagine in
[28:58] is u maybe hard for people to imagine in America um where you know politics has
[29:00] America um where you know politics has replaced religion because Elon had his
[29:02] replaced religion because Elon had his fora into politics that makes it hard
[29:04] fora into politics that makes it hard for some people in America to see him
[29:07] for some people in America to see him clearly which is sad because I do think
[29:10] clearly which is sad because I do think you know he's probably doing more for
[29:11] you know he's probably doing more for America than any other American. You
[29:14] America than any other American. You know he's single-handedly bringing
[29:16] know he's single-handedly bringing manufacturing back to America. He's
[29:18] manufacturing back to America. He's revived Dince Tech. SpaceX is in some
[29:21] revived Dince Tech. SpaceX is in some ways the most important defense
[29:22] ways the most important defense contractor in America. You know, what
[29:24] contractor in America. You know, what he's doing with Starlink is amazing for
[29:26] he's doing with Starlink is amazing for the world. He's creating all these blue
[29:29] the world. He's creating all these blue collar manufacturing jobs, which is like
[29:30] collar manufacturing jobs, which is like a goal, I think, of a lot of liberals
[29:32] a goal, I think, of a lot of liberals and good for America. He's done more
[29:34] and good for America. He's done more than any living human to decarbonize the
[29:37] than any living human to decarbonize the world. And if you are upset about data
[29:39] world. And if you are upset about data sitters on Earth for environmental
[29:41] sitters on Earth for environmental reasons, well, here you go. You know, so
[29:45] reasons, well, here you go. You know, so it's it's sad, but he is a living deity.
[29:49] it's it's sad, but he is a living deity. in China, Taiwan, South Korea, and
[29:53] in China, Taiwan, South Korea, and Japan.
[29:55] Japan. And having watched him for a long time,
[29:59] And having watched him for a long time, what he's going to do is they're going
[30:00] what he's going to do is they're going to recruit the best people because the
[30:04] to recruit the best people because the best engineers want to work for Elon,
[30:08] best engineers want to work for Elon, especially in hardware engineering. He's
[30:10] especially in hardware engineering. He's going to recruit incredible engineers.
[30:12] going to recruit incredible engineers. And then they'll be next to next to
[30:14] And then they'll be next to next to Turfab, they'll be a Taiwan town. Oh,
[30:17] Turfab, they'll be a Taiwan town. Oh, these are your favorite restaurants. I'm
[30:19] these are your favorite restaurants. I'm going to move them and their whole staff
[30:21] going to move them and their whole staff from Taiwan to Texas and we're going to
[30:24] from Taiwan to Texas and we're going to make everything the way they like it.
[30:26] make everything the way they like it. And then we'll have Japan Town. Same
[30:28] And then we'll have Japan Town. Same thing. Then we're going to have Korea
[30:29] thing. Then we're going to have Korea Town. We're going to have all these
[30:30] Town. We're going to have all these things exactly but dialed to recruit the
[30:35] things exactly but dialed to recruit the best engineers. And that's just not the
[30:39] best engineers. And that's just not the way that the people who run Intel at
[30:42] way that the people who run Intel at Seung think. So he's going to have the
[30:44] Seung think. So he's going to have the best talent. He's going to have the A
[30:46] best talent. He's going to have the A teams at the wafer fab equipment
[30:48] teams at the wafer fab equipment companies. He's he has intel which is
[30:51] companies. He's he has intel which is important. It's so good for all of any
[30:54] important. It's so good for all of any administration's political goals. And I
[30:56] administration's political goals. And I think it's different enough that it will
[30:58] think it's different enough that it will not alienate Taiwan SMI.
[31:00] not alienate Taiwan SMI. >> And these have long lead times, right?
[31:02] >> And these have long lead times, right? So like Terrafab is going to be pumping
[31:04] So like Terrafab is going to be pumping out Nvidia G or whatever GPUs, whatever
[31:06] out Nvidia G or whatever GPUs, whatever chips like quite quite a long time from
[31:08] chips like quite quite a long time from now.
[31:09] now. >> Elon tends to do things differently.
[31:10] >> Elon tends to do things differently. Everybody else has taken three years to
[31:12] Everybody else has taken three years to build a data center. He built one in 122
[31:14] build a data center. He built one in 122 days. You know, Samsung had to give him
[31:17] days. You know, Samsung had to give him an office in their fab in Texas because
[31:20] an office in their fab in Texas because he was so unhappy about like the pace at
[31:22] he was so unhappy about like the pace at which they're expanding a building.
[31:24] which they're expanding a building. We'll see. Are you surprised by you
[31:27] We'll see. Are you surprised by you mentioned Deep Seek earlier? The simple
[31:29] mentioned Deep Seek earlier? The simple reaction to that was okay, these models
[31:31] reaction to that was okay, these models are just going to get 95% as effective
[31:34] are just going to get 95% as effective for some tiny fraction of the cost to
[31:36] for some tiny fraction of the cost to still Chinese open source models. Like
[31:38] still Chinese open source models. Like we'll be able to use these for most of
[31:39] we'll be able to use these for most of what we want to do. Fast forwarded a
[31:41] what we want to do. Fast forwarded a little bit of time, you know, two years
[31:43] little bit of time, you know, two years from now, there's no reason I have to
[31:45] from now, there's no reason I have to spend a million dollars a year in my
[31:46] spend a million dollars a year in my small little firm on on tokens or
[31:48] small little firm on on tokens or something. But then the actual reality
[31:50] something. But then the actual reality seems quite different than this. And I'm
[31:52] seems quite different than this. And I'm curious why there's that dissonance in
[31:54] curious why there's that dissonance in your mind.
[31:54] your mind. >> I do think it's the fascinating the
[31:56] >> I do think it's the fascinating the returns to the frontier, all the
[31:58] returns to the frontier, all the economic returns to AI at the model
[32:01] economic returns to AI at the model layer, not all of them, but an
[32:04] layer, not all of them, but an overwhelming amount of them have been at
[32:05] overwhelming amount of them have been at the frontier, which is surprising to me.
[32:09] the frontier, which is surprising to me. And I think it's been surprising to a
[32:11] And I think it's been surprising to a lot of people and I think this is one of
[32:15] lot of people and I think this is one of the most important questions to be
[32:18] the most important questions to be answered and you need to have a
[32:19] answered and you need to have a hypothesis on it as an investor. Are
[32:21] hypothesis on it as an investor. Are frontier tokens going to continue
[32:24] frontier tokens going to continue capturing the overwhelming majority of
[32:27] capturing the overwhelming majority of economic value created at the model
[32:29] economic value created at the model layer? And it is surprising like I just
[32:31] layer? And it is surprising like I just I remember when Gemini 3.1 Pro came out
[32:35] I remember when Gemini 3.1 Pro came out and it was it was mind-blowing to me. It
[32:37] and it was it was mind-blowing to me. It was so good. And today it's intolerable.
[32:42] was so good. And today it's intolerable. >> Intolerable.
[32:43] >> Intolerable. And you know there's probably a little
[32:45] And you know there's probably a little bit of a dynamic where companies
[32:46] bit of a dynamic where companies prototype with Frontiers then when they
[32:48] prototype with Frontiers then when they put something into production you're
[32:50] put something into production you're hearing a lot of people do use Vertex or
[32:52] hearing a lot of people do use Vertex or you know open source. But still it is it
[32:56] you know open source. But still it is it is a fact today that the overwhelming
[32:58] is a fact today that the overwhelming majority of these economic turns come
[33:00] majority of these economic turns come from Frontier tokens. And that's
[33:02] from Frontier tokens. And that's surprising and whether or not it
[33:04] surprising and whether or not it continues I think is a very interesting
[33:07] continues I think is a very interesting question. And I'm much more open-minded
[33:09] question. And I'm much more open-minded to that having had the experience I've
[33:11] to that having had the experience I've had with Gemini 3.1
[33:14] had with Gemini 3.1 and then Opus. Um, and then I do use Gro
[33:18] and then Opus. Um, and then I do use Gro 4.3. It is on the paro frontier. like
[33:21] 4.3. It is on the paro frontier. like the companies that are on the paro
[33:22] the companies that are on the paro frontier are and this is by the way a
[33:24] frontier are and this is by the way a big change in a a consequence of what we
[33:26] big change in a a consequence of what we talked about last time. Google losing
[33:28] talked about last time. Google losing their percost token leadership as a
[33:31] their percost token leadership as a result of making very conservative
[33:32] result of making very conservative design decisions with TPU V8 to try and
[33:35] design decisions with TPU V8 to try and take it away partially from Broadcom and
[33:37] take it away partially from Broadcom and Nvidia um continuing to make aggressive
[33:40] Nvidia um continuing to make aggressive choices. Uh but Google dominated the
[33:43] choices. Uh but Google dominated the prao frontier. The prao frontier being
[33:45] prao frontier. The prao frontier being intelligence versus cost. And I think
[33:47] intelligence versus cost. And I think this is the most important thing to look
[33:48] this is the most important thing to look at to analyze AI labs. Google dominated
[33:51] at to analyze AI labs. Google dominated that nine months ago. They at every
[33:53] that nine months ago. They at every point on the paro frontier. OpenAI, XAI
[33:58] point on the paro frontier. OpenAI, XAI and Anthropic were inside of them. Now
[34:01] and Anthropic were inside of them. Now the Paro frontier is dominated by
[34:03] the Paro frontier is dominated by Enthropic, OpenAI. And then Grock 4.3 is
[34:06] Enthropic, OpenAI. And then Grock 4.3 is on the paro frontier. It's clearly like
[34:09] on the paro frontier. It's clearly like the, you know, the best lowest cost 500
[34:11] the, you know, the best lowest cost 500 billion parameter model. And then Gemini
[34:14] billion parameter model. And then Gemini 3.1 is like hanging on to the paro
[34:17] 3.1 is like hanging on to the paro frontier. And if I were to bet or bet
[34:19] frontier. And if I were to bet or bet that they're subsidizing that out of
[34:20] that they're subsidizing that out of pride, I would just say one a violation
[34:23] pride, I would just say one a violation of Richard Sutton's bitter lesson is for
[34:25] of Richard Sutton's bitter lesson is for sure the biggest risk to this trade
[34:27] sure the biggest risk to this trade >> to all of AI. Now the closer someone is
[34:30] >> to all of AI. Now the closer someone is to AI, the more skeptical they are this
[34:32] to AI, the more skeptical they are this will occur. One thing I think
[34:34] will occur. One thing I think contributed to weakness in March was,
[34:36] contributed to weakness in March was, you know, a much more stupid version of
[34:39] you know, a much more stupid version of DeepSeek, which was this thing called
[34:40] DeepSeek, which was this thing called Turboquant. and Turbo Quan is some
[34:42] Turboquant. and Turbo Quan is some Google memory optimization that was
[34:44] Google memory optimization that was written up in a paper a year ago. And
[34:46] written up in a paper a year ago. And then during the middle of an agreement
[34:49] then during the middle of an agreement while Google was negotiating with
[34:51] while Google was negotiating with Micron, Samsung and Highex to sign, you
[34:53] Micron, Samsung and Highex to sign, you know, some LTA that would lock in really
[34:55] know, some LTA that would lock in really high prices for a long time. They
[34:57] high prices for a long time. They released this. You know, what people do
[34:59] released this. You know, what people do is always more important than they say.
[35:00] is always more important than they say. And they just kind of publicize it on X
[35:03] And they just kind of publicize it on X and it goes viral like, "Oh my god, DRM
[35:06] and it goes viral like, "Oh my god, DRM is cooked. Here's this DRAM
[35:08] is cooked. Here's this DRAM optimization." I was unable to find a
[35:10] optimization." I was unable to find a single AI engineer on planet earth who
[35:13] single AI engineer on planet earth who believed that turbo quant would have any
[35:15] believed that turbo quant would have any impact on DRAM demand but nonetheless a
[35:18] impact on DRAM demand but nonetheless a violation of Richard Sutton's bitter
[35:20] violation of Richard Sutton's bitter lesson you know more compute will always
[35:21] lesson you know more compute will always outperform human algorithmic ingenuity
[35:23] outperform human algorithmic ingenuity more computing data or chin beyond
[35:25] more computing data or chin beyond chinchilla optimal I guess what what
[35:27] chinchilla optimal I guess what what people increasingly do today that's a
[35:30] people increasingly do today that's a real risk man and I think the people who
[35:33] real risk man and I think the people who are building these models are skeptical
[35:35] are building these models are skeptical of that risk the reason I am a little
[35:38] of that risk the reason I am a little less skep skeptical is I think we are
[35:40] less skep skeptical is I think we are very close to ASI and who knows if the
[35:43] very close to ASI and who knows if the bitter lesson holds for 400 IQ models
[35:47] bitter lesson holds for 400 IQ models just you know or maybe we get a
[35:49] just you know or maybe we get a temporary
[35:51] temporary period where these you know if you get
[35:52] period where these you know if you get to ASI the first thing it wants is
[35:55] to ASI the first thing it wants is probably to be smarter and have more
[35:57] probably to be smarter and have more resources. How does it do that? It makes
[35:58] resources. How does it do that? It makes itself more efficient. I think that is
[36:02] itself more efficient. I think that is an actual risk that humans the bitter
[36:06] an actual risk that humans the bitter lesson literally I believe includes
[36:08] lesson literally I believe includes humans in it. So we're about to find out
[36:11] humans in it. So we're about to find out whether the bitter lesson we'll find out
[36:12] whether the bitter lesson we'll find out if it applies to 300 IQ ahis then 400
[36:16] if it applies to 300 IQ ahis then 400 then 500 and 600 and at some point we
[36:20] then 500 and 600 and at some point we may have like a temporary violation of
[36:22] may have like a temporary violation of the bitter lesson based upon AI and ASI.
[36:27] the bitter lesson based upon AI and ASI. So I'm curious how you think about some
[36:29] So I'm curious how you think about some other parts of the innovation around the
[36:32] other parts of the innovation around the model continual learning and memory
[36:34] model continual learning and memory being two that see people seem to be
[36:36] being two that see people seem to be most focused on as things that might
[36:37] most focused on as things that might create yet another you know new paradigm
[36:39] create yet another you know new paradigm that we would enter. What do you think
[36:40] that we would enter. What do you think about the role of those two things?
[36:42] about the role of those two things? Yeah. Well, I think we've done a lot
[36:43] Yeah. Well, I think we've done a lot with memory through these harnesses. And
[36:46] with memory through these harnesses. And it turns out that harness engineering is
[36:49] it turns out that harness engineering is not as important as the model, but it
[36:52] not as important as the model, but it really matters. And these harnesses in
[36:54] really matters. And these harnesses in these models are increasingly being
[36:56] these models are increasingly being co-developed. One of the big things a
[36:58] co-developed. One of the big things a harness does, which you just think of as
[36:59] harness does, which you just think of as like a a runtime that the model operates
[37:03] like a a runtime that the model operates in, knows where the pool tools are. It
[37:06] in, knows where the pool tools are. It like creates context, memory, state, um,
[37:11] like creates context, memory, state, um, you know, has very specific,
[37:13] you know, has very specific, you know, prompts or instructions and
[37:16] you know, prompts or instructions and just
[37:16] just >> makes a huge difference. Even simple
[37:18] >> makes a huge difference. Even simple versions,
[37:19] versions, >> it makes an incredible difference. I
[37:20] >> it makes an incredible difference. I think the last time I was on here or one
[37:22] think the last time I was on here or one of the other times I just said like,
[37:23] of the other times I just said like, "Hey, as an investor, it's very
[37:26] "Hey, as an investor, it's very important that you pay for the $250 a
[37:29] important that you pay for the $250 a month version to get like your own
[37:31] month version to get like your own intuitive sense." that's no longer
[37:33] intuitive sense." that's no longer possible to understand what frontier AI
[37:35] possible to understand what frontier AI is capable of today even for like a
[37:39] is capable of today even for like a non-coding use case you need to have
[37:40] non-coding use case you need to have cloud code or codeex and you need to be
[37:43] cloud code or codeex and you need to be on an enterprise plan and the reason for
[37:45] on an enterprise plan and the reason for this is and this is another I think this
[37:49] this is and this is another I think this is another dynamic that's enabled by
[37:51] is another dynamic that's enabled by Google losing their cost leadership is
[37:54] Google losing their cost leadership is these AI models just shifted to
[37:56] these AI models just shifted to usagebased pricing and if you're on that
[37:59] usagebased pricing and if you're on that $250 or$300 or $280 month plan or
[38:01] $250 or$300 or $280 month plan or whatever it is you are getting severely
[38:04] whatever it is you are getting severely rate limited. You are getting a
[38:06] rate limited. You are getting a labbotomized version of the AI because
[38:09] labbotomized version of the AI because like we talked about Claude now produces
[38:11] like we talked about Claude now produces 70% less tokens. You want the tokens
[38:13] 70% less tokens. You want the tokens that Claude and its harness really think
[38:16] that Claude and its harness really think it needs to produce to get you a good
[38:17] it needs to produce to get you a good answer, you need to be on a usage based
[38:19] answer, you need to be on a usage based plan. And by the way, this is so bullish
[38:23] plan. And by the way, this is so bullish for AI. I was a telecom analyst in ' 05
[38:25] for AI. I was a telecom analyst in ' 05 to07 and cellular had been a great
[38:28] to07 and cellular had been a great growth industry really for the last 10
[38:30] growth industry really for the last 10 years and the reason was you had a
[38:32] years and the reason was you had a combination of fixed pricing you had 900
[38:35] combination of fixed pricing you had 900 minutes for whatever it was and then
[38:37] minutes for whatever it was and then usage based pricing over that and when
[38:40] usage based pricing over that and when did cellular stop being a great growth
[38:42] did cellular stop being a great growth industry when everybody just went to all
[38:44] industry when everybody just went to all you can eat. And and by the way long
[38:47] you can eat. And and by the way long distance was the same thing. AI is just
[38:48] distance was the same thing. AI is just shifting from all you can eat to pay by
[38:51] shifting from all you can eat to pay by the drink. And it turns out people
[38:52] the drink. And it turns out people really like to talk to their friends
[38:54] really like to talk to their friends long distance. They really like to talk
[38:56] long distance. They really like to talk to their friends on the phone. And
[38:57] to their friends on the phone. And people really like to use AI and
[39:00] people really like to use AI and particularly now that one person can
[39:02] particularly now that one person can have a 100 agents working. So I think
[39:03] have a 100 agents working. So I think the shift to usage based pricing is
[39:07] the shift to usage based pricing is probably why you will see OpenAI and
[39:10] probably why you will see OpenAI and Anthropic exceed well over $200 billion
[39:13] Anthropic exceed well over $200 billion in ARR this year. because not only is
[39:15] in ARR this year. because not only is more compute going to become online, but
[39:17] more compute going to become online, but they're going to be able to push
[39:19] they're going to be able to push frontier token pricing with these usage
[39:21] frontier token pricing with these usage enterprise models, but it's it's sad.
[39:24] enterprise models, but it's it's sad. It's sad for the world and because it
[39:26] It's sad for the world and because it just means if you can't afford that,
[39:27] just means if you can't afford that, you're not at the frontier. But yeah,
[39:30] you're not at the frontier. But yeah, continual learning, man, I mean, if we
[39:32] continual learning, man, I mean, if we solve that,
[39:32] solve that, >> how do you conceptualize that? Like
[39:34] >> how do you conceptualize that? Like >> there's so many mysteries about the
[39:35] >> there's so many mysteries about the human mind, like we're such sample
[39:38] human mind, like we're such sample efficient learners relative to AI. Like
[39:42] efficient learners relative to AI. Like I forget what it is, but like an AI
[39:44] I forget what it is, but like an AI needs orders of magnitude.
[39:45] needs orders of magnitude. >> Yeah. Many orders of magnitude. Now we
[39:47] >> Yeah. Many orders of magnitude. Now we have a crude variant of continual
[39:49] have a crude variant of continual learning today when something is
[39:51] learning today when something is verifiable and that's just, you know,
[39:53] verifiable and that's just, you know, reinforcement learning during
[39:54] reinforcement learning during mid-training. But yeah, continual
[39:56] mid-training. But yeah, continual learning is a model that dynamically
[39:58] learning is a model that dynamically adjusts its weights or adjusts in some
[40:02] adjusts its weights or adjusts in some way in real time. Like as a human,
[40:05] way in real time. Like as a human, >> that's what you do.
[40:06] >> that's what you do. >> Yeah. Like if I the first time I touch
[40:08] >> Yeah. Like if I the first time I touch or you know put my hand in a fire, I've
[40:11] or you know put my hand in a fire, I've learned I never put it in there before.
[40:13] learned I never put it in there before. That model today needs to put its hand
[40:15] That model today needs to put its hand in the fire a million times and then
[40:18] in the fire a million times and then have, you know, the designers
[40:20] have, you know, the designers effectively put a fire in the next
[40:23] effectively put a fire in the next training run or an RL gym for it to
[40:26] training run or an RL gym for it to learn. I think it has to be dynamically
[40:29] learn. I think it has to be dynamically updating the weights, but I think people
[40:31] updating the weights, but I think people are working on really smart techniques
[40:33] are working on really smart techniques beyond this. But if we get that then we
[40:37] beyond this. But if we get that then we have a really fast takeoff and people
[40:40] have a really fast takeoff and people seem
[40:42] seem confident that continual learning is
[40:45] confident that continual learning is kind of just around the corner. And I do
[40:47] kind of just around the corner. And I do think this is like the third big
[40:50] think this is like the third big question. Bitter le violation as a
[40:52] question. Bitter le violation as a result of ASI or less likely human
[40:55] result of ASI or less likely human ingenuity. Will Frontier tokens still
[40:57] ingenuity. Will Frontier tokens still command the premium they do? And will we
[41:00] command the premium they do? And will we get continual learning? And if so, when?
[41:02] get continual learning? And if so, when? As your business scales up, everything
[41:04] As your business scales up, everything gets more complex, especially your
[41:05] gets more complex, especially your compliance and security needs. With so
[41:07] compliance and security needs. With so many tools offering band-aids and
[41:09] many tools offering band-aids and patches, it's unfortunately far too easy
[41:11] patches, it's unfortunately far too easy for something to slip through the
[41:12] for something to slip through the cracks. Fortunately, Vanta is a powerful
[41:14] cracks. Fortunately, Vanta is a powerful tool designed to simplify and automate
[41:16] tool designed to simplify and automate your security work and deliver a single
[41:18] your security work and deliver a single source of truth for compliance and risk.
[41:20] source of truth for compliance and risk. There's a reason that Ramp, Cursor, and
[41:22] There's a reason that Ramp, Cursor, and Snowflake all use Vanta. It frees them
[41:24] Snowflake all use Vanta. It frees them to focus on building amazing
[41:26] to focus on building amazing differentiated products knowing that
[41:27] differentiated products knowing that compliance and security are under
[41:29] compliance and security are under control. Invest like the best listeners.
[41:31] control. Invest like the best listeners. Get a special offer of $1,000 off Vanta
[41:33] Get a special offer of $1,000 off Vanta when you go to vanta.com/invest.
[41:36] when you go to vanta.com/invest. I know firsthand how complex the tech
[41:38] I know firsthand how complex the tech stack is for asset management firms. And
[41:40] stack is for asset management firms. And seemingly every new tool and data source
[41:42] seemingly every new tool and data source makes the problem even worse, adding
[41:44] makes the problem even worse, adding more complexity, more headcount, and
[41:46] more complexity, more headcount, and more risk. Ridgeline offers a better way
[41:48] more risk. Ridgeline offers a better way forward. One unified platform that
[41:49] forward. One unified platform that automates away that complexity across
[41:51] automates away that complexity across portfolio accounting, reconciliation,
[41:53] portfolio accounting, reconciliation, reporting, trading, compliance, and
[41:55] reporting, trading, compliance, and more. All at scale. Ridgeline is
[41:58] more. All at scale. Ridgeline is revolutionizing investment management.
[41:59] revolutionizing investment management. helping ambitious firms scale faster,
[42:01] helping ambitious firms scale faster, operate smarter, and stay ahead of the
[42:02] operate smarter, and stay ahead of the curve. See what Ridgeline can unlock for
[42:04] curve. See what Ridgeline can unlock for your firm. Schedule a demo at
[42:06] your firm. Schedule a demo at ridgeline.ai.
[42:07] ridgeline.ai. What is the role of new chip companies
[42:10] What is the role of new chip companies in all of this? Like we talked a lot
[42:11] in all of this? Like we talked a lot about Nvidia and you know their their
[42:13] about Nvidia and you know their their sort of relationship with TSMC and Intel
[42:15] sort of relationship with TSMC and Intel and all these sorts of things. There's a
[42:17] and all these sorts of things. There's a thousand flowers blooming. I think
[42:19] thousand flowers blooming. I think literally probably a thousand flowers
[42:20] literally probably a thousand flowers blooming trying to create a new chip to
[42:23] blooming trying to create a new chip to address some part of this bottleneck.
[42:26] address some part of this bottleneck. I'm curious how you process this space,
[42:28] I'm curious how you process this space, this opportunity, what role it will
[42:29] this opportunity, what role it will play, what role they'll play.
[42:31] play, what role they'll play. >> So, I think this is good and healthy for
[42:32] >> So, I think this is good and healthy for the world. It's good for Jensen too. Um,
[42:35] the world. It's good for Jensen too. Um, you know, because a different
[42:36] you know, because a different administration might take a different
[42:39] administration might take a different view. Competition, I think, is good for
[42:41] view. Competition, I think, is good for everyone. In in tank design, they talk
[42:42] everyone. In in tank design, they talk about the iron triangle. The iron
[42:44] about the iron triangle. The iron triangle of tank design is that all
[42:46] triangle of tank design is that all designers of a tank, they have to make
[42:48] designers of a tank, they have to make trade-offs between attack, defense, and
[42:50] trade-offs between attack, defense, and mobility. And you know, for obvious
[42:51] mobility. And you know, for obvious reasons, the more defense you have,
[42:53] reasons, the more defense you have, which is just armor, the heavier the
[42:55] which is just armor, the heavier the tank is, the less mobile it is. So you
[42:58] tank is, the less mobile it is. So you have to live in this triangle and make
[43:00] have to live in this triangle and make tradeoffs. Okay? Like the marava in
[43:03] tradeoffs. Okay? Like the marava in Israel, it's optimized for defense.
[43:05] Israel, it's optimized for defense. Russian tanks and like the Leopard are
[43:08] Russian tanks and like the Leopard are generally more optimized for mobility.
[43:10] generally more optimized for mobility. Chip design is the same. And you there
[43:13] Chip design is the same. And you there there are these fundamental constraints
[43:16] there are these fundamental constraints imposed by the laws of physics has
[43:19] imposed by the laws of physics has embedded in the Taiwan semi design rules
[43:22] embedded in the Taiwan semi design rules that you need to live within and you
[43:26] that you need to live within and you have TPU tranium and AMD which are all
[43:30] have TPU tranium and AMD which are all um you know essentially trying to be a
[43:34] um you know essentially trying to be a better GPU and today I think probably
[43:37] better GPU and today I think probably Tranium is doing the best. Nobody's a
[43:39] Tranium is doing the best. Nobody's a better GPU, but Trrenium is is I think
[43:42] better GPU, but Trrenium is is I think their, you know, they're they're tugging
[43:44] their, you know, they're they're tugging on Superman's cape
[43:46] on Superman's cape >> and and this hadn't started yet. The
[43:48] >> and and this hadn't started yet. The Tranium 3 needs to ramp into production
[43:50] Tranium 3 needs to ramp into production because it has a switch scale up
[43:52] because it has a switch scale up network, which you really need to
[43:53] network, which you really need to economically inference models. You know,
[43:56] economically inference models. You know, a lot of companies have a Taurus
[43:58] a lot of companies have a Taurus architecture. Um that that's where
[44:00] architecture. Um that that's where Google was and AMD. We'll see. The
[44:03] Google was and AMD. We'll see. The MI450, we don't know yet. We'll see. We
[44:06] MI450, we don't know yet. We'll see. We probably know more about Trinidium 3
[44:07] probably know more about Trinidium 3 than the MI450, but that's a hard game
[44:10] than the MI450, but that's a hard game to play. So you have to do something
[44:14] to play. So you have to do something different and you have to do something
[44:16] different and you have to do something different that is also hard to do. So I
[44:20] different that is also hard to do. So I think the best path for these startups
[44:22] think the best path for these startups like my rule of thumb is 1% market share
[44:24] like my rule of thumb is 1% market share is going to be worth 100 billion. 100
[44:26] is going to be worth 100 billion. 100 billion is a pretty good venture
[44:27] billion is a pretty good venture outcome. I think what Jensen would say
[44:29] outcome. I think what Jensen would say is like, okay, if something somebody
[44:30] is like, okay, if something somebody does something different and it gets to
[44:33] does something different and it gets to one or two or 3% share, we'll make that
[44:36] one or two or 3% share, we'll make that chip and that's that's coming for
[44:39] chip and that's that's coming for everyone. But if you're trying to make a
[44:41] everyone. But if you're trying to make a better GPU, good luck. If you were doing
[44:43] better GPU, good luck. If you were doing something different, it also needs to be
[44:47] something different, it also needs to be hard to do. And you can make different
[44:49] hard to do. And you can make different trade-offs. you know, the disagregation
[44:51] trade-offs. you know, the disagregation of prefill and inference really have
[44:53] of prefill and inference really have opened the aperture um for making these
[44:56] opened the aperture um for making these different trade-offs because you can
[44:58] different trade-offs because you can make very aggressive trade-offs for
[44:59] make very aggressive trade-offs for decode, aggressive trade-offs for
[45:01] decode, aggressive trade-offs for prefill.
[45:02] prefill. >> Prefill being taking in the context,
[45:03] >> Prefill being taking in the context, decode being, you know, write the
[45:05] decode being, you know, write the output.
[45:06] output. >> Yeah. I have a great colleague named
[45:07] >> Yeah. I have a great colleague named Andrew Fox who said, "Picture, you know,
[45:09] Andrew Fox who said, "Picture, you know, a British naval ship from the 18th
[45:11] a British naval ship from the 18th century. Prefill is loading the cannon,
[45:13] century. Prefill is loading the cannon, decode is firing it." And what prefill
[45:15] decode is firing it." And what prefill literally is is just the model
[45:17] literally is is just the model understanding the question, the prompt
[45:19] understanding the question, the prompt and then kind of keeping track of its
[45:20] and then kind of keeping track of its own dec.
[45:22] own dec. And that is fundamentally a memory
[45:24] And that is fundamentally a memory capacity bound problem. Decode is a
[45:27] capacity bound problem. Decode is a process of generating new tokens and
[45:28] process of generating new tokens and that is memory bandwidth constraint. And
[45:31] that is memory bandwidth constraint. And so if you're a chip designer, this gives
[45:33] so if you're a chip designer, this gives you a richer canvas to to paint on. But
[45:36] you a richer canvas to to paint on. But even so, it needs to be hard because if
[45:39] even so, it needs to be hard because if you make different trade-offs in that
[45:40] you make different trade-offs in that iron triangle to optimize for memory
[45:43] iron triangle to optimize for memory capacity, and they're not hard
[45:44] capacity, and they're not hard trade-offs to make, well then Nvidia is
[45:47] trade-offs to make, well then Nvidia is going to make those same trade-offs.
[45:49] going to make those same trade-offs. They get better prices from Taiwan Semi
[45:52] They get better prices from Taiwan Semi than you're ever going to get. Um, and
[45:54] than you're ever going to get. Um, and good luck. Good luck. And they have the
[45:56] good luck. Good luck. And they have the advantage of working with every model
[45:58] advantage of working with every model company and optimizing their designs.
[46:00] company and optimizing their designs. And by the way, another very funny thing
[46:02] And by the way, another very funny thing is if you're a VC
[46:04] is if you're a VC and you're investing in semiconductor
[46:06] and you're investing in semiconductor company that is telling you they are
[46:08] company that is telling you they are going to have an advantage because of a
[46:10] going to have an advantage because of a Taiwan semi process that they have
[46:12] Taiwan semi process that they have special access to. I promise you that
[46:15] special access to. I promise you that Jensen saw that process when it was a
[46:19] Jensen saw that process when it was a twinkle in Taiwan Simmy's eyes and it
[46:22] twinkle in Taiwan Simmy's eyes and it they know more about it than this little
[46:25] they know more about it than this little company with 200 people can imagine.
[46:28] company with 200 people can imagine. Taiwan, CMI, everybody in the supply
[46:30] Taiwan, CMI, everybody in the supply chain is showing Jensen everything the
[46:32] chain is showing Jensen everything the same way they're showing Amazon
[46:34] same way they're showing Amazon everything, AMD everything, TPU
[46:37] everything, AMD everything, TPU everything. And that's another reason
[46:38] everything. And that's another reason don't go try to make a better GPU. So
[46:41] don't go try to make a better GPU. So you can do something different. You can
[46:42] you can do something different. You can paint in the pre-filled canvas. You can
[46:44] paint in the pre-filled canvas. You can paint in the decode canvas, but you also
[46:46] paint in the decode canvas, but you also have to do something hard because if it
[46:48] have to do something hard because if it gets to scale, you're going to have
[46:51] gets to scale, you're going to have those four companies has very fast
[46:53] those four companies has very fast followers. My firm was a was a um
[46:56] followers. My firm was a was a um venture investor in Cerebras. What
[46:59] venture investor in Cerebras. What Cerebrris has done is something hard and
[47:00] Cerebrris has done is something hard and fundamentally different way for scale
[47:02] fundamentally different way for scale computing. And it it comes with a set of
[47:05] computing. And it it comes with a set of trade-offs, but that architectural
[47:08] trade-offs, but that architectural decision they made was hard and lets
[47:11] decision they made was hard and lets them do something that no one else can
[47:13] them do something that no one else can do. And we'll find out how big that is.
[47:17] do. And we'll find out how big that is. And you know, they're working on really
[47:18] And you know, they're working on really cool things like um one of the problems
[47:21] cool things like um one of the problems Cerebras has. Once you start needing to
[47:22] Cerebras has. Once you start needing to glue a lot of chips together and scale
[47:24] glue a lot of chips together and scale up networks or scale out networks, you
[47:27] up networks or scale out networks, you need a lot of IO and IO is bound by
[47:31] need a lot of IO and IO is bound by what's called the shoreline, the sides
[47:32] what's called the shoreline, the sides of the chip. And so Cerebris has an
[47:35] of the chip. And so Cerebris has an overwhelming ratio of onchip computed
[47:37] overwhelming ratio of onchip computed memory relative to shoreline IO. Well,
[47:40] memory relative to shoreline IO. Well, they're really smart people. They did
[47:41] they're really smart people. They did something really hard. They're trying to
[47:43] something really hard. They're trying to see if they can put an optical wafer
[47:45] see if they can put an optical wafer right on top of that and then that
[47:46] right on top of that and then that solves that problem. Um, I'm sure
[47:48] solves that problem. Um, I'm sure they're looking at hybrid bonding of
[47:50] they're looking at hybrid bonding of DRAM, you know, to get around these
[47:52] DRAM, you know, to get around these alleged limitations that are not true. A
[47:54] alleged limitations that are not true. A cerebrus machine can theoretically run
[47:56] cerebrus machine can theoretically run any size model. There are sizes of
[47:58] any size model. There are sizes of models where they're much better than
[47:59] models where they're much better than other sizes. So, Cerebras, what I think
[48:02] other sizes. So, Cerebras, what I think is interesting is they did something
[48:03] is interesting is they did something different that's hard to do, really hard
[48:05] different that's hard to do, really hard to do wafer scale computing. So, I do
[48:08] to do wafer scale computing. So, I do think there's a role for these and you
[48:10] think there's a role for these and you know, I would just encourage them all
[48:12] know, I would just encourage them all make a different trade-off
[48:14] make a different trade-off and try and do something hard. because
[48:18] and try and do something hard. because everybody's going to get funded after
[48:20] everybody's going to get funded after the Cerebrus IPO. It's not going to be a
[48:22] the Cerebrus IPO. It's not going to be a problem. But it took it took Cerebras
[48:25] problem. But it took it took Cerebras three generations of chips to get it
[48:28] three generations of chips to get it right. And it's really hard. Like Andrew
[48:31] right. And it's really hard. Like Andrew Feldman, the CEO, you can just see
[48:36] Feldman, the CEO, you can just see >> how hard it was
[48:39] >> how hard it was and that whole team did
[48:42] and that whole team did >> to get where they are today. And they
[48:44] >> to get where they are today. And they need to have the grit to do that, the
[48:45] need to have the grit to do that, the resilience. This first chip is a
[48:47] resilience. This first chip is a failure. It happens. Can you come back
[48:48] failure. It happens. Can you come back and make a second chip? But the one last
[48:50] and make a second chip? But the one last thing on this topic, this is going to be
[48:52] thing on this topic, this is going to be amazing for the useful lives of GPUs and
[48:55] amazing for the useful lives of GPUs and may single-handedly save private credit.
[48:57] may single-handedly save private credit. >> Say more about that. What what do you
[48:58] >> Say more about that. What what do you mean by the private credit? Well, just,
[49:00] mean by the private credit? Well, just, you know, private credit, they're in
[49:01] you know, private credit, they're in pain from these SAS loans and however
[49:03] pain from these SAS loans and however much they're marked down, they probably
[49:04] much they're marked down, they probably need to be marked down more because if
[49:06] need to be marked down more because if the public companies are struggling to
[49:07] the public companies are struggling to adapt, how's like a debtleaden company
[49:10] adapt, how's like a debtleaden company going going to adapt um and invest in
[49:13] going going to adapt um and invest in what is a very different margin
[49:15] what is a very different margin structure business? But there's a lot of
[49:17] structure business? But there's a lot of private credit and GPUs too. They were
[49:19] private credit and GPUs too. They were underwriting that to I think three or
[49:21] underwriting that to I think three or four years. And the disagregation of
[49:24] four years. And the disagregation of inference means that I think these GPUs
[49:28] inference means that I think these GPUs are going to have 10 or 15 year lives.
[49:30] are going to have 10 or 15 year lives. The AI skeptics are like, oh, these
[49:32] The AI skeptics are like, oh, these companies are all cooking their books.
[49:33] companies are all cooking their books. You know, the useful life of GU GPU is
[49:35] You know, the useful life of GU GPU is only a year or two. The useful life of a
[49:37] only a year or two. The useful life of a CPU is only four years because the rapid
[49:39] CPU is only four years because the rapid technological change. No. What rapid
[49:42] technological change. No. What rapid technological change has done with the
[49:44] technological change has done with the disagregation of pre-fill and inference
[49:46] disagregation of pre-fill and inference is mean that you you know you can put a
[49:48] is mean that you you know you can put a cerebra system or grock LPUs that Nvidia
[49:51] cerebra system or grock LPUs that Nvidia acquired effectively in front of a
[49:53] acquired effectively in front of a hopper or even an ampier use that hopper
[49:56] hopper or even an ampier use that hopper and ampear for prefill and extend the
[49:58] and ampear for prefill and extend the useful life of that GPU
[50:00] useful life of that GPU until it melts. Now they do melts they
[50:02] until it melts. Now they do melts they do melt so they have a time but you know
[50:04] do melt so they have a time but you know maybe you you don't have to run them as
[50:06] maybe you you don't have to run them as fast. This is going to be really good
[50:08] fast. This is going to be really good for the whole private credit industry.
[50:10] for the whole private credit industry. It's going to help finance the AI
[50:11] It's going to help finance the AI buildout because if you can start to
[50:13] buildout because if you can start to finance GPUs at more like you know 5% or
[50:17] finance GPUs at more like you know 5% or 6% instead of I think Corw's lowest
[50:19] 6% instead of I think Corw's lowest financing was like low sevens that
[50:21] financing was like low sevens that actually mathematically changes the cost
[50:23] actually mathematically changes the cost to finance this buildout. We had this
[50:25] to finance this buildout. We had this technological innovation that it's going
[50:27] technological innovation that it's going to lower the cost of financing extend
[50:29] to lower the cost of financing extend the useful life of compute on Earth. And
[50:31] the useful life of compute on Earth. And then I do think the one last thing
[50:32] then I do think the one last thing that's interesting about that is um my
[50:35] that's interesting about that is um my friend Jamon from Kotu just did a
[50:37] friend Jamon from Kotu just did a podcast and Cotu had a deck and they
[50:39] podcast and Cotu had a deck and they talked about hey you know the sellers of
[50:42] talked about hey you know the sellers of shortage are doing so much better than
[50:43] shortage are doing so much better than the buyers of shortage. Buyers of
[50:44] the buyers of shortage. Buyers of shortage being you know the the
[50:46] shortage being you know the the hyperscalers
[50:48] hyperscalers but if you own a giant installed base of
[50:52] but if you own a giant installed base of what is currently in shortage that's
[50:55] what is currently in shortage that's also a very very good place to be. And
[50:57] also a very very good place to be. And we're hearing, you know, CPUs are way
[50:59] we're hearing, you know, CPUs are way more important than they were in an
[51:00] more important than they were in an agentic world. They do all these things
[51:02] agentic world. They do all these things around orchestration, tool calls, etc.,
[51:03] around orchestration, tool calls, etc., etc., etc. The biggest CPU fleets in the
[51:06] etc., etc. The biggest CPU fleets in the world sit at the hyperscalers. So, I
[51:08] world sit at the hyperscalers. So, I think some of these hyperscalers may
[51:09] think some of these hyperscalers may have,
[51:10] have, >> you know, may may catch up a little bit
[51:12] >> you know, may may catch up a little bit to the sellers of shortage.
[51:13] to the sellers of shortage. >> I want to talk about this idea of
[51:15] >> I want to talk about this idea of different and hard applied outside of
[51:17] different and hard applied outside of the infrastructure piece of this. So,
[51:20] the infrastructure piece of this. So, now you're starting to interact with new
[51:21] now you're starting to interact with new founders, um, existing CEOs and founders
[51:24] founders, um, existing CEOs and founders that have to adjust to this new world.
[51:26] that have to adjust to this new world. What are you seeing like the most AI
[51:28] What are you seeing like the most AI native founders that aren't building
[51:30] native founders that aren't building chips or infrastructure or models, but
[51:32] chips or infrastructure or models, but just people using this technology to
[51:34] just people using this technology to build other stuff? How do they feel the
[51:36] build other stuff? How do they feel the most different to you if if you've
[51:38] most different to you if if you've observed differences?
[51:39] observed differences? >> Well, one, I do think this is just for
[51:40] >> Well, one, I do think this is just for chip design. To me, it's always been a
[51:42] chip design. To me, it's always been a fundamental question for venture. So,
[51:44] fundamental question for venture. So, there are different ideas that are
[51:47] there are different ideas that are obvious to everyone on planet Earth as
[51:48] obvious to everyone on planet Earth as soon as they hear it. And if that's
[51:50] soon as they hear it. And if that's where you are in venture, if it's not
[51:52] where you are in venture, if it's not hard to do, if it becomes obvious to the
[51:55] hard to do, if it becomes obvious to the world before you have built um scale,
[51:59] world before you have built um scale, scale is the ultimate advantage, you're
[52:00] scale is the ultimate advantage, you're in trouble. And the great thing Amazon
[52:03] in trouble. And the great thing Amazon had was um you know, I think it was
[52:06] had was um you know, I think it was obvious to a lot of people, but it
[52:08] obvious to a lot of people, but it wasn't obvious to the retail CEOs. And
[52:10] wasn't obvious to the retail CEOs. And Amazon, they were very smart.
[52:13] Amazon, they were very smart. Any e-commerce company that VCs invested
[52:16] Any e-commerce company that VCs invested in, they would destroy. They'd be like,
[52:18] in, they would destroy. They'd be like, "Oh, that's so cute. We're gonna we're
[52:20] "Oh, that's so cute. We're gonna we're gonna take our margins of that to
[52:22] gonna take our margins of that to negative 10,000%."
[52:24] negative 10,000%." And that's what like like the guys at
[52:26] And that's what like like the guys at Wayfair, they did something hard and
[52:27] Wayfair, they did something hard and Amazon tried to kill them and they
[52:28] Amazon tried to kill them and they failed. Those are like tough
[52:30] failed. Those are like tough operationally
[52:31] operationally like really competent CEOs. For me in
[52:34] like really competent CEOs. For me in venture, I always look, is this going to
[52:36] venture, I always look, is this going to be obvious to the world before this
[52:39] be obvious to the world before this company could build scale
[52:42] company could build scale or is this both not obvious, different,
[52:45] or is this both not obvious, different, and really hard to do? I think a lot of
[52:48] and really hard to do? I think a lot of founders are really struggling with this
[52:51] founders are really struggling with this >> in AI like I think people are
[52:56] >> in AI like I think people are becoming worried you know today in that
[52:59] becoming worried you know today in that in Jensen's five layer cake of AI
[53:02] in Jensen's five layer cake of AI and the profits they're acrewing to
[53:04] and the profits they're acrewing to energy they're crewing to data centers
[53:06] energy they're crewing to data centers they're crewing to chips they're
[53:07] they're crewing to chips they're acrewing to models they're not really
[53:09] acrewing to models they're not really acrewing to the applications cursor and
[53:12] acrewing to the applications cursor and cognition you know got to a scale you
[53:15] cognition you know got to a scale you know they focused on coding
[53:17] know they focused on coding you know 18 months ago the people were
[53:19] you know 18 months ago the people were focusing on coding. OpenAI was doing
[53:20] focusing on coding. OpenAI was doing everything under the sun. The people
[53:22] everything under the sun. The people focused on coding were cursor cognition
[53:24] focused on coding were cursor cognition and um anthropic and it was really right
[53:27] and um anthropic and it was really right to focus on code. Um I'm MSAD the
[53:30] to focus on code. Um I'm MSAD the founder of Replet tweeted something that
[53:32] founder of Replet tweeted something that I thought was so smart just it was
[53:34] I thought was so smart just it was something like you know bitter lesson
[53:36] something like you know bitter lesson adjacent is the fact that coding might
[53:39] adjacent is the fact that coding might be the shortest path to ASI and useful
[53:41] be the shortest path to ASI and useful AI because if you're really good at
[53:43] AI because if you're really good at coding you can write yourself code to do
[53:45] coding you can write yourself code to do anything. So I think it was really smart
[53:46] anything. So I think it was really smart of those companies to focus intensely on
[53:48] of those companies to focus intensely on coding and I think they all probably got
[53:51] coding and I think they all probably got to a scale where they they have a place.
[53:54] to a scale where they they have a place. I think cognition is doing something
[53:55] I think cognition is doing something really really different but I think a
[53:57] really really different but I think a lot of founders are really struggling
[53:59] lot of founders are really struggling man they're really struggling
[54:03] man they're really struggling and you know I think they're trying to
[54:04] and you know I think they're trying to get confidence that in nichier areas
[54:08] get confidence that in nichier areas >> that they can get to them and get like a
[54:12] >> that they can get to them and get like a you know a data moat
[54:14] you know a data moat >> before the model companies get to that
[54:16] >> before the model companies get to that niche or that it's a small enough niche
[54:18] niche or that it's a small enough niche that the model companies won't do it
[54:20] that the model companies won't do it themselves but it can still produce a
[54:22] themselves but it can still produce a venture outcome. Is this related to what
[54:23] venture outcome. Is this related to what you would call like the token path? I
[54:24] you would call like the token path? I know you've used that phrase with me
[54:26] know you've used that phrase with me before.
[54:26] before. >> Yeah, I he comes from a guy um at
[54:28] >> Yeah, I he comes from a guy um at alttimeter, Jamon Ball. He just said if
[54:30] alttimeter, Jamon Ball. He just said if you're a software company or an AI
[54:32] you're a software company or an AI company of any kind, you have to be in
[54:33] company of any kind, you have to be in the token path. So, data bricks that's
[54:35] the token path. So, data bricks that's in the token path. Comparable companies
[54:37] in the token path. Comparable companies are in the token path. If you're not in
[54:39] are in the token path. If you're not in the token path
[54:41] the token path and you're not in some really niche
[54:45] and you're not in some really niche thing, life may be hard. And even for
[54:49] thing, life may be hard. And even for these vertical niches, I think if you
[54:51] these vertical niches, I think if you talk to the people at the model
[54:54] talk to the people at the model companies, they're even skeptical of
[54:56] companies, they're even skeptical of some of these because all of the data
[54:59] some of these because all of the data that's, you know, being generated in
[55:01] that's, you know, being generated in these niches come from humans. But then
[55:03] these niches come from humans. But then you're betting that you're able to use
[55:05] you're betting that you're able to use that proprietary data in this narrow
[55:07] that proprietary data in this narrow vertical to train a model that's lower
[55:10] vertical to train a model that's lower cost than the frontier labs can ever get
[55:12] cost than the frontier labs can ever get to. And maybe that's a good bet, but I
[55:14] to. And maybe that's a good bet, but I just think you have to be very very
[55:16] just think you have to be very very careful. Now on the other hand, if the
[55:19] careful. Now on the other hand, if the returns to these frontier tokens
[55:21] returns to these frontier tokens relative to other tokens come down,
[55:24] relative to other tokens come down, there's going to be an explosion in
[55:26] there's going to be an explosion in value creation at the application layer.
[55:29] value creation at the application layer. And I think another really important
[55:31] And I think another really important point is
[55:34] point is I have a belief that whenever he wants,
[55:39] I have a belief that whenever he wants, Jensen can probably get pretty close to
[55:41] Jensen can probably get pretty close to the frontier
[55:43] the frontier >> with his own model.
[55:44] >> with his own model. >> With his own model, they're doing some
[55:45] >> With his own model, they're doing some really cool things. Neimatronics
[55:47] really cool things. Neimatronics >> commoditize your compliment as
[55:49] >> commoditize your compliment as >> say I don't think he wants to do that
[55:52] >> say I don't think he wants to do that that is what open AI and you know
[55:56] that is what open AI and you know anthropic are kind of trying to do to
[55:58] anthropic are kind of trying to do to him unsuccessfully
[56:01] him unsuccessfully but so it's just like he's a very
[56:02] but so it's just like he's a very logical thinker this is the logical
[56:04] logical thinker this is the logical counter move
[56:06] counter move >> and I think you will see that like
[56:08] >> and I think you will see that like opensource frontier which today consists
[56:12] opensource frontier which today consists of you know Chinese models with stolen
[56:15] of you know Chinese models with stolen American tokens you know Somebody told
[56:16] American tokens you know Somebody told me that like Deep Seek
[56:19] me that like Deep Seek uh the latest one or maybe the original
[56:21] uh the latest one or maybe the original one was only 150,000 reasoning traces.
[56:23] one was only 150,000 reasoning traces. There's many ways to launder this if
[56:25] There's many ways to launder this if you're a Chinese company. You know, you
[56:28] you're a Chinese company. You know, you can hit all these different APIs. You
[56:31] can hit all these different APIs. You can make it hard. Now, the American labs
[56:32] can make it hard. Now, the American labs are working really hard on
[56:34] are working really hard on anti-distillation technology. But I I
[56:36] anti-distillation technology. But I I just think Chinese open source, they're
[56:38] just think Chinese open source, they're doing really impressive things in a very
[56:40] doing really impressive things in a very resource constrained way. But there's a
[56:42] resource constrained way. But there's a lot of distillation. And this is why I
[56:45] lot of distillation. And this is why I think in addition to there not being
[56:46] think in addition to there not being enough compute to serve Mythos
[56:49] enough compute to serve Mythos just they did not want it to be
[56:52] just they did not want it to be distilled. they wanted to use Mythos,
[56:55] distilled. they wanted to use Mythos, you know, distill it themselves, use it
[56:57] you know, distill it themselves, use it to RL their next model, whatever it is.
[57:00] to RL their next model, whatever it is. And then I think what they and
[57:02] And then I think what they and eventually I think if OpenAI gets to,
[57:04] eventually I think if OpenAI gets to, you know, economics feel good about
[57:06] you know, economics feel good about anyone on the frontier will do is just
[57:07] anyone on the frontier will do is just say, you know, there's going to be some
[57:09] say, you know, there's going to be some very interesting game theory because
[57:11] very interesting game theory because it's it is it's a new kind of prisoner's
[57:13] it's it is it's a new kind of prisoner's dilemma. You know, we talked about the
[57:14] dilemma. You know, we talked about the old prisoners dilemma being just around
[57:17] old prisoners dilemma being just around like, hey, you you're in a prisoner's
[57:18] like, hey, you you're in a prisoner's dilemma where you have to spend. The new
[57:20] dilemma where you have to spend. The new prisoners dilemma is going to be if you
[57:22] prisoners dilemma is going to be if you were at the frontier, do you release
[57:24] were at the frontier, do you release that model via API or not?
[57:26] that model via API or not? >> And if everyone at the frontier agrees
[57:30] >> And if everyone at the frontier agrees not to do that, then Chinese open source
[57:33] not to do that, then Chinese open source is quickly
[57:34] is quickly >> if one person defects, they're going to
[57:37] >> if one person defects, they're going to have the best model. They're going to
[57:39] have the best model. They're going to have a lot of revenue and cash flow and
[57:41] have a lot of revenue and cash flow and then of course resources equal
[57:42] then of course resources equal intelligence. So they'll start to pull
[57:44] intelligence. So they'll start to pull ahead and then that will lead to, you
[57:47] ahead and then that will lead to, you know, everybody else releasing it. So
[57:48] know, everybody else releasing it. So it's a new game theory. It's kind of the
[57:50] it's a new game theory. It's kind of the same game theory that you have with
[57:51] same game theory that you have with Taiwan semi Samsung and Intel. The
[57:53] Taiwan semi Samsung and Intel. The reality is like if if a company like
[57:55] reality is like if if a company like Nvidia were or AMD were to ever really
[57:58] Nvidia were or AMD were to ever really really use one of these other
[58:01] really use one of these other foundaries, that foundry would get
[58:02] foundaries, that foundry would get better really quickly. So I do think
[58:06] better really quickly. So I do think Jensen is going to keep open source a
[58:09] Jensen is going to keep open source a certain time frame behind the frontier.
[58:12] certain time frame behind the frontier. I think that's going to be a very
[58:14] I think that's going to be a very interesting thing to watch. And then by
[58:16] interesting thing to watch. And then by the way, open source gets monetized.
[58:17] the way, open source gets monetized. There's this misnomer that open source
[58:19] There's this misnomer that open source is free. Open source tokens, they cost
[58:21] is free. Open source tokens, they cost energy. They, you know, they cost energy
[58:23] energy. They, you know, they cost energy to produce. You need to make up on GPUs
[58:25] to produce. You need to make up on GPUs and the open source model companies
[58:26] and the open source model companies almost always get a revenue share.
[58:28] almost always get a revenue share. >> How are you preparing a trades for the
[58:31] >> How are you preparing a trades for the world of Mythos 3, Mythos 4.
[58:34] world of Mythos 3, Mythos 4. >> We're just trying to overinvest in cyber
[58:36] >> We're just trying to overinvest in cyber security. You know, something I've like,
[58:37] security. You know, something I've like, you know, said in multiple forums and I
[58:39] you know, said in multiple forums and I really believe is you everybody needs to
[58:42] really believe is you everybody needs to have a safe word. Everybody needs to go
[58:46] have a safe word. Everybody needs to go leave your digital devices behind.
[58:48] leave your digital devices behind. Literally go to the ocean and have a
[58:50] Literally go to the ocean and have a family safe word or a company safe word.
[58:52] family safe word or a company safe word. And it can't be one that can be like
[58:54] And it can't be one that can be like socially engineered. And this is just to
[58:56] socially engineered. And this is just to avoid like cyber crime where like what
[58:58] avoid like cyber crime where like what looks like your son or your daughter or
[59:00] looks like your son or your daughter or your your grandparents or your parents
[59:02] your your grandparents or your parents or whatever facetimes you. It's an
[59:05] or whatever facetimes you. It's an utterly accurate
[59:08] utterly accurate simulation of them. they know everything
[59:10] simulation of them. they know everything and can extrapolate based on what
[59:12] and can extrapolate based on what they've said, what they're likely to
[59:13] they've said, what they're likely to say, and says, you know, wire me a
[59:16] say, and says, you know, wire me a million bucks.
[59:17] million bucks. >> That's defensive. What about what will
[59:18] >> That's defensive. What about what will you still be able to do that it won't be
[59:19] you still be able to do that it won't be able to do, I guess,
[59:21] able to do, I guess, >> on the analytical side.
[59:22] >> on the analytical side. >> So, it's a good question. I did just
[59:23] >> So, it's a good question. I did just have I I just watched The Last Samurai
[59:25] have I I just watched The Last Samurai and I asked um people at my firm to
[59:27] and I asked um people at my firm to watch it. And The Last Samurai, if you
[59:29] watch it. And The Last Samurai, if you haven't seen it, I highly recommend
[59:31] haven't seen it, I highly recommend watching it. It's actually a movie
[59:32] watching it. It's actually a movie that's aged really well. Tom Cruz movie
[59:34] that's aged really well. Tom Cruz movie from 20 years ago. You know, the conceit
[59:36] from 20 years ago. You know, the conceit is Tom Cruz is this like bitter, washed
[59:38] is Tom Cruz is this like bitter, washed up Civil War veteran who's actually a
[59:40] up Civil War veteran who's actually a very good soldier. He's bitter and
[59:42] very good soldier. He's bitter and washed up because he feels like he
[59:44] washed up because he feels like he participated in negative actions against
[59:46] participated in negative actions against the Native Americans. He's hired by
[59:48] the Native Americans. He's hired by Japan to train just during the Miji
[59:50] Japan to train just during the Miji restoration. And he's hired by the
[59:52] restoration. And he's hired by the modern elements of the Japanese
[59:54] modern elements of the Japanese government to train like an army of
[59:56] government to train like an army of peasants
[59:57] peasants >> how to fight the samurai. There's a
[59:59] >> how to fight the samurai. There's a first battle. Of course, the samurai win
[01:00:01] first battle. Of course, the samurai win even though they don't have guns. He
[01:00:03] even though they don't have guns. He fights valiantly. So the samurai decide
[01:00:05] fights valiantly. So the samurai decide not to kill him, take him to their
[01:00:06] not to kill him, take him to their village. He becomes a samurai. It feels
[01:00:08] village. He becomes a samurai. It feels like the civil war to him. So he fights
[01:00:10] like the civil war to him. So he fights on the side of the samurai.
[01:00:12] on the side of the samurai. And at the end, he's massacred by a
[01:00:14] And at the end, he's massacred by a peasant with a machine gun. And like the
[01:00:16] peasant with a machine gun. And like the machine gun is here and if we do not all
[01:00:21] machine gun is here and if we do not all become masters of the machine gun, we're
[01:00:23] become masters of the machine gun, we're going to get mastered. So I am trying to
[01:00:25] going to get mastered. So I am trying to become a master of the machine gun. And
[01:00:27] become a master of the machine gun. And then, you know, I'm optimistic. There's
[01:00:30] then, you know, I'm optimistic. There's a long period of time where just like if
[01:00:33] a long period of time where just like if you were a 50year-old samurai veteran of
[01:00:37] you were a 50year-old samurai veteran of many wars, I fought many wars, master
[01:00:39] many wars, I fought many wars, master dwarf. Um, you will have advantages
[01:00:42] dwarf. Um, you will have advantages using the machine gun. And I'm
[01:00:43] using the machine gun. And I'm optimistic as a lifelong student of
[01:00:46] optimistic as a lifelong student of investing. I'm going to be able to
[01:00:47] investing. I'm going to be able to master the machine gun, this new
[01:00:49] master the machine gun, this new technology, um, integrate it into my own
[01:00:52] technology, um, integrate it into my own process, integrate it into our firm's
[01:00:54] process, integrate it into our firm's process in ways that, you know, let me
[01:00:57] process in ways that, you know, let me contribute value as a human being for a
[01:01:00] contribute value as a human being for a long time. But, you know, like everyone,
[01:01:01] long time. But, you know, like everyone, like, you know, I have agents running
[01:01:03] like, you know, I have agents running all the time now.
[01:01:04] all the time now. >> What's your most useful agent? The most
[01:01:05] >> What's your most useful agent? The most useful agent honestly is as and I think
[01:01:08] useful agent honestly is as and I think I told you this and I don't want to hurt
[01:01:10] I told you this and I don't want to hurt your business, but my single most useful
[01:01:12] your business, but my single most useful agent is a really good summary of the
[01:01:16] agent is a really good summary of the points that would be interesting to me
[01:01:18] points that would be interesting to me from podcasts. There's like six hours a
[01:01:21] from podcasts. There's like six hours a day of stuff that I feel like it's in my
[01:01:23] day of stuff that I feel like it's in my job description to watch, you know,
[01:01:25] job description to watch, you know, every time every time somebody from
[01:01:27] every time every time somebody from OpenAI, XAI,
[01:01:31] OpenAI, XAI, Google,
[01:01:32] Google, you know, Cursor,
[01:01:34] you know, Cursor, Fireworks, Bin, I say nothing of like
[01:01:37] Fireworks, Bin, I say nothing of like Jensen, Elon, Daario. Um, I feel
[01:01:42] Jensen, Elon, Daario. Um, I feel compelled to watch and I just don't have
[01:01:44] compelled to watch and I just don't have that much time. And there's some real
[01:01:47] that much time. And there's some real needles and hay stacks. There's a set of
[01:01:48] needles and hay stacks. There's a set of things I always like to see like I'm
[01:01:50] things I always like to see like I'm very sensitive to management
[01:01:51] very sensitive to management compensation. What are they incented to
[01:01:53] compensation. What are they incented to do? They do they just have stupid RSUs
[01:01:56] do? They do they just have stupid RSUs or do they have PSUs? And if they have
[01:01:58] or do they have PSUs? And if they have PSUs, what are those PSUs incent them to
[01:02:00] PSUs, what are those PSUs incent them to do? I think systems that do a very good
[01:02:02] do? I think systems that do a very good first pass at that and you know that
[01:02:05] first pass at that and you know that saves people a lot of time. It frees
[01:02:07] saves people a lot of time. It frees them up for more creative work than like
[01:02:10] them up for more creative work than like you know going through the proxy pulling
[01:02:12] you know going through the proxy pulling the PSU thing looking at how it's
[01:02:15] the PSU thing looking at how it's changed versus all the proxies because
[01:02:17] changed versus all the proxies because there's signal in that and that's very
[01:02:19] there's signal in that and that's very labor intensive and that's so good for
[01:02:20] labor intensive and that's so good for an AI and there's obviously all sorts of
[01:02:22] an AI and there's obviously all sorts of same things within investing. This is
[01:02:24] same things within investing. This is the most exciting thrilling time to be
[01:02:26] the most exciting thrilling time to be an investor
[01:02:28] an investor >> and there is and it is I am a little I'm
[01:02:30] >> and there is and it is I am a little I'm getting a little bit worried
[01:02:32] getting a little bit worried >> the diversity breakdown thing. Yeah, I'm
[01:02:34] >> the diversity breakdown thing. Yeah, I'm getting
[01:02:35] getting >> Say just like a little bit more about
[01:02:36] >> Say just like a little bit more about like the kinds of people that are
[01:02:37] like the kinds of people that are >> I don't know anyone like me who's not
[01:02:39] >> I don't know anyone like me who's not really bullish on DRM.
[01:02:42] really bullish on DRM. >> No one.
[01:02:42] >> No one. >> No one. There's all these interesting
[01:02:44] >> No one. There's all these interesting things happening with AI right now. So,
[01:02:46] things happening with AI right now. So, one is cross-sectionally the valuations
[01:02:48] one is cross-sectionally the valuations do not make sense. They just flat out do
[01:02:51] do not make sense. They just flat out do not make sense. They cannot all be true.
[01:02:54] not make sense. They cannot all be true. You have semicap equipment companies
[01:02:56] You have semicap equipment companies trading at 40 times next quarter's
[01:02:58] trading at 40 times next quarter's annualized earnings and DRAM companies
[01:03:00] annualized earnings and DRAM companies trading at mid-s single digit. at the
[01:03:02] trading at mid-s single digit. at the peak of the last cycle that was like
[01:03:04] peak of the last cycle that was like five verse 12. At one point it was like
[01:03:06] five verse 12. At one point it was like three verse 45. Those can't both be
[01:03:09] three verse 45. Those can't both be true. And yes, semiconductor capex
[01:03:12] true. And yes, semiconductor capex business models have improved more than
[01:03:14] business models have improved more than the memory business models. We don't
[01:03:16] the memory business models. We don't know how much HBM is going to improve
[01:03:18] know how much HBM is going to improve memory business models yet. Yes, they
[01:03:20] memory business models yet. Yes, they have some element of recurring revenue
[01:03:22] have some element of recurring revenue with parts and maintenance, but it's not
[01:03:25] with parts and maintenance, but it's not worth a,000% multiple gap. I think it's
[01:03:27] worth a,000% multiple gap. I think it's hard to square like the valuation of
[01:03:29] hard to square like the valuation of something like Nvidia which is still you
[01:03:31] something like Nvidia which is still you know in in in early April was
[01:03:33] know in in in early April was essentially as cheap as it gets relative
[01:03:35] essentially as cheap as it gets relative to the market like in the last 10 or 12
[01:03:37] to the market like in the last 10 or 12 years or whatever it is and very cheap
[01:03:39] years or whatever it is and very cheap absolute it's very hard to square that
[01:03:41] absolute it's very hard to square that valuation with something like GE
[01:03:44] valuation with something like GE Vernova's valuation
[01:03:46] Vernova's valuation >> because it builds in like an
[01:03:48] >> because it builds in like an unfathomable amount of share loss for
[01:03:51] unfathomable amount of share loss for Nvidia. So valuations cross-sectionally
[01:03:53] Nvidia. So valuations cross-sectionally are really different because we are in
[01:03:56] are really different because we are in shortages.
[01:03:58] shortages. The lowest quality companies are doing
[01:04:00] The lowest quality companies are doing the best. So if you're an oil and gas
[01:04:03] the best. So if you're an oil and gas investor or you know a mighty investor,
[01:04:05] investor or you know a mighty investor, natural resources investor and you're,
[01:04:08] natural resources investor and you're, you know, you're well versed in thinking
[01:04:09] you know, you're well versed in thinking of costs, this is very intuitive to you.
[01:04:11] of costs, this is very intuitive to you. In a real bull market for a commodity,
[01:04:14] In a real bull market for a commodity, the commodity suppliers with the highest
[01:04:16] the commodity suppliers with the highest costs go up the most because it's the
[01:04:19] costs go up the most because it's the most beneficial to them. They go from on
[01:04:21] most beneficial to them. They go from on the verge of bankruptcy to gushing cash.
[01:04:23] the verge of bankruptcy to gushing cash. And this is, I think, one reason
[01:04:25] And this is, I think, one reason commodity investing is really, really
[01:04:26] commodity investing is really, really hard because quality outperforms during
[01:04:29] hard because quality outperforms during the cycles, but you get all of the
[01:04:31] the cycles, but you get all of the outperformance during the downturns when
[01:04:33] outperformance during the downturns when the high-cost guys that moon during the
[01:04:36] the high-cost guys that moon during the shortages and the commodity bull
[01:04:37] shortages and the commodity bull markets, you know, go bankrupt or
[01:04:39] markets, you know, go bankrupt or whatever. You're seeing that happen in
[01:04:40] whatever. You're seeing that happen in every industry. the lowest quality
[01:04:43] every industry. the lowest quality players in, you know, these different
[01:04:45] players in, you know, these different industries that are hated and detested
[01:04:49] industries that are hated and detested by the hyperscalers and the buyers
[01:04:51] by the hyperscalers and the buyers because they have high costs, they're
[01:04:53] because they have high costs, they're unreliable, the parts fail at a high
[01:04:55] unreliable, the parts fail at a high rate, etc., etc. They're sold out and
[01:04:57] rate, etc., etc. They're sold out and raising prices. Um, and then that
[01:04:59] raising prices. Um, and then that activity gets the interest of like these
[01:05:02] activity gets the interest of like these retail accounts on X and these stocks
[01:05:05] retail accounts on X and these stocks get bid to the moon. whereas some of the
[01:05:08] get bid to the moon. whereas some of the higher quality expressions
[01:05:10] higher quality expressions have like actually really underperformed
[01:05:13] have like actually really underperformed and you know as an investor it's it's
[01:05:15] and you know as an investor it's it's hard because you know within a like
[01:05:20] hard because you know within a like shadow of a doubt that that thing that's
[01:05:23] shadow of a doubt that that thing that's moved you know 10x in 3 months or 6
[01:05:26] moved you know 10x in 3 months or 6 months is going to go right back down
[01:05:30] months is going to go right back down subject to what they do with all the
[01:05:31] subject to what they do with all the cash. But like these low quality
[01:05:33] cash. But like these low quality companies really do smart stuff with
[01:05:34] companies really do smart stuff with cash. And so it worries me a little bit
[01:05:36] cash. And so it worries me a little bit that people who were very skeptical a
[01:05:38] that people who were very skeptical a year ago are no longer skeptical. But
[01:05:41] year ago are no longer skeptical. But then I just contrast that with like the
[01:05:43] then I just contrast that with like the valuations of these like highquality
[01:05:46] valuations of these like highquality companies which are just not extended
[01:05:49] companies which are just not extended and it makes me feel better. But it does
[01:05:51] and it makes me feel better. But it does kind of feel like, you know, I always
[01:05:52] kind of feel like, you know, I always thought it was funny in 24 and 25 that
[01:05:54] thought it was funny in 24 and 25 that anyone asked about an AI bubble or
[01:05:56] anyone asked about an AI bubble or talked about it because it's like you
[01:05:58] talked about it because it's like you have this nuclear bubble and this
[01:05:59] have this nuclear bubble and this quantum bubble right here, right in
[01:06:01] quantum bubble right here, right in front of you. What are we talking about?
[01:06:03] front of you. What are we talking about? This is so real. Some of that nuclear
[01:06:06] This is so real. Some of that nuclear quantum silliness is maybe spread into
[01:06:09] quantum silliness is maybe spread into more speculative, lower quality, smaller
[01:06:12] more speculative, lower quality, smaller cap names where if you have a big
[01:06:15] cap names where if you have a big presence on X or Reddit, it's easy to
[01:06:18] presence on X or Reddit, it's easy to move them. And that frightens me a
[01:06:20] move them. And that frightens me a little bit, but I just wish there were
[01:06:22] little bit, but I just wish there were more AI bears. Like I wish there were
[01:06:24] more AI bears. Like I wish there were more memory bears. You know, one reason
[01:06:26] more memory bears. You know, one reason I'm um you know, Astera is a stock I've
[01:06:29] I'm um you know, Astera is a stock I've been close to a long time. There's a lot
[01:06:31] been close to a long time. There's a lot of bears on that. I love that. Great.
[01:06:34] of bears on that. I love that. Great. You know, I first invested in the series
[01:06:36] You know, I first invested in the series C. Good luck thinking you're going to
[01:06:37] C. Good luck thinking you're going to price that, you know, differentially for
[01:06:40] price that, you know, differentially for me. You know, good luck thinking that's
[01:06:41] me. You know, good luck thinking that's a copper loser. And then there's also
[01:06:44] a copper loser. And then there's also you can feel the baskets in the market
[01:06:46] you can feel the baskets in the market and the leverage baskets and what
[01:06:48] and the leverage baskets and what baskets you're in is really important.
[01:06:50] baskets you're in is really important. You know, copper, optical, DRAM, NAND.
[01:06:54] You know, copper, optical, DRAM, NAND. Um, and a very interesting thing that's
[01:06:55] Um, and a very interesting thing that's happened this year, um, is in 24 and 25
[01:06:58] happened this year, um, is in 24 and 25 the AI trade traded together. So like
[01:07:02] the AI trade traded together. So like you could be long GPU compute, scale up
[01:07:05] you could be long GPU compute, scale up networking, and optical scale across and
[01:07:08] networking, and optical scale across and like short power. that trade worked from
[01:07:12] like short power. that trade worked from like a riskmanagement sense because you
[01:07:13] like a riskmanagement sense because you know I'm very factor aware that all blew
[01:07:16] know I'm very factor aware that all blew out in January of this year it's like
[01:07:19] out in January of this year it's like you know scale up networking would go
[01:07:21] you know scale up networking would go crazy while scale out was going down or
[01:07:23] crazy while scale out was going down or DRM massively underperforming NAND and
[01:07:27] DRM massively underperforming NAND and HDDs which had not h happened so these
[01:07:30] HDDs which had not h happened so these cross-sectional correlations within AI
[01:07:34] cross-sectional correlations within AI really fell apart and you had to get
[01:07:36] really fell apart and you had to get very fine grained you couldn't hedge
[01:07:39] very fine grained you couldn't hedge your memory
[01:07:40] your memory anymore with like some semicap equipment
[01:07:43] anymore with like some semicap equipment or nan everything cross-sectionally
[01:07:48] or nan everything cross-sectionally really changed and in a very interesting
[01:07:50] really changed and in a very interesting way in January and I think maybe one
[01:07:53] way in January and I think maybe one reason for that was you know the AI got
[01:07:55] reason for that was you know the AI got to a quality where it was all of a
[01:07:58] to a quality where it was all of a sudden really easy for a bunch of people
[01:08:00] sudden really easy for a bunch of people to get really smart on these different
[01:08:02] to get really smart on these different subsectors start trading them and then
[01:08:04] subsectors start trading them and then they get put into baskets and those
[01:08:07] they get put into baskets and those baskets
[01:08:07] baskets >> yeah creating price efficiency Yeah,
[01:08:09] >> yeah creating price efficiency Yeah, exactly. And then it's like if you like
[01:08:12] exactly. And then it's like if you like I think some of the biggest
[01:08:13] I think some of the biggest opportunities outside of these higher
[01:08:15] opportunities outside of these higher quality names that I think can compound
[01:08:16] quality names that I think can compound for a long time and they're safe unlike
[01:08:19] for a long time and they're safe unlike these lowquality names which are
[01:08:20] these lowquality names which are terrifying is in names that are
[01:08:22] terrifying is in names that are miscatategorized
[01:08:24] miscatategorized like Astera was in a lot of copper loser
[01:08:26] like Astera was in a lot of copper loser baskets. Astera their biggest product is
[01:08:29] baskets. Astera their biggest product is going to be a switch. You use both
[01:08:32] going to be a switch. You use both copper and optics to connect switches to
[01:08:35] copper and optics to connect switches to accelerators.
[01:08:37] accelerators. And so definitionally, if you're a
[01:08:40] And so definitionally, if you're a switch company or an accelerator
[01:08:42] switch company or an accelerator company, you cannot be a copper loser
[01:08:45] company, you cannot be a copper loser because you're going to be on the other
[01:08:46] because you're going to be on the other side of that connection. I
[01:08:47] side of that connection. I >> I wonder if you could riff just for like
[01:08:49] >> I wonder if you could riff just for like a sentence or two on each of the major
[01:08:51] a sentence or two on each of the major companies. I feel like I always forget
[01:08:52] companies. I feel like I always forget to ask you like Google, Microsoft,
[01:08:54] to ask you like Google, Microsoft, Amazon, you know, the the major players
[01:08:55] Amazon, you know, the the major players that are public that all the
[01:08:57] that are public that all the conversation is centered around these
[01:08:59] conversation is centered around these exciting new companies.
[01:09:00] exciting new companies. >> Yeah. So Google um it was incredible
[01:09:02] >> Yeah. So Google um it was incredible last year because they had that TPU
[01:09:04] last year because they had that TPU advantage which is now gone. The reason
[01:09:05] advantage which is now gone. The reason I think they're still in a great
[01:09:07] I think they're still in a great position is just they have the most
[01:09:09] position is just they have the most compute of everyone. We talked about the
[01:09:11] compute of everyone. We talked about the value of installed bases being higher as
[01:09:13] value of installed bases being higher as a result of shortages.
[01:09:15] a result of shortages. >> They have the biggest installed base of
[01:09:16] >> They have the biggest installed base of compute. Yeah,
[01:09:18] compute. Yeah, >> I am a little surprised
[01:09:22] >> I am a little surprised by
[01:09:24] by their inability and Google IO is this um
[01:09:28] their inability and Google IO is this um is this week
[01:09:30] is this week >> and um like if they don't release
[01:09:34] >> and um like if they don't release something that even slightly leapfrogs
[01:09:39] something that even slightly leapfrogs open AI
[01:09:40] open AI andor um clawed
[01:09:43] andor um clawed like that that's interesting and it's
[01:09:46] like that that's interesting and it's not a disaster. faster for Google. It's
[01:09:48] not a disaster. faster for Google. It's just interesting and it just means this
[01:09:50] just interesting and it just means this Nvidia effect we discussed is even more
[01:09:52] Nvidia effect we discussed is even more powerful than maybe I'd imagined. But
[01:09:53] powerful than maybe I'd imagined. But I'm very curious to see what the paro
[01:09:56] I'm very curious to see what the paro frontier looks like literally in 5 days
[01:09:58] frontier looks like literally in 5 days after Google's announced its new stuff.
[01:10:01] after Google's announced its new stuff. This is a big card for them. But Google,
[01:10:03] This is a big card for them. But Google, you know, between um the amount of data
[01:10:05] you know, between um the amount of data they have and the YouTube data is
[01:10:07] they have and the YouTube data is actually really genuinely valuable. It's
[01:10:09] actually really genuinely valuable. It's actually it is valuable in a world of
[01:10:12] actually it is valuable in a world of robotics. The amount of compute they
[01:10:14] robotics. The amount of compute they have and you know the search business
[01:10:16] have and you know the search business they have. Google's never not going to
[01:10:18] they have. Google's never not going to be in a good position. And then you see
[01:10:20] be in a good position. And then you see that with GCP going crazy. You got to
[01:10:22] that with GCP going crazy. You got to give Zuckerberg immense credit. Um what
[01:10:25] give Zuckerberg immense credit. Um what he's done in terms of making Meta an AI
[01:10:27] he's done in terms of making Meta an AI first company internally and I do think
[01:10:30] first company internally and I do think he is the only one of those true
[01:10:32] he is the only one of those true internet giants to have done that. And I
[01:10:36] internet giants to have done that. And I give him a lot of credit for that. I
[01:10:37] give him a lot of credit for that. I give him a lot of credit for um paying
[01:10:40] give him a lot of credit for um paying up when he did for you know all those
[01:10:43] up when he did for you know all those you know those billion dollar contracts
[01:10:45] you know those billion dollar contracts that talent
[01:10:46] that talent >> and Muse I think was a really big upside
[01:10:49] >> and Muse I think was a really big upside surprise um you know was the first model
[01:10:53] surprise um you know was the first model from MSL and it's not on the paro
[01:10:56] from MSL and it's not on the paro frontier with you know XAI Google's one
[01:10:59] frontier with you know XAI Google's one entrant and then openAI and claude but
[01:11:02] entrant and then openAI and claude but it's pretty close that was very
[01:11:04] it's pretty close that was very impressive to me so I think meta is in a
[01:11:07] impressive to me so I think meta is in a better position. Still not as strong of
[01:11:09] better position. Still not as strong of an absolute position as Google, but like
[01:11:11] an absolute position as Google, but like they're better position and rates of
[01:11:13] they're better position and rates of change matter more than level as you
[01:11:15] change matter more than level as you know in markets particularly over short
[01:11:17] know in markets particularly over short like three-year time frames over like
[01:11:19] like three-year time frames over like long time frames level of competitive
[01:11:21] long time frames level of competitive advantages tends to dominate but even
[01:11:23] advantages tends to dominate but even within that you know the changes changes
[01:11:25] within that you know the changes changes are really matter. Amazon I think is in
[01:11:28] are really matter. Amazon I think is in a really strong position because of
[01:11:29] a really strong position because of Trrenium. you're going to see like real
[01:11:32] Trrenium. you're going to see like real P&L efficiencies from robotics over the
[01:11:34] P&L efficiencies from robotics over the next 18 months in their retail business.
[01:11:36] next 18 months in their retail business. I actually think Nova their internal
[01:11:38] I actually think Nova their internal models are not where Muse is, but
[01:11:41] models are not where Muse is, but they're better than they get credit for.
[01:11:43] they're better than they get credit for. Microsoft, I think Satya is a really
[01:11:45] Microsoft, I think Satya is a really brilliant man, but you know, in in
[01:11:48] brilliant man, but you know, in in investor conversations,
[01:11:50] investor conversations, people just don't talk about him the way
[01:11:52] people just don't talk about him the way that they did. I I like Satya. I admire
[01:11:55] that they did. I I like Satya. I admire him. I think he's an exceptional CEO
[01:11:59] him. I think he's an exceptional CEO and I give him a lot of credit for the
[01:12:01] and I give him a lot of credit for the decisions he's made, but you know, he
[01:12:04] decisions he's made, but you know, he did go from we're going to make Google
[01:12:05] did go from we're going to make Google dance to being the product manager of
[01:12:07] dance to being the product manager of Copilot in like three years. I I would
[01:12:10] Copilot in like three years. I I would love to know during the coup attempt
[01:12:12] love to know during the coup attempt against OpenAI, does Satcha regret his
[01:12:15] against OpenAI, does Satcha regret his decisions?
[01:12:17] decisions? Does Satia wish that he had supported
[01:12:20] Does Satia wish that he had supported Ilia and instead of Sam and that kind of
[01:12:23] Ilia and instead of Sam and that kind of Ilia and Meera were really running
[01:12:27] Ilia and Meera were really running OpenAI today? In his heart of hearts, I
[01:12:30] OpenAI today? In his heart of hearts, I would love to know because I think the
[01:12:33] would love to know because I think the Microsoft OpenAI partnership might look
[01:12:35] Microsoft OpenAI partnership might look very different in that world. I think
[01:12:38] very different in that world. I think that's a very interesting question that
[01:12:40] that's a very interesting question that we'll never know the answer to.
[01:12:43] we'll never know the answer to. But I give him a lot of credit like he
[01:12:45] But I give him a lot of credit like he is what he is doing now he's taking risk
[01:12:50] is what he is doing now he's taking risk so they could earn you know this goes to
[01:12:52] so they could earn you know this goes to the decisions you have to make in that
[01:12:54] the decisions you have to make in that cone of uncertainty are not only how
[01:12:56] cone of uncertainty are not only how much you spend but what you're going to
[01:12:58] much you spend but what you're going to spend it on I think Microsoft flinched
[01:13:03] spend it on I think Microsoft flinched for like a moment in early 25 you know
[01:13:06] for like a moment in early 25 you know they have this algorithm we spend this
[01:13:08] they have this algorithm we spend this much capex dollars we get this return
[01:13:10] much capex dollars we get this return that algorithm was kind of off and if
[01:13:13] that algorithm was kind of off and if you flinch you lose position
[01:13:15] you flinch you lose position >> you lose all these allocations and it's
[01:13:17] >> you lose all these allocations and it's difficult to get it back. So they
[01:13:19] difficult to get it back. So they flinched and now the decision Satya is
[01:13:21] flinched and now the decision Satya is making which the market has punished him
[01:13:23] making which the market has punished him for but I think is the right decision is
[01:13:26] for but I think is the right decision is we're going to use our compute rather
[01:13:28] we're going to use our compute rather than making I mean who knows how fast
[01:13:30] than making I mean who knows how fast Azure could be growing if they're
[01:13:32] Azure could be growing if they're willing to just sell GPUs to OpenAI.
[01:13:35] willing to just sell GPUs to OpenAI. We're going to use our compute
[01:13:37] We're going to use our compute internally to make our own products
[01:13:39] internally to make our own products better. You know, one reason C-pilot is
[01:13:41] better. You know, one reason C-pilot is so bad or has been so bad is just not
[01:13:43] so bad or has been so bad is just not enough compute available. They're fixing
[01:13:44] enough compute available. They're fixing that. He's the product manager of
[01:13:47] that. He's the product manager of Copilot. I do think he's a great CEO and
[01:13:51] Copilot. I do think he's a great CEO and they're trying to use their compute to
[01:13:52] they're trying to use their compute to train their own models. I don't I am a
[01:13:55] train their own models. I don't I am a little skeptical that they have the
[01:13:56] little skeptical that they have the right team to succeed there but you know
[01:13:59] right team to succeed there but you know they can certainly like just like Meta
[01:14:01] they can certainly like just like Meta they can afford to hire maybe maybe a
[01:14:04] they can afford to hire maybe maybe a different team but I think he's making
[01:14:07] different team but I think he's making good decisions that are risky decisions
[01:14:11] good decisions that are risky decisions to position Microsoft from for this
[01:14:13] to position Microsoft from for this world where frontier models are are no
[01:14:17] world where frontier models are are no longer API accessible
[01:14:19] longer API accessible >> and I think it's a really courageous
[01:14:20] >> and I think it's a really courageous decision that I give him a lot of credit
[01:14:22] decision that I give him a lot of credit for and he is foregoing I Microsoft
[01:14:24] for and he is foregoing I Microsoft probably be an $800 stock today if they
[01:14:27] probably be an $800 stock today if they were using their GPUs to serve OpenAI
[01:14:30] were using their GPUs to serve OpenAI solely OpenAI and anthropics capacity
[01:14:32] solely OpenAI and anthropics capacity instead of using them for their own
[01:14:34] instead of using them for their own products. So I give him a lot of credit
[01:14:36] products. So I give him a lot of credit for making a great decision. What's
[01:14:38] for making a great decision. What's really interesting is the degree to
[01:14:41] really interesting is the degree to which these companies are outward facing
[01:14:44] which these companies are outward facing in their decisions. The two companies
[01:14:46] in their decisions. The two companies who are the most deeply engaged with
[01:14:48] who are the most deeply engaged with startups are Amazon and Nvidia by a
[01:14:51] startups are Amazon and Nvidia by a mile. Then there's a really intense
[01:14:55] mile. Then there's a really intense engagement with Google, their next most
[01:14:57] engagement with Google, their next most intense. Broadcom is engaged in a
[01:15:00] intense. Broadcom is engaged in a different way. They're just, you know,
[01:15:02] different way. They're just, you know, everybody's favorite AS6 supplier. Like
[01:15:05] everybody's favorite AS6 supplier. Like it's, you know, if you're a startup,
[01:15:06] it's, you know, if you're a startup, it's considered like a level up if you
[01:15:08] it's considered like a level up if you get to work with Broadcom for your
[01:15:09] get to work with Broadcom for your second gen chip. And it's considered
[01:15:11] second gen chip. And it's considered mana from heaven if Broadcom works with
[01:15:13] mana from heaven if Broadcom works with you for their first gen chip. And then
[01:15:15] you for their first gen chip. And then you see essentially
[01:15:17] you see essentially zero engagement with startups from AMD,
[01:15:22] zero engagement with startups from AMD, Microsoft, and Meta. And I just Yeah, I
[01:15:24] Microsoft, and Meta. And I just Yeah, I mean when I say zero, it's a little. And
[01:15:27] mean when I say zero, it's a little. And I just wonder about that decision
[01:15:30] I just wonder about that decision because some of the best teams
[01:15:35] because some of the best teams are no longer at big public companies.
[01:15:37] are no longer at big public companies. They're at these smaller startups.
[01:15:40] They're at these smaller startups. And I think it's going to end up being a
[01:15:42] And I think it's going to end up being a pretty big advantage for Nvidia, AMD,
[01:15:44] pretty big advantage for Nvidia, AMD, Google right behind them to have this
[01:15:47] Google right behind them to have this engagement
[01:15:49] engagement that you just don't see from these other
[01:15:53] that you just don't see from these other um hyperscalers.
[01:15:54] um hyperscalers. >> As we wrap up, I'm curious for you to
[01:15:55] >> As we wrap up, I'm curious for you to riff on any other like out there
[01:15:57] riff on any other like out there knock-on effects that you've started to
[01:15:59] knock-on effects that you've started to think about for this giant trend. We've
[01:16:01] think about for this giant trend. We've talked about the specific companies in a
[01:16:02] talked about the specific companies in a lot of detail that this most impacts. We
[01:16:05] lot of detail that this most impacts. We talked a little bit about the
[01:16:05] talked a little bit about the application layer and what would have to
[01:16:07] application layer and what would have to happen for there to be more value
[01:16:08] happen for there to be more value occurring to that layer of the stack.
[01:16:10] occurring to that layer of the stack. I'm curious like any other just fun
[01:16:12] I'm curious like any other just fun knock-on things that you've been
[01:16:13] knock-on things that you've been thinking about as this world changes so
[01:16:15] thinking about as this world changes so quickly.
[01:16:16] quickly. >> Yeah. And it is wild. I mean at the
[01:16:17] >> Yeah. And it is wild. I mean at the application layer, forget value
[01:16:18] application layer, forget value acrewing, just value has been destroyed.
[01:16:20] acrewing, just value has been destroyed. >> AI has net destroyed. Even if you count
[01:16:22] >> AI has net destroyed. Even if you count cursor cognition, the most successful AI
[01:16:25] cursor cognition, the most successful AI natives, value has been trillions of
[01:16:28] natives, value has been trillions of dollars of value has been destroyed by
[01:16:30] dollars of value has been destroyed by AI at the application layer. And just in
[01:16:32] AI at the application layer. And just in this context, I do think it's a little
[01:16:34] this context, I do think it's a little it's something we need to be aware of.
[01:16:36] it's something we need to be aware of. The companies that are doing the best
[01:16:38] The companies that are doing the best today that are seeing kind of their
[01:16:41] today that are seeing kind of their values increase the most that are
[01:16:43] values increase the most that are creating economic value are the
[01:16:45] creating economic value are the companies with the highest ratio,
[01:16:48] companies with the highest ratio, highest effective ratio of utilized GPUs
[01:16:51] highest effective ratio of utilized GPUs per human.
[01:16:52] per human. >> And you know, maybe this just means that
[01:16:54] >> And you know, maybe this just means that every human's going to get a lot of
[01:16:55] every human's going to get a lot of GPUs, but I think that's an interesting
[01:16:57] GPUs, but I think that's an interesting fact that we kind of need to be
[01:16:59] fact that we kind of need to be cognizant of. I will just say and maybe
[01:17:01] cognizant of. I will just say and maybe this is a little dark. I am more more
[01:17:04] this is a little dark. I am more more and more worried about personal safety
[01:17:06] and more worried about personal safety and I worry about this a lot more for
[01:17:08] and I worry about this a lot more for people who are you know have a much
[01:17:11] people who are you know have a much bigger public presence and are much more
[01:17:12] bigger public presence and are much more associated with AI but I really worry
[01:17:15] associated with AI but I really worry about personal safety. I hope nothing
[01:17:17] about personal safety. I hope nothing tragic happens, but like there is this
[01:17:19] tragic happens, but like there is this upsurge in political violence here in
[01:17:21] upsurge in political violence here in America and as AI increasingly becomes
[01:17:24] America and as AI increasingly becomes political, I worry that's going to get
[01:17:26] political, I worry that's going to get directed at more and more AI political
[01:17:28] directed at more and more AI political leaders. You know, just whatever we can
[01:17:30] leaders. You know, just whatever we can agree, you know, whatever whatever I may
[01:17:32] agree, you know, whatever whatever I may think or may not think of open AI like I
[01:17:35] think or may not think of open AI like I think it is terrible that someone threw
[01:17:36] think it is terrible that someone threw mal malatto cocktails at Sam Alman's
[01:17:39] mal malatto cocktails at Sam Alman's house. I am worried that we are headed
[01:17:42] house. I am worried that we are headed into a higher variance,
[01:17:46] into a higher variance, higher beta,
[01:17:48] higher beta, higher risk world because of AI. And
[01:17:51] higher risk world because of AI. And that's for me as an individual and then
[01:17:53] that's for me as an individual and then you know for people who are big players
[01:17:55] you know for people who are big players on the chess board. Think about what it
[01:17:57] on the chess board. Think about what it means geopolitically like we're watching
[01:18:00] means geopolitically like we're watching the Ukrainians are really starting to
[01:18:01] the Ukrainians are really starting to win. And the reason they're winning I I
[01:18:04] win. And the reason they're winning I I think is not really because they have
[01:18:05] think is not really because they have better drones. I think they do have
[01:18:07] better drones. I think they do have better drones. That's part of it. I
[01:18:08] better drones. That's part of it. I think the reason Ukraine is really
[01:18:09] think the reason Ukraine is really winning is they have the best
[01:18:11] winning is they have the best battlefield AI outside of probably
[01:18:14] battlefield AI outside of probably America and Israel and has China has our
[01:18:19] America and Israel and has China has our adversaries begin to process that
[01:18:22] adversaries begin to process that like how do they respond? Like if the
[01:18:25] like how do they respond? Like if the United States because of its edge in AI
[01:18:28] United States because of its edge in AI um it's great if you're America but it
[01:18:32] um it's great if you're America but it is destabilizing for the rest of the
[01:18:34] is destabilizing for the rest of the world. Something I think a lot about is
[01:18:36] world. Something I think a lot about is creating a charity to just like educate
[01:18:37] creating a charity to just like educate the world on how awesome the west has
[01:18:39] the world on how awesome the west has been. Slavery was endemic to essentially
[01:18:41] been. Slavery was endemic to essentially almost every civilization and slavery
[01:18:43] almost every civilization and slavery was really ended by the British Empire.
[01:18:45] was really ended by the British Empire. Tell that story. Um but America after
[01:18:49] Tell that story. Um but America after 1945
[01:18:51] 1945 we had the nuclear bomb. No one else had
[01:18:53] we had the nuclear bomb. No one else had it. We could have controlled the world
[01:18:56] it. We could have controlled the world forever. Instead, we rebuilt Germany and
[01:18:59] forever. Instead, we rebuilt Germany and Japan and now we're America's most
[01:19:03] Japan and now we're America's most reliable allies. Israel, South Korea,
[01:19:05] reliable allies. Israel, South Korea, Japan. That's a testament to like the
[01:19:07] Japan. That's a testament to like the American spirit in our country. We
[01:19:08] American spirit in our country. We didn't take over the world. You know,
[01:19:10] didn't take over the world. You know, there were these fears, you know, that
[01:19:11] there were these fears, you know, that were documented at the time that the
[01:19:13] were documented at the time that the American generals and, you know,
[01:19:15] American generals and, you know, MacArthur was a little bit of an
[01:19:16] MacArthur was a little bit of an American emperor in Japan,
[01:19:19] American emperor in Japan, but um we're just going to take over the
[01:19:21] but um we're just going to take over the world. And they could have and they
[01:19:23] world. And they could have and they didn't. They came home, we demilitarized
[01:19:26] didn't. They came home, we demilitarized and then you had this, you know, this
[01:19:28] and then you had this, you know, this period of of great global stability
[01:19:30] period of of great global stability between, you know, it was scary. They
[01:19:31] between, you know, it was scary. They were turn America. Yeah. You had the Pax
[01:19:33] were turn America. Yeah. You had the Pax Americana.
[01:19:34] Americana. >> So maybe it's not destabilizing. Maybe
[01:19:36] >> So maybe it's not destabilizing. Maybe it leads to the another Pax Americana
[01:19:40] it leads to the another Pax Americana >> informed by our AI dominance. And I'm so
[01:19:43] >> informed by our AI dominance. And I'm so optimistic that AI is going to be
[01:19:45] optimistic that AI is going to be amazing for the world. There's someone
[01:19:47] amazing for the world. There's someone like me whose daughter was diagnosed
[01:19:49] like me whose daughter was diagnosed with a very rare mutation. there's no
[01:19:52] with a very rare mutation. there's no cure. He was able to assemble a lot of
[01:19:54] cure. He was able to assemble a lot of resources. He was able to get a lot of
[01:19:56] resources. He was able to get a lot of compute from the labs. Um we were made
[01:19:59] compute from the labs. Um we were made aware of what was happening, spun up an
[01:20:01] aware of what was happening, spun up an immense amount of agents, came up using
[01:20:05] immense amount of agents, came up using AI with a drug on the market that can
[01:20:07] AI with a drug on the market that can actually impact his daughter's disease
[01:20:09] actually impact his daughter's disease and then has spun up a company to cure
[01:20:12] and then has spun up a company to cure it.
[01:20:13] it. And like her life is already
[01:20:16] And like her life is already immeasurably different because of AI. So
[01:20:18] immeasurably different because of AI. So I'm like an AI I'm like an AI optimist
[01:20:21] I'm like an AI I'm like an AI optimist maximalist but I also just acknowledge
[01:20:23] maximalist but I also just acknowledge it's like an event horizon. It for sure
[01:20:26] it's like an event horizon. It for sure I think is going to be a discontinuity.
[01:20:28] I think is going to be a discontinuity. We need to navigate has societ as
[01:20:30] We need to navigate has societ as society. I think the lites are going to
[01:20:32] society. I think the lites are going to be wrong but we need to be like really
[01:20:34] be wrong but we need to be like really thoughtful in how we address their
[01:20:36] thoughtful in how we address their concerns. We need to make sure that it's
[01:20:38] concerns. We need to make sure that it's good for everyone. Like it is a little
[01:20:40] good for everyone. Like it is a little dystopian that now the best AI is only
[01:20:42] dystopian that now the best AI is only available to people with a lot of money.
[01:20:44] available to people with a lot of money. Like we need to solve that. We need to
[01:20:47] Like we need to solve that. We need to approach this with humility, recognize
[01:20:48] approach this with humility, recognize there's a lot of uncertainty, and be
[01:20:50] there's a lot of uncertainty, and be thoughtful.
[01:20:50] thoughtful. >> When I do this with you, I tell people
[01:20:52] >> When I do this with you, I tell people afterwards, I'm like, "May you find
[01:20:53] afterwards, I'm like, "May you find something that you love as much as Gavin
[01:20:55] something that you love as much as Gavin loves markets and companies and
[01:20:57] loves markets and companies and capitalism and history uh on display
[01:21:00] capitalism and history uh on display today as always." Gavin, thanks so much
[01:21:01] today as always." Gavin, thanks so much for your time.
[01:21:02] for your time. >> Thank you. Thanks, Patrick.
[01:21:06] >> You know how small advantages compound
[01:21:08] >> You know how small advantages compound over time? That's true in investing and
[01:21:10] over time? That's true in investing and just as true in how you run your
[01:21:11] just as true in how you run your company. Your spending system is your
[01:21:13] company. Your spending system is your capital allocation strategy. Ramp makes
[01:21:15] capital allocation strategy. Ramp makes it smarter by default. Better data,
[01:21:17] it smarter by default. Better data, better decisions, better economics over
[01:21:19] better decisions, better economics over time. See how at ramp.com/invest.
[01:21:22] time. See how at ramp.com/invest. As your business grows, Vant scales with
[01:21:23] As your business grows, Vant scales with you, automating compliance and giving
[01:21:25] you, automating compliance and giving you a single source of truth for
[01:21:27] you a single source of truth for security and risk. Learn more at
[01:21:29] security and risk. Learn more at vanta.com/invest.
[01:21:30] vanta.com/invest. Every investment firm is unique and
[01:21:32] Every investment firm is unique and generic AI doesn't understand your
[01:21:34] generic AI doesn't understand your process. Rogo does. It's an AI platform
[01:21:36] process. Rogo does. It's an AI platform built specifically for Wall Street,
[01:21:37] built specifically for Wall Street, connected to your data, understanding
[01:21:39] connected to your data, understanding your process, and producing real
[01:21:40] your process, and producing real outputs. Check them out at
[01:21:42] outputs. Check them out at rogo.ai/invest.
[01:21:44] rogo.ai/invest. The best AI and software companies from
[01:21:45] The best AI and software companies from OpenAI to Cursor to Perplexity use work
[01:21:48] OpenAI to Cursor to Perplexity use work OS to become enterprise ready overnight,
[01:21:50] OS to become enterprise ready overnight, not in months. Visit works.com to skip
[01:21:53] not in months. Visit works.com to skip the unglamorous infrastructure work and
[01:21:54] the unglamorous infrastructure work and focus on your product. Ridgeline is
[01:21:56] focus on your product. Ridgeline is redefining asset management technology
[01:21:58] redefining asset management technology as a true partner, not just a software
[01:22:00] as a true partner, not just a software vendor. They've helped firms 5x and
[01:22:02] vendor. They've helped firms 5x and scale, enabling faster growth, smarter
[01:22:04] scale, enabling faster growth, smarter operations, and a competitive edge.
[01:22:06] operations, and a competitive edge. Visit ridgelineapps.com
[01:22:08] Visit ridgelineapps.com to see what they can unlock for your
[01:22:09] to see what they can unlock for your firm.

Full Transcript

https://www.youtube.com/watch?v=Mmj_G9RlW-I
Translation: zh-CN

[00:00] 我认为人工智能领域正在发生的事情是资本主义历史上、美国商业历史上最非凡的时刻。
[00:08] Anthropic 他们增加了 110 亿美元的年经常性收入。
[00:11] 过去 10 到 12 年成立的三家知名度最高的 SAS 公司是 Palunteer、Snowflake 和 Data Bricks。
[00:21] 这三家公司花了 10 年时间建立自己的业务。
[00:23] Anthropic 在一个月内增加了他们合并的业务。
[00:30] 这在资本主义历史上是前所未有的。
[00:35] 忘掉我的职业生涯吧。
[00:35] 就说资本主义的纯粹历史，商业的历史吧。
[00:53] 好的。
[00:53] 所以，这是我们第六次做这件事了，你敢信吗，这让你回到了第一名，或者至少是和 Girly 并列第一。
[01:01] 至少与Girly并列第一。
[01:03] 而且我认为，自从上次我们做这件事以来，那真是令人兴奋和壮观，我认为我们现在正处于一个更有趣的时期。
[01:10] 也许我们可以先谈谈今年三月和四月的生活感受，对我来说，那就像一个完全独特的经济、技术和市场环境。
[01:22] 你是历史和这些时代最伟大的研究者。
[01:24] 所以感觉如何？
[01:25] 总的来说，我认为有两种类型的下跌。
[01:27] 一种是你错了，公司估计错误，你的假设被证伪，你不得不接受损失并确认亏损。
[01:41] 然后是下跌或表现不佳的时期，你表现不佳是因为你非常了解的公司，并且你强烈反对价格走势，你可以坚持下去，而不是确认负面表现，你可以积累潜在的阿尔法，未来的表现，对我来说，这就是
[02:03] 表现，对我来说，三月就是这样的感觉。
[02:05] 感觉就像，嗯，你三月感觉就像这样。
[02:08] 感觉就像，嗯，你知道纳斯达克在抛售，同时人工智能领域发生的事情，我认为是资本主义历史上，美国商业历史上最非凡的时刻，我之所以这么说，是因为Anthropic他们增加了110亿美元的年经常性收入，而令我惊讶的是，SaaS和云计算革命创造了，我们称之为5到10万亿美元的价值，我想说，可以说是过去10到12年里成立的三个最高调的SaaS公司是Palunteer、Snowflake和Data Bricks。
[02:51] 这三家公司雇佣了成千上万的人，总共数万人。
[02:54] 他们都花了10年时间建立自己的业务。
[02:57] 而Anthropic在一个月内增加了他们合并的业务。
[03:00] 月。
[03:02] 从来没有过这样的事情。
[03:04] 这种情况在资本主义的历史上是前所未有的。
[03:06] 发生在资本主义的历史上。
[03:09] 忘掉我的职业生涯吧。就资本主义的历史，商业的历史而言。
[03:12] 我疯了。然后克里希纳来到这个节目，分享了一些数据。
[03:15] 在DR方面增长了500%。
[03:19] 是的。你算算这三年的账。太疯狂了。
[03:22] 我们，所以这根本没有先例可循。
[03:25] 我们，你知道，科技科技投资者，你会听到很多关于S曲线和投资指数增长的讨论。
[03:28] 我从未见过这样的指数增长。
[03:34] 它感觉比Deepseek还要极端，而Deepseek的设置非常相似。
[03:36] 如果我们回到25，Deepseek出现了大规模抛售，
[03:41] 这很奇怪，因为论文在Deepseek周一发布前7天就已发表。
[03:43] 我记得是在一个美国的假日周一发表的。
[03:54] 我读了它，我想，嗯，你知道，这感觉可能不会对……你知道，AI……读起来那么积极。
[04:07] 那对，嗯，你知道的，人工智能交易来说是积极的。
[04:10] 你我采取了行动。
[04:10] 我们有 DeepSeek 星期一，人工智能在那之后一周真的崩溃了，这真的很奇怪。
[04:18] 因为到了 DeepSec 星期一，很明显这将是对计算需求最积极的事情。
[04:25] 亚洲 AWS 可用区域的价格已经翻了一番。
[04:31] 你看到 GPU 的可用性在下降。
[04:34] 这只是我们第一次看到推理模型在推理过程中比非推理模型更需要计算能力。
[04:46] 所以那是一个类似的情况，但你必须做一些工作才能看到这一点。
[04:51] 我的意思是，并不难说，哦，哇，股票正在下跌。
[04:55] DRAM 的价格正在飞涨。
[04:57] 亚洲 GPU 的价格正在飞涨。
[05:00] GPU 的可用性正在下降。
[05:02] 然后大约两三天后，你知道的，美国 GPU 的价格开始上涨，GPU 租金价格。
[05:07] 你所要做的就是
[05:09] 上涨，GPU租金价格。
[05:13] 三月份你所要做的就是简单地观察Anthropic发生了什么。
[05:14] 有这么多人似乎后悔了，
[05:16] 后悔，你知道，没有在22年购买，没有在COVID期间购买，没有在Deep Seek期间购买。
[05:22] 四月初，你面临着同样的估值设定和一个更清晰的人工智能拐点。
[05:32] 所以有这么多的机会投资人工智能，当然，使事情复杂化的是直率的。
[05:39] 首先，我成了一个信徒，并且是一个信徒，我认为市场可能在错误定价一件事。
[05:47] 我不是宏观经济专家，我确实做了很多关于国家安全的投资。
[05:53] 所以我确实能接触到那些专家，他们很乐意与我分享他们的想法和观点。
[06:02] 荷姆海峡的关闭对美国来说实际上是相对有利的。
[06:05] 对美国来说是极好的。
[06:07] 为什么？
[06:08] 因为特别是为了...的目标
[06:10] 因为特别是为了当前政府的目标。
[06:12] 当前政府。所以电力是一个非常重要的工业或制造业投入。
[06:17] 进入美国电力价格的关键投入，而这又影响到人工智能，是在G1天然气上，根据彭博社的数据，天然气价格下跌了20%。
[06:26] 而亚洲、欧洲以及其他所有地方的天然气价格都翻了一番或翻了两番。
[06:35] 所以我们相对的制造业竞争力一夜之间就提高了，无论好坏，这似乎是特朗普政府关心的。
[06:43] 他们非常关注美国的相对地位。
[06:48] 我想很多人还记得20世纪70年代。
[06:53] 而70年代之所以如此痛苦，不仅仅是因为价格上涨，更是因为出现了实际的天然气短缺。
[07:01] 然后你再看，好吧，美国的经济比以前的能源密集度大大降低了。
[07:07] 美国经济，美国现在是世界上最大的石油和天然气生产国，我们已经成为
[07:12] 石油和天然气生产商，现在我们已成为世界上最大的石油出口国。
[07:15] 现在是世界上最大的石油和天然气出口国，在此之上还有。
[07:18] 和天然气，在此之上还有这种相对的制造优势。
[07:21] 这种相对的制造优势，我认为这使得更容易。
[07:26] 因此，我认为这使得更容易专注于人工智能的基本面，保持。
[07:30] 专注于人工智能的基本面，专注于历史上。
[07:33] 专注于历史上具有吸引力的估值，我认为在。
[07:35] 有吸引力的估值，我认为在相对基础上。
[07:37] 相对基础上，科技基本上变得像它一样便宜。
[07:38] 科技基本上变得像它与市场上其他部分相比一样便宜。
[07:41] 在过去10年中的任何时候都比市场上其他部分便宜，并且。
[07:44] 在过去10年中的任何时候，并且只是从...的角度来考虑它。
[07:45] 只是从市场效率的角度来考虑它。我们拥有最。
[07:47] 市场效率。我们拥有历史上最非凡的时刻。
[07:49] 资本主义历史上非凡的时刻，这对人工智能来说是极度看好的。
[07:53] 资本主义，这对人工智能来说是极度看好的，而你有机会购买人工智能。
[07:56] 而你有机会以非常诱人的估值购买人工智能。你对此怎么看。
[07:59] 以非常诱人的估值。你如何看待那些。
[08:01] 你如何看待特别是Anthropic和OpenAI的倍数，它们。
[08:03] 特别是Anthropic和OpenAI，在我看来，它们就像参考资产。
[08:05] 在我看来，它们就像参考资产，是对此趋势最纯粹的演绎。
[08:07] 对此趋势最纯粹的演绎，真的不是那么疯狂？
[08:10] 这个趋势真的不是那么疯狂？就像如果你只看销售额。
[08:11] 就像如果你只看销售倍数并将其与可能是什么进行比较。
[08:13] 多个，并将其与数据砖和雪花以及这些公司进行比较。
[08:15] 数据砖和雪花以及这些公司在其高峰期交易，你如何理解它？
[08:18] 我认为OpenAI和Anthropic在资本效率方面是截然不同的。
[08:19] OpenAI和Anthropic在资本效率方面是截然不同的。
[08:21] 而且Anthropic每token的成本显然比OpenAI低得多。
[08:24] 而且Anthropic每token的成本显然比OpenAI低得多。
[08:26] 它们就是如此。
[08:30] 你可以从他们为了达到大致相似的收入规模而烧掉的钱的数量中看出这一点。
[08:31] 你可以从他们为了达到大致相似的收入规模而烧掉的钱的数量中看出这一点。
[08:35] 我认为他们可能烧掉了比OpenAI少80%的钱。
[08:37] 我认为他们可能烧掉了比OpenAI少80%的钱。
[08:39] 比OpenAI。
[08:41] 因此，作为企业，它们显然具有非常不同的结构性股本回报率。
[08:44] 因此，作为企业，它们显然具有非常不同的结构性股本回报率。
[08:46] 我认为OpenAI正在做很多事情。
[08:48] 我认为Sarah Frier是最杰出的CFO之一。
[08:49] CFO。
[08:51] 我认为他们正在做很多事情来尝试改进这一点。
[08:53] 并且他们已经获得了大量的计算资源。
[08:55] 并且他们已经获得了大量的计算资源。
[08:55] 比他们获得的计算资源多得多。
[08:57] 这是另一个很大的区别。
[08:59] 嗯，事实证明激进确实得到了回报，但是的，我只是认为Anthropic以9000亿美元的估值，500亿美元的年收入，你知道，年收入，你知道我。
[09:02] 激进确实得到了回报，但是的，我只是认为Anthropic以9000亿美元的估值，500亿美元的年收入，你知道，年收入，你知道我。
[09:05] 增长了百分之九百。
[09:09] 十亿美元，你知道年收入，你知道我。
[09:10] 是的，以惊人的速度增长。
[09:13] 也许一个真实的说法是，如果Anthropic
[09:15] 一个真实的说法是，如果Anthropic拥有所有的计算能力，他们今天可能在一百亿以上，也许是一千五百亿。
[09:26] 而且我确实知道，他们已经明显降低了Claude的智能。
[09:31] 有一项分析显示，Claude在Opus上为完全相同的问题生成的token数量减少了70%。
[09:37] 而且你知道，正如我们上次讨论的，token的数量在某种程度上等于答案的质量和思考的质量。
[09:43] 你知道，每个token的智能密度也很重要，你知道，我认为我作为用户感受到了这一点，所以我认为他们会做得更好，1000亿、1500亿，也许2000亿，所以你可能会以大约五倍的未受限制的价格购买它。
[09:59] 我将编造一个新数字，呃，未受限制的运行收入，是的。
[10:09] 你为什么认为他们不以3万亿美元的估值或类似的价格筹集1000亿美元呢？
[10:14] 就像如果你是
[10:16] 类似这样。
[10:18] 就像如果你是Anthropic的首席财务官，呃Krishna是Anthropic的首席财务官，呃Krishna很棒。
[10:19] 我们刚请了他来。
[10:21] 或者如果你是Open，如果你是Sarah，当然，如果我收到的关于Krishna那一集的入站信息有任何指示，我遇到的每个人都在试图投资这两家公司。
[10:24] 关于Krishna那一集的入站信息有任何指示，我遇到的每个人都在试图投资这两家公司。
[10:27] 指示，我遇到的每个人都在试图投资这两家公司。
[10:28] 试图投资这两家公司。
[10:29] 公司。
[10:33] 所以我觉得明智的做法是，未来是不确定的。
[10:36] 未来是不确定的。
[10:38] 你显然是在一个非常资本密集的游戏中，即使你是，你知道，Enthropic
[10:41] 密集的游戏，即使你是，你知道，Enthropic
[10:43] Enthropic，嗯，我敢肯定它今天在推理方面有非常高的毛利率，我认为它今年可能会开始产生现金，如果它们还没有产生现金的话，而我认为这很可能是事实。
[10:45] 嗯，我敢肯定它今天在推理方面有非常高的毛利率，我认为它今年可能会开始产生现金，如果它们还没有产生现金的话，而我认为这很可能是事实。
[10:48] 推理方面的毛利率，我认为它今年可能会开始产生现金，如果它们还没有产生现金的话，而我认为这很可能是事实。
[10:49] 可能会开始产生现金，如果它们还没有产生现金的话，而我认为这很可能是事实。
[10:51] 年，如果它们还没有产生现金的话，而我认为这很可能是事实。
[10:55] 现金，而我认为这很可能是事实，但你仍然可能想能够筹集更多资本，获得更多计算能力，世界是不确定的，乌克兰
[10:57] 但你仍然可能想能够筹集更多资本，获得更多计算能力，世界是不确定的，乌克兰
[10:59] 筹集更多资本，获得更多计算能力，世界是不确定的，乌克兰开始真正地真正地获胜，俄罗斯将如何回应，你知道，我认为伊朗仍然存在很多不确定性。
[11:01] 开始真正地真正地获胜，俄罗斯将如何回应，你知道，我认为伊朗仍然存在很多不确定性。
[11:04] 俄罗斯将如何回应，你知道，我认为伊朗仍然存在很多不确定性。
[11:06] 我认为伊朗仍然存在很多不确定性。
[11:07] 在伊朗。我认为所有这些不确定性可能加剧台湾的地缘政治不确定性。
[11:10] 我认为所有这些不确定性可能加剧台湾的地缘政治不确定性。
[11:12] 可能加剧台湾的地缘政治不确定性。
[11:14] 对台湾的不确定性。所以，这是一个不确定的世界。如果，如果我想到
[11:16] 不确定的世界。如果，如果我想起埃隆，埃隆总是能让投资者赚钱。
[11:19] 埃隆，埃隆总是能让投资者赚钱。他将其视为神圣的契约。
[11:21] 钱。他将其视为神圣的契约。因此，因为他过去20年一直为人们赚钱，所以他拥有超能力。
[11:24] 契约。因此，因为他过去20年一直为人们赚钱，所以他拥有超能力。那就是他可以根据自己的意愿，在任何时候筹集到他想要的任何数额的资金。
[11:27] 为人们赚钱已有20年，他拥有超能力。那就是他可以根据自己的意愿，在任何时候筹集到他想要的任何数额的资金。
[11:29] 拥有超能力。那就是他可以根据自己的意愿，在任何时候筹集到他想要的任何数额的资金。我认为这些公司采取的做法是明智的，我不知道他们是否这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。
[11:32] 可以根据自己的意愿，在任何时候筹集到他想要的任何数额的资金。我认为这些公司采取的做法是明智的，我不知道他们是否这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。
[11:35] 想要的时候，随时都可以。我认为这些公司采取的做法是明智的，我不知道他们是否这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。
[11:37] 这些公司采取的做法是明智的，我不知道他们是否这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。
[11:38] 采取，我不知道他们是否这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。
[11:41] 这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。
[11:45] 专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。它们可以持续未来20到30年。
[11:49] 仅仅持续一两年。它们可以持续未来20到30年。
[11:52] 仅仅持续一两年。它们可以持续未来20到30年。
[11:54] 可以持续未来20到30年。>> 埃隆做到这一点的方式是系统性地低估SpaceX或其他任何东西的价格。
[11:57] >> 埃隆做到这一点的方式是系统性地低估SpaceX或其他任何东西的价格。实际方法是什么？
[11:59] 系统性地低估SpaceX或其他任何东西的价格。实际方法是什么？就是从不在估值上贪婪，从不推高估值。
[12:00] 任何其他东西。实际方法是什么？就是从不在估值上贪婪，从不推高估值。
[12:02] 方法？就是从不在估值上贪婪，从不推高估值。>> 就这么简单。
[12:04] 就是从不在估值上贪婪，从不推高估值。>> 就这么简单。
[12:06] 从不推高估值。>> 就这么简单。>> 你知道，我的朋友安东尼奥指出，SpaceX的年复合增长率，你知道，大约是30%，持续了十年。
[12:08] >> 就这么简单。>> 你知道，我的朋友安东尼奥指出，SpaceX的年复合增长率，你知道，大约是30%，持续了十年。
[12:09] >> 你知道，我的朋友安东尼奥指出，SpaceX的年复合增长率，你知道，大约是30%，持续了十年。而这仅仅是因为埃隆，我
[12:12] SpaceX的年复合增长率，你知道，大约是30%，持续了十年。而这仅仅是因为埃隆，我
[12:16] 年，持续了十年。而这仅仅是因为埃隆，我
[12:18] 这仅仅是因为埃隆，我认为，专注于维护超级大国，并试图在投资者和员工之间取得公平的平衡。
[12:26] 但我我认为这是明智的。
[12:28] 但Anthropic能否以至少100%的溢价筹集资金，达到这个传闻中的最新估值？
[12:38] 当然，大多数软件公司都试图最大化你在他们应用程序上的时间，以提高参与度。
[12:41] RAMP则恰恰相反。
[12:43] RAMP明白没有人想花几个小时去追逐收据、审查费用报告和检查政策违规行为。
[12:51] 所以，他们构建了他们的工具，利用人工智能将85%的费用审查自动化，准确率达到99%，从而将这些时间还给用户。
[12:59] 而且由于Ramp为公司节省了5%的成本，难怪Shopify、Stripe以及我的公司都在使用RAM。
[13:05] 想看看当你消除繁琐的工作时会发生什么，请访问ramp.com/invest。
[13:10] Rogo的Felix是一款个人理财助手，它能将一个简单的提示，利用你公司的模板、上下文和标准，转化为客户准备就绪的最终工作成果。
[13:17] 给Felix发一封电子邮件，比如，
[13:19] 标准。给菲利克斯发一封邮件，说：“把这些评论转给我，或者用这些邮件的上下文更新我的跟踪器。”
[13:25] 或者，“对这个买家运行支付能力数学计算，然后菲利克斯会发回完成的PowerPoint演示文稿、Excel模型和搜集到的研究资料。”
[13:31] 菲利克斯的工作方式就像你们团队已经做的那样，全天候快速准确地交付工作。
[13:36] 了解更多信息，请访问rogo.ai/felix。
[13:40] OpenAI、Cursor、Enthropic、Perplexity和Versell都有一个共同点。
[13:44] 它们都使用WorkOS。
[13:47] 原因如下。
[13:49] 为了实现大规模的企业采用，您必须提供核心功能，如SSO、skim、arbback和审计日志。
[13:55] 这就是Work OS的作用。
[13:57] 与其花费数月时间自己构建这些关键任务功能，不如直接使用Work OS API，在第一天就获得所有这些功能。
[14:02] 这就是为什么您听说过的许多顶级AI团队都已经运行在Work OS上。
[14:07] Work OS是为企业做好准备并专注于最重要事项——您的产品的最快途径。
[14:13] 访问works.com开始使用。
[14:15] 让我们来谈谈讨论中关于沃森晶圆的部分。
[14:19] 这总是我最喜欢和你谈论的事情。
[14:19] 呃，重要性
[14:21] 和你谈谈。
[14:21] 呃，这次基础设施建设的重要性。
[14:24] 我觉得这次基础设施建设。
[14:24] 我觉得好像每次都感觉它过热了，然后下次和你谈话时，似乎我们应该做得比实际做得多得多。
[14:27] 过热了，然后下次和你谈话时，似乎我们应该做得比实际做得多得多。
[14:29] 你研究了S曲线和那些S曲线的陡峭程度很多。
[14:33] 呃，而且你对历史也很了解。
[14:35] 给我们讲讲你今天如何看待沃森晶圆是这一切的关键输入。
[14:42] 是的，我想说我认为资本主义将解决瓦特短缺问题，除非有大的监管政治抵制。
[14:49] 我认为这是很有可能发生的。
[14:53] 一家大型私募股权公司数据中心基础设施投资的负责人。
[14:58] 你知道，我认为黑石、阿波罗、KKR说，过去能源和芯片是我们最大的限制因素。
[15:04] 现在，分区和审批变得更加重要了。
[15:10] 而且我认为很多公司在采取行动之前会等到中期选举之后，比如可能裁员。
[15:17] 没有人想成为，你知道，在...期间的替罪羊。
[15:22] 要成为，你知道的，中期选举中的皮纳塔。
[15:24] 中期选举，但你知道，你已经看到很多制造涡轮机的公司很重要。
[15:26] 但你知道，你已经看到很多制造涡轮机的公司宣布了大幅提高产能的计划。
[15:30] 宣布了大幅提高产能的计划。
[15:32] 有大约两台这样的机器可以铸造这些大型叶片。
[15:34] 这些机器可以铸造这些大型叶片。
[15:37] 我们在西方已经80年没有制造过一个了。
[15:39] 我们不知道如何再制造它们了，等等，等等，等等。
[15:42] 所有这些都是真的。
[15:44] 我绝不是在轻视，你知道的，工业工程，你知道的，那些东西所包含的魔法和艺术性，但资本主义非常擅长随着时间的推移解决这类问题。
[15:49] 那些东西所包含的艺术性，但资本主义非常擅长随着时间的推移解决这类问题。
[15:52] 除了这些涡轮机之外，还有其他能源，而且时间跨度更长。
[15:54] 所以我认为瓦特短缺可能会在27、28年开始缓解，然后我认为轨道计算将真正解决这个问题。
[15:59] 我确实想重新定义轨道计算，因为我认为当人们听到太空中的数据中心时，就像我们在上期节目中讨论的那样，他们会想象一个五角大楼大小的太空建筑。
[16:09] 他们会说，嗯，我们不能做
[16:23] 在太空中。
[16:23] 他们就像，嗯，我们不能那样做。
[16:23] 那不是它。
[16:28] 一个布莱克韦尔那个。
[16:28] 那不是它。
[16:28] 一个布莱克韦尔机架重 3000 磅。
[16:31] 它是 8 英尺的机架重 3000 磅。
[16:31] 它高 8 英尺。
[16:31] 它深 4 英尺。
[16:31] 宽 3 英尺。
[16:35] 高。
[16:35] 它深 4 英尺。
[16:35] 宽 3 英尺。
[16:35] 它是太空中的机架。
[16:38] SpaceX 有它在太空中的机架。
[16:38] SpaceX 向你展示了一张插图，它是一个
[16:41] 展示了一张插图，它是一个机架。
[16:41] 那就是卫星。
[16:41] 呃，但它可能和布莱克韦尔差不多大
[16:44] 机架。
[16:44] 那就是卫星。
[16:44] 呃，但它可能和布莱克韦尔差不多大
[16:46] 可能差不多是一个布莱克韦尔机架的大小。
[16:46] 它有这些太阳能翼，它们是
[16:48] 机架。
[16:48] 它有这些太阳能翼，每边可能长 500 英尺。
[16:51] 可能每边长 500 英尺。
[16:51] 你把它放在太阳同步轨道上。
[16:54] 放在太阳同步轨道上。
[16:54] 所以那些太阳能电池板总是在
[16:56] 那些太阳能电池板总是在阳光下。
[16:56] 然后因为它在一个精确的
[16:58] 阳光下。
[16:58] 然后因为它在一个精确的太阳同步轨道上，散热器它
[17:01] 太阳同步轨道上，散热器它延伸在它后面数百英尺。
[17:05] 延伸在它后面数百英尺。
[17:05] >>这是一个常见的批评。
[17:05] 是的。
[17:06] 如何
[17:06] 你要去。
[17:07] >>我花了很多时间在星港
[17:11] 这些年来，我和很多
[17:13] SpaceX 的工程师谈过，我认为它是
[17:16] 最才华横溢的工程师团队
[17:19] 地球上，他们非常有信心
[17:21] 他们已经解决了这个问题，而且他们没有
[17:23] 他们已经解决了这个问题，而且他们并不总是那么自信，就像我认为可能的那样。
[17:27] 总是自信，就像我认为可能的那样，你知道，需要做一些工程才能将星舰变成火星殖民运输船。
[17:30] 需要做一些工程才能将星舰变成火星殖民运输船。
[17:32] 他们会这样做吗？绝对会。
[17:34] 他们更关注什么？我想可能是，你知道，维修和保养。
[17:37] 更关注什么？我想可能是，你知道，维修和保养。
[17:39] 这是两个大的，你知道，两个大的响应，散热器以及如何修复机架中出现的任何问题。
[17:40] 大的响应，散热器以及如何修复机架中出现的任何问题。
[17:42] 如何修复机架中出现的任何问题。
[17:44] 答案是，直到你可能拥有，你知道，浮动的优化器。
[17:45] 在机架中出错了。答案是，直到你可能拥有，你知道，浮动的优化器。
[17:47] 你不会。现在，我认为星舰将以我们无法想象的方式改变太空经济。
[17:50] 优化器。你不会。现在，我认为星舰将以我们无法想象的方式改变太空经济。
[17:52] 星舰将以我们无法想象的方式改变太空经济。特别是如果监管成为数据中心的限制，这一切都将无关紧要。
[17:54] 经济，以我们无法想象的方式。特别是如果监管成为数据中心的限制，这一切都将无关紧要。
[17:56] 特别是如果监管成为数据中心的限制，这一切都将无关紧要。你将销售你能制造的尽可能多的轨道计算能力。
[17:57] 对数据中心的限制，这一切都将无关紧要。你将销售你能制造的尽可能多的轨道计算能力。
[17:59] 将无关紧要。你将销售你能制造的尽可能多的轨道计算能力。然后显然，你将使用激光在真空中连接这些机架，这些激光已经存在于每个星链上。
[18:01] 尽可能多的轨道计算能力。然后显然，你将使用激光在真空中连接这些机架，这些激光已经存在于每个星链上。
[18:04] 然后显然，你将使用激光在真空中连接这些机架，这些激光已经存在于每个星链上。对我来说，这简直令人难以置信，SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。
[18:07] 使用激光在真空中旅行，这些激光已经存在于每个星链上。对我来说，这简直令人难以置信，SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。
[18:10] 已经存在于每个星链上。对我来说，这简直令人难以置信，SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。
[18:11] 对我来说，这简直令人难以置信，SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。
[18:15] 简直令人难以置信，SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。
[18:17] SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。每一个
[18:24] 在轨卫星的总数。
[18:24] 每一个星链，他们今天都在降温。
[18:28] 星链，他们今天都在降温。
[18:28] 而且你知道，我认为星链V3将运行在20千瓦。
[18:33] 一个 Blackwell 机架只有100千瓦。
[18:37] 人们谈论了很多关于密度的问题。
[18:39] 嗯，如果你通过真空用激光连接机架，你知道，你可以让机架在物理上更大。
[18:45] 你关注的是重量，而不是大小。
[18:49] 在地球上的数据中心，你试图连接机架，理想情况下使用铜线，尽量缩短长度等等。
[18:53] 布线是一笔巨大的开销。
[18:56] 嗯，你确实希望那个机架小一些，因为你知道，在可能的情况下使用铜线，在必须的情况下使用光纤。
[19:01] 但在太空，你知道，SpaceX 可以做各种各样的事情，我认为也许一些反对者没有考虑到，但他们只是运营的卫星比你想要的要多。
[19:12] 他们今天有一个20千瓦的卫星，所以也许你一开始就将其扩展到60千瓦。
[19:18] 他们似乎非常有信心，将直接达到100到120。
[19:23] 而且他们也是同一家公司，现在也运营着最大的
[19:26] 公司现在还运营着地球上最大的数据中心。
[19:28] 他们拥有地球上的数据中心。
[19:31] 他们拥有世界上最好的硬件工程师和各种各样的人，几乎所有这些人都不够聪明或不够实用，无法在SpaceX工作，他们是这些纸上谈兵的怀疑论者。
[19:37] 不够聪明或不够实用，无法在SpaceX工作，他们是这些纸上谈兵的怀疑论者。
[19:40] 在SpaceX工作，他们是这些纸上谈兵的怀疑论者。
[19:42] 怀疑论者。
[19:43] 你知道，我不想引用拉里·埃里森的话，但有人，你知道，持怀疑态度，而拉里，拉里就像说，“听着，他正在那里着陆火箭。
[19:47] 持怀疑态度，而拉里，拉里就像说，“听着，他正在那里着陆火箭。
[19:49] 就像说，“听着，他正在那里着陆火箭。
[19:51] 着陆火箭。
[19:53] 我没看到其他人着陆火箭。
[19:56] 而现实是，10年后，没有其他公司能够持续地着陆和完全重复使用轨道火箭。
[19:59] 能够持续地着陆和完全重复使用轨道火箭。
[20:03] 并完全重复使用轨道火箭。
[20:05] 而这一切的运作如果没有可重用性就没有意义。
[20:06] 这意味着你必须着陆它。
[20:08] 我想重新定义轨道计算，它在太空中拥有机架，而不是在太空中拥有巨大的漂浮的五角形数据中心，这简直是，你知道，这很愚蠢。
[20:14] 在太空中漂浮的五角形数据中心，这简直是，你知道，这很愚蠢。
[20:17] 在太空中数据中心，这简直是，你知道，这很愚蠢。
[20:19] 但你可以，你知道，使数据中心成为数据中心的是你用激光连接这些机架。
[20:23] 用激光连接这些机架。
[20:24] 所以它将是在太空中用激光连接成虚拟数据中心的机架。
[20:26] 用激光连接成虚拟数据中心。
[20:26] 而如果你想到

Full Transcript (Bilingual)

https://www.youtube.com/watch?v=Mmj_G9RlW-I
Translation: zh-CN

[00:00] What was happening in AI was I think the most extraordinary moment in the history of capitalism, the history of American business.
我认为人工智能领域正在发生的事情是资本主义历史上、美国商业历史上最非凡的时刻。

[00:08] Anthropic they added 11 billion of ARR.
Anthropic 他们增加了 110 亿美元的年经常性收入。

[00:11] The three highest profile SAS companies founded in the last 10 12 years are Palunteer, Snowflake and Data Bricks.
过去 10 到 12 年成立的三家知名度最高的 SAS 公司是 Palunteer、Snowflake 和 Data Bricks。

[00:21] And these three companies spent 10 years building their businesses.
这三家公司花了 10 年时间建立自己的业务。

[00:23] Anthropic added their combined businesses in one month.
Anthropic 在一个月内增加了他们合并的业务。

[00:30] That's just nothing like that has ever happened in the history of capitalism.
这在资本主义历史上是前所未有的。

[00:35] Forget my career.
忘掉我的职业生涯吧。

[00:35] Just the flatout history of capitalism, the history of business.
就说资本主义的纯粹历史，商业的历史吧。

[00:53] All right.
好的。

[00:53] So, this is our uh sixth time doing this, if you can believe it, which puts you back into first place uh or at least tied for first place with Girly.
所以，这是我们第六次做这件事了，你敢信吗，这让你回到了第一名，或者至少是和 Girly 并列第一。

[01:01] Least tied for first place with Girly.
至少与Girly并列第一。

[01:03] And I think even since last time when we did this, which was so exciting and spectacular, I think we're in an even more interesting time now.
而且我认为，自从上次我们做这件事以来，那真是令人兴奋和壮观，我认为我们现在正处于一个更有趣的时期。

[01:10] Maybe just start by riffing on how it felt for you living through March and April of this year, which which felt to me just like a completely unique economic, technology, and market environment.
也许我们可以先谈谈今年三月和四月的生活感受，对我来说，那就像一个完全独特的经济、技术和市场环境。

[01:22] And you're the biggest student of of of history and of these times.
你是历史和这些时代最伟大的研究者。

[01:24] So what does it feel like?
所以感觉如何？

[01:25] I would say broadly speaking there are two kinds of draw downs.
总的来说，我认为有两种类型的下跌。

[01:27] They're drawowns where you're wrong, a company misestimates, your hypothesis was invalidated and you have to take your medicine and you crystallize that loss.
一种是你错了，公司估计错误，你的假设被证伪，你不得不接受损失并确认亏损。

[01:41] And then there are draw downs or periods of underperformance where you you're underperforming because of companies you know really really well and where you profoundly disagree with the price action and you can lean in and instead of crystallizing uh negative performance you can kind of build pent up alpha pent up future performance and for me that is what
然后是下跌或表现不佳的时期，你表现不佳是因为你非常了解的公司，并且你强烈反对价格走势，你可以坚持下去，而不是确认负面表现，你可以积累潜在的阿尔法，未来的表现，对我来说，这就是

[02:03] performance and for me that is what March felt like.
表现，对我来说，三月就是这样的感觉。

[02:05] It felt like uh you March felt like.
感觉就像，嗯，你三月感觉就像这样。

[02:08] It felt like uh you know the NASDAQ was selling off and at the same time what was happening in AI was I think the most extraordinary moment in the history of capitalism the history of American business and what I just mean by that is that anthropic they added 11 billion of ARR and what is astonishing to me about this is that the SAS and cloud revolution it created we'll call it between 5 and 10 trillion dollars of value and I would say Arguably the three highest profile SAS companies to have kind of been founded in the last 10 12 years are Palunteer, Snowflake and Data Bricks.
感觉就像，嗯，你知道纳斯达克在抛售，同时人工智能领域发生的事情，我认为是资本主义历史上，美国商业历史上最非凡的时刻，我之所以这么说，是因为Anthropic他们增加了110亿美元的年经常性收入，而令我惊讶的是，SaaS和云计算革命创造了，我们称之为5到10万亿美元的价值，我想说，可以说是过去10到12年里成立的三个最高调的SaaS公司是Palunteer、Snowflake和Data Bricks。

[02:51] And these three companies have spent employ thousands of people, tens of thousands collectively.
这三家公司雇佣了成千上万的人，总共数万人。

[02:54] They've all spent 10 years building their businesses.
他们都花了10年时间建立自己的业务。

[02:57] And Anthropic added their combined businesses in one month.
而Anthropic在一个月内增加了他们合并的业务。

[03:00] month.
月。

[03:02] That's just nothing like that has ever
从来没有过这样的事情。

[03:04] That's just nothing like that has ever happened in the history of capitalism.
这种情况在资本主义的历史上是前所未有的。

[03:06] happened in the history of capitalism.
发生在资本主义的历史上。

[03:09] Forget my career. Just the flatout history of capitalism, the history of business.
忘掉我的职业生涯吧。就资本主义的历史，商业的历史而言。

[03:12] I wild. And then Krishna comes on this show and shares some stats.
我疯了。然后克里希纳来到这个节目，分享了一些数据。

[03:15] 500% in DR.
在DR方面增长了500%。

[03:19] Yeah. You do the math on that for three years. Insanity.
是的。你算算这三年的账。太疯狂了。

[03:22] We So there's just no precedent for this.
我们，所以这根本没有先例可循。

[03:25] And we, you know, tech tech investors, you hear a lot of discussions about S-curves and investing in exponentials.
我们，你知道，科技科技投资者，你会听到很多关于S曲线和投资指数增长的讨论。

[03:28] I've just never seen an exponential like this.
我从未见过这样的指数增长。

[03:34] It felt even more extreme than Deepseek, which was a very similar setup.
它感觉比Deepseek还要极端，而Deepseek的设置非常相似。

[03:36] If we go back to 25 and there's a huge sell-off on Deepseek,
如果我们回到25，Deepseek出现了大规模抛售，

[03:41] which was very strange because the paper gets published 7 days before Deepseek Monday.
这很奇怪，因为论文在Deepseek周一发布前7天就已发表。

[03:43] got published, I believe, on a Monday that was a holiday in America.
我记得是在一个美国的假日周一发表的。

[03:54] And I read it, I thought, hm, you know, this this feels like it might not read that positively for um you know, the AI
我读了它，我想，嗯，你知道，这感觉可能不会对……你知道，AI……读起来那么积极。

[04:07] That positively for, um, you know, the AI trade.
那对，嗯，你知道的，人工智能交易来说是积极的。

[04:10] You I took action.
你我采取了行动。

[04:10] We had DeepSeek Monday where AI really imploded a week later and that was really strange.
我们有 DeepSeek 星期一，人工智能在那之后一周真的崩溃了，这真的很奇怪。

[04:18] Because by DeepSec Monday it was super clear that this was going to be the most positive thing that had ever happened to compute demand.
因为到了 DeepSec 星期一，很明显这将是对计算需求最积极的事情。

[04:25] Prices in the AWS available availability zones in Asia had already like doubled.
亚洲 AWS 可用区域的价格已经翻了一番。

[04:31] You were seeing GPU availability go down.
你看到 GPU 的可用性在下降。

[04:34] And this was just the first time we saw how much more compute-hungry reasoning models are during inference than non-reasoning models.
这只是我们第一次看到推理模型在推理过程中比非推理模型更需要计算能力。

[04:46] And so that was a similar setup, but you had to do some work to see that.
所以那是一个类似的情况，但你必须做一些工作才能看到这一点。

[04:51] I mean, it's not that hard to say, oh wow, stocks are selling off.
我的意思是，并不难说，哦，哇，股票正在下跌。

[04:55] The price of DRAM is going vertical.
DRAM 的价格正在飞涨。

[04:57] The price of GPUs in Asia going vertical.
亚洲 GPU 的价格正在飞涨。

[05:00] GPU availability is going down.
GPU 的可用性正在下降。

[05:02] And then like two or three days later, you know, GPU prices in in America started going up, GPU rental prices.
然后大约两三天后，你知道的，美国 GPU 的价格开始上涨，GPU 租金价格。

[05:07] All you had to do
你所要做的就是

[05:09] Up, GPU rental prices.
上涨，GPU租金价格。

[05:13] All you had to do in in March was just simply observe what was happening to Anthropic.
三月份你所要做的就是简单地观察Anthropic发生了什么。

[05:14] There there's all these people who seem to regret,
有这么多人似乎后悔了，

[05:16] regret, you know, not buying during '22, not buying during COVID, not buying during Deep Seek.
后悔，你知道，没有在22年购买，没有在COVID期间购买，没有在Deep Seek期间购买。

[05:22] You had the same valuation setup at the beginning of April and an even clearer AI inflection.
四月初，你面临着同样的估值设定和一个更清晰的人工智能拐点。

[05:32] And so there have been all these chances to buy into AI and then of course what complicated it was the straight of.
所以有这么多的机会投资人工智能，当然，使事情复杂化的是直率的。

[05:39] Foremost I became a believer and am a believer that I think maybe one thing that the market was mispricing.
首先，我成了一个信徒，并且是一个信徒，我认为市场可能在错误定价一件事。

[05:47] And I'm I'm no macro expert, I do do a lot of pro national security investing.
我不是宏观经济专家，我确实做了很多关于国家安全的投资。

[05:53] And so I do have access to people who are experts and are excited to share their thoughts and opinions with me.
所以我确实能接触到那些专家，他们很乐意与我分享他们的想法和观点。

[06:02] That the straight of horm being closed is actually relatively awesome for America.
荷姆海峡的关闭对美国来说实际上是相对有利的。

[06:05] Awesome for America.
对美国来说是极好的。

[06:07] Why?
为什么？

[06:08] Because particularly for the goals of
因为特别是为了...的目标

[06:10] Because particularly for the goals of the current administration.
因为特别是为了当前政府的目标。

[06:12] The current administration. So electricity is a very important industrial or manufacturing input.
当前政府。所以电力是一个非常重要的工业或制造业投入。

[06:17] The key input into American electricity prices which feeds into AI is in G1 natural gas on Bloomberg that was down 20%.
进入美国电力价格的关键投入，而这又影响到人工智能，是在G1天然气上，根据彭博社的数据，天然气价格下跌了20%。

[06:26] And natural gas in Asia, Europe, everywhere else doubled or tripled.
而亚洲、欧洲以及其他所有地方的天然气价格都翻了一番或翻了两番。

[06:35] So our relative manufacturing competitiveness improved overnight and for better or worse that is what the Trump administration seems to care about.
所以我们相对的制造业竞争力一夜之间就提高了，无论好坏，这似乎是特朗普政府关心的。

[06:43] They are very focused on America's relative position.
他们非常关注美国的相对地位。

[06:48] And I think a lot of people had memories of the 1970s.
我想很多人还记得20世纪70年代。

[06:53] And what made the 70s so traumatic was it wasn't just that prices went up, it's that there were actual gas shortages.
而70年代之所以如此痛苦，不仅仅是因为价格上涨，更是因为出现了实际的天然气短缺。

[07:01] And then you go through, okay, well the US economy is dramatically less energy intensive than it was.
然后你再看，好吧，美国的经济比以前的能源密集度大大降低了。

[07:07] US econ the United States is now the world's largest producer of oil and gas and we've become
美国经济，美国现在是世界上最大的石油和天然气生产国，我们已经成为

[07:12] producer of oil and gas and we've become now the world's largest exporter of oil.
石油和天然气生产商，现在我们已成为世界上最大的石油出口国。

[07:15] now the world's largest exporter of oil and gas and then on top of that there's.
现在是世界上最大的石油和天然气出口国，在此之上还有。

[07:18] and gas and then on top of that there's this relative manufacturing advantage.
和天然气，在此之上还有这种相对的制造优势。

[07:21] this relative manufacturing advantage and so that made it I think easier to.
这种相对的制造优势，我认为这使得更容易。

[07:26] and so that made it I think easier to stay focused on AI fundamentals stay.
因此，我认为这使得更容易专注于人工智能的基本面，保持。

[07:30] stay focused on AI fundamentals stay focused on what were historically.
专注于人工智能的基本面，专注于历史上。

[07:33] focused on what were historically attractive valuations I think on a.
专注于历史上具有吸引力的估值，我认为在。

[07:35] attractive valuations I think on a relative basis.
有吸引力的估值，我认为在相对基础上。

[07:37] relative basis tech essentially got as cheap as it's.
相对基础上，科技基本上变得像它一样便宜。

[07:38] tech essentially got as cheap as it's been versus the rest of the market has.
科技基本上变得像它与市场上其他部分相比一样便宜。

[07:41] been versus the rest of the market has at any point over the last 10 years and.
在过去10年中的任何时候都比市场上其他部分便宜，并且。

[07:44] at any point over the last 10 years and just think about that in the context of.
在过去10年中的任何时候，并且只是从...的角度来考虑它。

[07:45] just think about that in the context of market efficiency. We have the most.
只是从市场效率的角度来考虑它。我们拥有最。

[07:47] market efficiency. We have the most extraordinary moment in the history of.
市场效率。我们拥有历史上最非凡的时刻。

[07:49] extraordinary moment in the history of capitalism that's wildly bullish for AI.
资本主义历史上非凡的时刻，这对人工智能来说是极度看好的。

[07:53] capitalism that's wildly bullish for AI and you get a chance to buy AI.
资本主义，这对人工智能来说是极度看好的，而你有机会购买人工智能。

[07:56] and you get a chance to buy AI at really attractive valuation. What do.
而你有机会以非常诱人的估值购买人工智能。你对此怎么看。

[07:59] at really attractive valuation. What do you make of the multiples that.
以非常诱人的估值。你如何看待那些。

[08:01] you make of the multiples that specifically Anthropic and OpenAI, which.
你如何看待特别是Anthropic和OpenAI的倍数，它们。

[08:03] specifically Anthropic and OpenAI, which in my mind are like the reference assets.
特别是Anthropic和OpenAI，在我看来，它们就像参考资产。

[08:05] in my mind are like the reference assets that are the most pure play takes on.
在我看来，它们就像参考资产，是对此趋势最纯粹的演绎。

[08:07] that are the most pure play takes on this trend really being not that crazy?
对此趋势最纯粹的演绎，真的不是那么疯狂？

[08:10] this trend really being not that crazy? Like if you just look at the sales.
这个趋势真的不是那么疯狂？就像如果你只看销售额。

[08:11] Like if you just look at the sales multiple and compare it to maybe what.
就像如果你只看销售倍数并将其与可能是什么进行比较。

[08:13] multiple and compare it to maybe what data bricks and snowflake and these.
多个，并将其与数据砖和雪花以及这些公司进行比较。

[08:15] data bricks and snowflake and these companies traded at at their peak like how do you make sense of it?
数据砖和雪花以及这些公司在其高峰期交易，你如何理解它？

[08:18] I do think OpenAI and Enthropic are pretty different animals from a capital efficiency perspective.
我认为OpenAI和Anthropic在资本效率方面是截然不同的。

[08:19] OpenAI and Enthropic are pretty different animals from a capital efficiency perspective.
OpenAI和Anthropic在资本效率方面是截然不同的。

[08:21] And Enthropic clearly is has a dramatically lower cost per token than OpenAI.
而且Anthropic每token的成本显然比OpenAI低得多。

[08:24] And Enthropic clearly is has a dramatically lower cost per token than OpenAI.
而且Anthropic每token的成本显然比OpenAI低得多。

[08:26] They just do.
它们就是如此。

[08:30] And you can just see that in the amount of money that they have burned to get to a roughly similar revenue scale.
你可以从他们为了达到大致相似的收入规模而烧掉的钱的数量中看出这一点。

[08:31] you can just see that in the amount of money that they have burned to get to a roughly similar revenue scale.
你可以从他们为了达到大致相似的收入规模而烧掉的钱的数量中看出这一点。

[08:35] I think have have they burned maybe 80% less than OpenAI.
我认为他们可能烧掉了比OpenAI少80%的钱。

[08:37] I think have have they burned maybe 80% less than OpenAI.
我认为他们可能烧掉了比OpenAI少80%的钱。

[08:39] than OpenAI.
比OpenAI。

[08:41] So as businesses, they clearly have very different structural ROIC's.
因此，作为企业，它们显然具有非常不同的结构性股本回报率。

[08:44] So as businesses, they clearly have very different structural ROIC's.
因此，作为企业，它们显然具有非常不同的结构性股本回报率。

[08:46] I think OpenAI is doing a lot.
我认为OpenAI正在做很多事情。

[08:48] I think Sarah Frier is one of the most exceptional CFOs.
我认为Sarah Frier是最杰出的CFO之一。

[08:49] CFOs.
CFO。

[08:51] I think they're doing a lot of things to try to improve this.
我认为他们正在做很多事情来尝试改进这一点。

[08:53] and they've secured a lot of compute.
并且他们已经获得了大量的计算资源。

[08:55] and they've secured a lot of compute.
并且他们已经获得了大量的计算资源。

[08:55] more more than they've secured a lot of compute.
比他们获得的计算资源多得多。

[08:57] That's another big difference.
这是另一个很大的区别。

[08:59] Um it turns out being aggressive really paid but yeah I just anthropic at 900 billion for 50 billion and you know ARR and you know I.
嗯，事实证明激进确实得到了回报，但是的，我只是认为Anthropic以9000亿美元的估值，500亿美元的年收入，你知道，年收入，你知道我。

[09:02] being aggressive really paid but yeah I just anthropic at 900 billion for 50 billion and you know ARR and you know I.
激进确实得到了回报，但是的，我只是认为Anthropic以9000亿美元的估值，500亿美元的年收入，你知道，年收入，你知道我。

[09:05] growing a thousand%.
增长了百分之九百。

[09:09] billion and you know ARR and you know I.
十亿美元，你知道年收入，你知道我。

[09:10] Yeah, growing at ridiculous rates.
是的，以惊人的速度增长。

[09:13] Maybe a true statement is that if Anthropic
也许一个真实的说法是，如果Anthropic

[09:15] A true statement is that if Anthropic had all the compute, they'd probably be doing well north of hundred billion dollars today, maybe 150.
一个真实的说法是，如果Anthropic拥有所有的计算能力，他们今天可能在一百亿以上，也许是一千五百亿。

[09:26] And I do, you know, they have clearly deprecated the intelligence of Claude.
而且我确实知道，他们已经明显降低了Claude的智能。

[09:31] There's an analysis Claude is even on Opus is generating 70% less tokens for the exact same question.
有一项分析显示，Claude在Opus上为完全相同的问题生成的token数量减少了70%。

[09:37] And you know, as we talked about last time, token quantity equals quality of answer and quality of thinking at some level.
而且你知道，正如我们上次讨论的，token的数量在某种程度上等于答案的质量和思考的质量。

[09:43] You know and there is an intelligence density per token that also matters you know I think I felt that as as a user so I think they would be doing materially more 100 150 maybe 200 billion so you might be buying it at more like five times unconstrained.
你知道，每个token的智能密度也很重要，你知道，我认为我作为用户感受到了这一点，所以我认为他们会做得更好，1000亿、1500亿，也许2000亿，所以你可能会以大约五倍的未受限制的价格购买它。

[09:59] I'm going to make up a new number urr unconstrained run rate revenue yes.
我将编造一个新数字，呃，未受限制的运行收入，是的。

[10:09] Why do you think they don't raise $und00 billion at a $3 trillion valuation or something like this.
你为什么认为他们不以3万亿美元的估值或类似的价格筹集1000亿美元呢？

[10:14] Like if you were
就像如果你是

[10:16] something like this.
类似这样。

[10:18] Like if you were the anthropic CFO, uh Krishna is the anthropic CFO, uh Krishna is awesome.
就像如果你是Anthropic的首席财务官，呃Krishna是Anthropic的首席财务官，呃Krishna很棒。

[10:19] We just had him on.
我们刚请了他来。

[10:21] Or if you're the open if you're Sarah, certainly if if the inbound I received following the Krishna episode is any indication, everyone I've ever met is trying to invest in in both these companies.
或者如果你是Open，如果你是Sarah，当然，如果我收到的关于Krishna那一集的入站信息有任何指示，我遇到的每个人都在试图投资这两家公司。

[10:24] following the Krishna episode is any indication, everyone I've ever met is trying to invest in in both these companies.
关于Krishna那一集的入站信息有任何指示，我遇到的每个人都在试图投资这两家公司。

[10:27] indication, everyone I've ever met is trying to invest in in both these companies.
指示，我遇到的每个人都在试图投资这两家公司。

[10:28] trying to invest in in both these companies.
试图投资这两家公司。

[10:29] companies.
公司。

[10:33] So I think it's wise it the future is uncertain.
所以我觉得明智的做法是，未来是不确定的。

[10:36] it the future is uncertain.
未来是不确定的。

[10:38] you are clearly in a very capital intensive game even if you are you know Enthropic
你显然是在一个非常资本密集的游戏中，即使你是，你知道，Enthropic

[10:41] intensive game even if you are you know Enthropic
密集的游戏，即使你是，你知道，Enthropic

[10:43] Enthropic um I'm sure is at very positive gross margins on inference today I think probably starts generating cash this year if they are not already generating cash which I think is probably the case
Enthropic，嗯，我敢肯定它今天在推理方面有非常高的毛利率，我认为它今年可能会开始产生现金，如果它们还没有产生现金的话，而我认为这很可能是事实。

[10:45] um I'm sure is at very positive gross margins on inference today I think probably starts generating cash this year if they are not already generating cash which I think is probably the case
嗯，我敢肯定它今天在推理方面有非常高的毛利率，我认为它今年可能会开始产生现金，如果它们还没有产生现金的话，而我认为这很可能是事实。

[10:48] margins on inference today I think probably starts generating cash this year if they are not already generating cash which I think is probably the case
推理方面的毛利率，我认为它今年可能会开始产生现金，如果它们还没有产生现金的话，而我认为这很可能是事实。

[10:49] probably starts generating cash this year if they are not already generating cash which I think is probably the case
可能会开始产生现金，如果它们还没有产生现金的话，而我认为这很可能是事实。

[10:51] year if they are not already generating cash which I think is probably the case
年，如果它们还没有产生现金的话，而我认为这很可能是事实。

[10:55] cash which I think is probably the case but still you probably want to be able to raise more capital access more compute the world is uncertain Ukraine
现金，而我认为这很可能是事实，但你仍然可能想能够筹集更多资本，获得更多计算能力，世界是不确定的，乌克兰

[10:57] but still you probably want to be able to raise more capital access more compute the world is uncertain Ukraine
但你仍然可能想能够筹集更多资本，获得更多计算能力，世界是不确定的，乌克兰

[10:59] to raise more capital access more compute the world is uncertain Ukraine is starting to really really win how is Russia going to respond ond, you know, I think there's still a lot of uncertainty in Iran.
筹集更多资本，获得更多计算能力，世界是不确定的，乌克兰开始真正地真正地获胜，俄罗斯将如何回应，你知道，我认为伊朗仍然存在很多不确定性。

[11:01] is starting to really really win how is Russia going to respond ond, you know, I think there's still a lot of uncertainty in Iran.
开始真正地真正地获胜，俄罗斯将如何回应，你知道，我认为伊朗仍然存在很多不确定性。

[11:04] Russia going to respond ond, you know, I think there's still a lot of uncertainty in Iran.
俄罗斯将如何回应，你知道，我认为伊朗仍然存在很多不确定性。

[11:06] think there's still a lot of uncertainty in Iran.
我认为伊朗仍然存在很多不确定性。

[11:07] in Iran. All this uncertainty, I think, probably amplifies geopolitical uncertainty over Taiwan.
在伊朗。我认为所有这些不确定性可能加剧台湾的地缘政治不确定性。

[11:10] All this uncertainty, I think, probably amplifies geopolitical uncertainty over Taiwan.
我认为所有这些不确定性可能加剧台湾的地缘政治不确定性。

[11:12] probably amplifies geopolitical uncertainty over Taiwan.
可能加剧台湾的地缘政治不确定性。

[11:14] uncertainty over Taiwan. So, it's an uncertain world. If if I think about
对台湾的不确定性。所以，这是一个不确定的世界。如果，如果我想到

[11:16] uncertain world. If if I think about Elon, Elon has always made investors money.
不确定的世界。如果，如果我想起埃隆，埃隆总是能让投资者赚钱。

[11:19] Elon, Elon has always made investors money. He treats it like a sacred covenant.
埃隆，埃隆总是能让投资者赚钱。他将其视为神圣的契约。

[11:21] money. He treats it like a sacred covenant. And as a result, because he's made people money for now 20 years, he has a superpower.
钱。他将其视为神圣的契约。因此，因为他过去20年一直为人们赚钱，所以他拥有超能力。

[11:24] covenant. And as a result, because he's made people money for now 20 years, he has a superpower. And that is he can essentially raise as much capital as he wants, whenever he wants.
契约。因此，因为他过去20年一直为人们赚钱，所以他拥有超能力。那就是他可以根据自己的意愿，在任何时候筹集到他想要的任何数额的资金。

[11:27] made people money for now 20 years, he has a superpower. And that is he can essentially raise as much capital as he wants, whenever he wants.
为人们赚钱已有20年，他拥有超能力。那就是他可以根据自己的意愿，在任何时候筹集到他想要的任何数额的资金。

[11:29] has a superpower. And that is he can essentially raise as much capital as he wants, whenever he wants. And I think it's wise that these companies are taking I don't know if that's how they think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
拥有超能力。那就是他可以根据自己的意愿，在任何时候筹集到他想要的任何数额的资金。我认为这些公司采取的做法是明智的，我不知道他们是否这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。

[11:32] essentially raise as much capital as he wants, whenever he wants. And I think it's wise that these companies are taking I don't know if that's how they think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
可以根据自己的意愿，在任何时候筹集到他想要的任何数额的资金。我认为这些公司采取的做法是明智的，我不知道他们是否这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。

[11:35] wants, whenever he wants. And I think it's wise that these companies are taking I don't know if that's how they think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
想要的时候，随时都可以。我认为这些公司采取的做法是明智的，我不知道他们是否这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。

[11:37] it's wise that these companies are taking I don't know if that's how they think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
这些公司采取的做法是明智的，我不知道他们是否这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。

[11:38] taking I don't know if that's how they think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
采取，我不知道他们是否这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。

[11:41] think about it, but I do think being focused on making investors money is wise and creates benefits that don't just last for like a year or two.
这样想，但我确实认为专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。

[11:45] focused on making investors money is wise and creates benefits that don't just last for like a year or two. They can last for the next 20 to 30 years.
专注于为投资者赚钱是明智的，并且会带来不仅仅是一两年的好处。它们可以持续未来20到30年。

[11:49] just last for like a year or two. They can last for the next 20 to 30 years.
仅仅持续一两年。它们可以持续未来20到30年。

[11:52] just last for like a year or two. They can last for the next 20 to 30 years.
仅仅持续一两年。它们可以持续未来20到30年。

[11:54] can last for the next 20 to 30 years. >> And the way Elon did this was sort of systematically underpricing SpaceX or whatever else.
可以持续未来20到30年。>> 埃隆做到这一点的方式是系统性地低估SpaceX或其他任何东西的价格。

[11:57] >> And the way Elon did this was sort of systematically underpricing SpaceX or whatever else. Like what is the actual method?
>> 埃隆做到这一点的方式是系统性地低估SpaceX或其他任何东西的价格。实际方法是什么？

[11:59] systematically underpricing SpaceX or whatever else. Like what is the actual method? Just never being greedy on valuation, never pushing valuation.
系统性地低估SpaceX或其他任何东西的价格。实际方法是什么？就是从不在估值上贪婪，从不推高估值。

[12:00] whatever else. Like what is the actual method? Just never being greedy on valuation, never pushing valuation.
任何其他东西。实际方法是什么？就是从不在估值上贪婪，从不推高估值。

[12:02] method? Just never being greedy on valuation, never pushing valuation. >> Just that simple.
方法？就是从不在估值上贪婪，从不推高估值。>> 就这么简单。

[12:04] Just never being greedy on valuation, never pushing valuation. >> Just that simple.
就是从不在估值上贪婪，从不推高估值。>> 就这么简单。

[12:06] never pushing valuation. >> Just that simple. >> You know, my friend Antonio pointed out SpaceX compounded, you know, low 30% per year for whatever that was a decade.
从不推高估值。>> 就这么简单。>> 你知道，我的朋友安东尼奥指出，SpaceX的年复合增长率，你知道，大约是30%，持续了十年。

[12:08] >> Just that simple. >> You know, my friend Antonio pointed out SpaceX compounded, you know, low 30% per year for whatever that was a decade.
>> 就这么简单。>> 你知道，我的朋友安东尼奥指出，SpaceX的年复合增长率，你知道，大约是30%，持续了十年。

[12:09] >> You know, my friend Antonio pointed out SpaceX compounded, you know, low 30% per year for whatever that was a decade. And and that was just because Elon was, I
>> 你知道，我的朋友安东尼奥指出，SpaceX的年复合增长率，你知道，大约是30%，持续了十年。而这仅仅是因为埃隆，我

[12:12] SpaceX compounded, you know, low 30% per year for whatever that was a decade. And and that was just because Elon was, I
SpaceX的年复合增长率，你知道，大约是30%，持续了十年。而这仅仅是因为埃隆，我

[12:16] year for whatever that was a decade. And and that was just because Elon was, I
年，持续了十年。而这仅仅是因为埃隆，我

[12:18] And that was just because Elon was, I think, focused on preserving the superpower and having trying to strike a fair balance between investors and employees.
这仅仅是因为埃隆，我认为，专注于维护超级大国，并试图在投资者和员工之间取得公平的平衡。

[12:26] But I I think it's wise.
但我我认为这是明智的。

[12:28] But could Anthropic raise money at probably at least a 100% premium to this rumored latest mark?
但Anthropic能否以至少100%的溢价筹集资金，达到这个传闻中的最新估值？

[12:38] Of course, most software companies try to maximize your time on their app to juice engagement.
当然，大多数软件公司都试图最大化你在他们应用程序上的时间，以提高参与度。

[12:41] RAMP does the exact opposite.
RAMP则恰恰相反。

[12:43] RAMP understands that no one wants to spend hours chasing receipts, reviewing expense reports, and checking for policy violations.
RAMP明白没有人想花几个小时去追逐收据、审查费用报告和检查政策违规行为。

[12:51] So, they built their tools to give that time back using AI to automate 85% of expense reviews with 99% accuracy.
所以，他们构建了他们的工具，利用人工智能将85%的费用审查自动化，准确率达到99%，从而将这些时间还给用户。

[12:59] And since Ramp saves companies 5%, it's no wonder that Shopify runs on RAM, Stripe runs on RAM, and my business does, too.
而且由于Ramp为公司节省了5%的成本，难怪Shopify、Stripe以及我的公司都在使用RAM。

[13:05] To see what happens when you eliminate the busy work, check out ramp.com/invest.
想看看当你消除繁琐的工作时会发生什么，请访问ramp.com/invest。

[13:10] Felix by Rogo is a personal finance agent that turns a single prompt into finished client ready work using your firm's own templates, context, and standards.
Rogo的Felix是一款个人理财助手，它能将一个简单的提示，利用你公司的模板、上下文和标准，转化为客户准备就绪的最终工作成果。

[13:17] Send Felix an email like,
给Felix发一封电子邮件，比如，

[13:19] Standards. Send Felix an email like, "Take these comments and turn them for me or update my tracker with the context of these emails."
标准。给菲利克斯发一封邮件，说：“把这些评论转给我，或者用这些邮件的上下文更新我的跟踪器。”

[13:25] Or, "Run the ability to pay math on this buyer and Felix sends back finished PowerPoint decks, Excel models, and sourced research."
或者，“对这个买家运行支付能力数学计算，然后菲利克斯会发回完成的PowerPoint演示文稿、Excel模型和搜集到的研究资料。”

[13:31] Felix works the way your team already does, delivering work quickly and accurately around the clock.
菲利克斯的工作方式就像你们团队已经做的那样，全天候快速准确地交付工作。

[13:36] Learn more at rogo.ai/felix.
了解更多信息，请访问rogo.ai/felix。

[13:40] OpenAI, Cursor, Enthropic, Perplexity, and Versell all have something in common.
OpenAI、Cursor、Enthropic、Perplexity和Versell都有一个共同点。

[13:44] They all use WorkOS.
它们都使用WorkOS。

[13:47] And here's why.
原因如下。

[13:49] To achieve enterprise adoption at scale, you have to deliver on core capabilities like SSO, skim, arbback, and audit logs.
为了实现大规模的企业采用，您必须提供核心功能，如SSO、skim、arbback和审计日志。

[13:55] That's where work OS comes in.
这就是Work OS的作用。

[13:57] Instead of spending months building these mission critical capabilities yourself, you can just use work OS APIs to gain all of them on day zero.
与其花费数月时间自己构建这些关键任务功能，不如直接使用Work OS API，在第一天就获得所有这些功能。

[14:02] That's why so many of the top AI teams you hear about already run on work OS.
这就是为什么您听说过的许多顶级AI团队都已经运行在Work OS上。

[14:07] Work OS is the fastest way to become enterprise ready and stay focused on what matters most, your product.
Work OS是为企业做好准备并专注于最重要事项——您的产品的最快途径。

[14:13] Visit works.com to get started.
访问works.com开始使用。

[14:15] Let's get to the Watson wafers part of the discussion.
让我们来谈谈讨论中关于沃森晶圆的部分。

[14:19] Always my favorite thing to talk about with you.
这总是我最喜欢和你谈论的事情。

[14:19] Uh the importance
呃，重要性

[14:21] Talk about with you.
和你谈谈。

[14:21] Uh the importance of this infrastructure buildout.
呃，这次基础设施建设的重要性。

[14:24] I feel of this infrastructure buildout.
我觉得这次基础设施建设。

[14:24] I feel like every time I feel like it's getting overheated and then the next time I talk to you, it seems like we should have done way more than we did.
我觉得好像每次都感觉它过热了，然后下次和你谈话时，似乎我们应该做得比实际做得多得多。

[14:27] overheated and then the next time I talk to you, it seems like we should have done way more than we did.
过热了，然后下次和你谈话时，似乎我们应该做得比实际做得多得多。

[14:29] And you've studied S-curves and the steepness of those S-curves a lot.
你研究了S曲线和那些S曲线的陡峭程度很多。

[14:33] Uh and you know a lot about history.
呃，而且你对历史也很了解。

[14:35] Talk us through how you're thinking about Watson wafers today as the key to inputs into this whole thing.
给我们讲讲你今天如何看待沃森晶圆是这一切的关键输入。

[14:42] Yeah, I would say I think capitalism is going to solve the Watts shortage absent big regulatory political blowback.
是的，我想说我认为资本主义将解决瓦特短缺问题，除非有大的监管政治抵制。

[14:49] which I think is a real possibility.
我认为这是很有可能发生的。

[14:53] the head of kind of data center infrain investing at one of the big PE firms.
一家大型私募股权公司数据中心基础设施投资的负责人。

[14:58] You know, I think Blackstone, Apollo, KKR said it used to be energy and chips were our biggest gating factors.
你知道，我认为黑石、阿波罗、KKR说，过去能源和芯片是我们最大的限制因素。

[15:04] Now it's zoning and approval much more important.
现在，分区和审批变得更加重要了。

[15:10] And I think a lot of companies are waiting till after the midterms to take action in terms of maybe workforce reductions.
而且我认为很多公司在采取行动之前会等到中期选举之后，比如可能裁员。

[15:17] Nobody wants to be, you know, piñata during the
没有人想成为，你知道，在...期间的替罪羊。

[15:22] To be, you know, piñata during the midterms.
要成为，你知道的，中期选举中的皮纳塔。

[15:24] Midterms, but you know, you've seen a lot of companies that make turbines significant.
中期选举，但你知道，你已经看到很多制造涡轮机的公司很重要。

[15:26] But you know, you've seen a lot of companies that make turbines significant announce of plans to significantly increase capacity.
但你知道，你已经看到很多制造涡轮机的公司宣布了大幅提高产能的计划。

[15:30] Announce of plans to significantly increase capacity.
宣布了大幅提高产能的计划。

[15:32] There's like two of these machines that can cast these big blades.
有大约两台这样的机器可以铸造这些大型叶片。

[15:34] These machines that can cast these big blades.
这些机器可以铸造这些大型叶片。

[15:37] We haven't made one in 80 years in the West.
我们在西方已经80年没有制造过一个了。

[15:39] We don't know how to make them anymore, etc., etc., etc.
我们不知道如何再制造它们了，等等，等等，等等。

[15:42] All of that is true.
所有这些都是真的。

[15:44] And I and and by no means am I minimizing, you know, the industrial engineering, you know, magic and artistry that goes into those, but capitalism is very good at solving problems like these over time.
我绝不是在轻视，你知道的，工业工程，你知道的，那些东西所包含的魔法和艺术性，但资本主义非常擅长随着时间的推移解决这类问题。

[15:49] And artistry that goes into those, but capitalism is very good at solving problems like these over time.
那些东西所包含的艺术性，但资本主义非常擅长随着时间的推移解决这类问题。

[15:52] There's other sources of energy besides these turbines with a longer time frame.
除了这些涡轮机之外，还有其他能源，而且时间跨度更长。

[15:54] So I think the watts shortage will probably begin to alleviate 27 28 and then I think orbital compute will really solve that.
所以我认为瓦特短缺可能会在27、28年开始缓解，然后我认为轨道计算将真正解决这个问题。

[15:59] And I do I do want to like reframe orbital compute because I think when people hear data centers in space they which we discussed on our last episode they picture a pentagon sized building in space.
我确实想重新定义轨道计算，因为我认为当人们听到太空中的数据中心时，就像我们在上期节目中讨论的那样，他们会想象一个五角大楼大小的太空建筑。

[16:09] They're like well we can't do
他们会说，嗯，我们不能做

[16:23] in space.
在太空中。

[16:23] They're like well we can't do that.
他们就像，嗯，我们不能那样做。

[16:23] That's not what it is.
那不是它。

[16:28] A blackwell that.
一个布莱克韦尔那个。

[16:28] That's not what it is.
那不是它。

[16:28] A blackwell rack weighs you 3,000 lbs.
一个布莱克韦尔机架重 3000 磅。

[16:31] It's 8 ft rack weighs you 3,000 lbs.
它是 8 英尺的机架重 3000 磅。

[16:31] It's 8 ft high.
它高 8 英尺。

[16:31] It's 4 feet deep.
它深 4 英尺。

[16:31] 3 feet wide.
宽 3 英尺。

[16:35] high.
高。

[16:35] It's 4 feet deep.
它深 4 英尺。

[16:35] 3 feet wide.
宽 3 英尺。

[16:35] It's racks in space.
它是太空中的机架。

[16:38] And SpaceX has it's racks in space.
SpaceX 有它在太空中的机架。

[16:38] And SpaceX has showed you an illustration and it's a
SpaceX 向你展示了一张插图，它是一个

[16:41] showed you an illustration and it's a rack.
展示了一张插图，它是一个机架。

[16:41] That's the satellite.
那就是卫星。

[16:41] Uh but it's probably about the size of a blackwell
呃，但它可能和布莱克韦尔差不多大

[16:44] rack.
机架。

[16:44] That's the satellite.
那就是卫星。

[16:44] Uh but it's probably about the size of a blackwell
呃，但它可能和布莱克韦尔差不多大

[16:46] probably about the size of a blackwell rack.
可能差不多是一个布莱克韦尔机架的大小。

[16:46] It has these solar wings that are
它有这些太阳能翼，它们是

[16:48] rack.
机架。

[16:48] It has these solar wings that are probably 500 ft long on each side.
它有这些太阳能翼，每边可能长 500 英尺。

[16:51] probably 500 ft long on each side.
可能每边长 500 英尺。

[16:51] You keep it in a sunsynchronous orbit.
你把它放在太阳同步轨道上。

[16:54] keep it in a sunsynchronous orbit.
放在太阳同步轨道上。

[16:54] So those solar panels are always in the
所以那些太阳能电池板总是在

[16:56] those solar panels are always in the sun.
那些太阳能电池板总是在阳光下。

[16:56] And then because it's in an exactly
然后因为它在一个精确的

[16:58] sun.
阳光下。

[16:58] And then because it's in an exactly sunsynchronous orbit, the radiator which
然后因为它在一个精确的太阳同步轨道上，散热器它

[17:01] sunsynchronous orbit, the radiator which extends behind it for hundreds of feet.
太阳同步轨道上，散热器它延伸在它后面数百英尺。

[17:05] extends behind it for hundreds of feet.
延伸在它后面数百英尺。

[17:05] >> This is a common criticism.
>>这是一个常见的批评。

[17:05] Yeah.
是的。

[17:06] how
如何

[17:06] you going to go.
你要去。

[17:07] >> I've spent a lot of time at Starbase
>>我花了很多时间在星港

[17:11] over the years and I've talked to a lot
这些年来，我和很多

[17:13] of SpaceX engineers and I do think it is
SpaceX 的工程师谈过，我认为它是

[17:16] the most talented group of engineers on
最才华横溢的工程师团队

[17:19] planet Earth and they're very confident
地球上，他们非常有信心

[17:21] they have solved this and they're not
他们已经解决了这个问题，而且他们没有

[17:23] they have solved this and they're not always confident like I think probably.
他们已经解决了这个问题，而且他们并不总是那么自信，就像我认为可能的那样。

[17:27] always confident like I think probably you know there's some engineering that needs to happen to turn the starship into a Mars colonial transporter.
总是自信，就像我认为可能的那样，你知道，需要做一些工程才能将星舰变成火星殖民运输船。

[17:30] needs to happen to turn the starship into a Mars colonial transporter.
需要做一些工程才能将星舰变成火星殖民运输船。

[17:32] Will they do that? Absolutely.
他们会这样做吗？绝对会。

[17:34] What are they more focused on? I would say probably, you know, the repair and maintenance.
他们更关注什么？我想可能是，你知道，维修和保养。

[17:37] more focused on? I would say probably, you know, the repair and maintenance.
更关注什么？我想可能是，你知道，维修和保养。

[17:39] These are the two big, you know, the two big responses, the radiator and and how do you repair the whatever issue goes wrong in the rack.
这是两个大的，你知道，两个大的响应，散热器以及如何修复机架中出现的任何问题。

[17:40] big responses, the radiator and and how do you repair the whatever issue goes wrong in the rack.
大的响应，散热器以及如何修复机架中出现的任何问题。

[17:42] do you repair the whatever issue goes wrong in the rack.
如何修复机架中出现的任何问题。

[17:44] And the answer is like until you have probably an, you know, floating optimuses.
答案是，直到你可能拥有，你知道，浮动的优化器。

[17:45] wrong in the rack. And the answer is like until you have probably an, you know, floating optimuses.
在机架中出错了。答案是，直到你可能拥有，你知道，浮动的优化器。

[17:47] You don't. Now, I do think Starship is going to change the space economy in ways we cannot imagine.
你不会。现在，我认为星舰将以我们无法想象的方式改变太空经济。

[17:50] optimuses. You don't. Now, I do think Starship is going to change the space economy in ways we cannot imagine.
优化器。你不会。现在，我认为星舰将以我们无法想象的方式改变太空经济。

[17:52] Starship is going to change the space economy in ways we cannot imagine. Particularly if regulation becomes a constraint to data centers, none of it's going to matter.
星舰将以我们无法想象的方式改变太空经济。特别是如果监管成为数据中心的限制，这一切都将无关紧要。

[17:54] economy in ways we cannot imagine. Particularly if regulation becomes a constraint to data centers, none of it's going to matter.
经济，以我们无法想象的方式。特别是如果监管成为数据中心的限制，这一切都将无关紧要。

[17:56] Particularly if regulation becomes a constraint to data centers, none of it's going to matter. you're going to sell as much orbital compute as you can make.
特别是如果监管成为数据中心的限制，这一切都将无关紧要。你将销售你能制造的尽可能多的轨道计算能力。

[17:57] constraint to data centers, none of it's going to matter. you're going to sell as much orbital compute as you can make.
对数据中心的限制，这一切都将无关紧要。你将销售你能制造的尽可能多的轨道计算能力。

[17:59] going to matter. you're going to sell as much orbital compute as you can make. And then obviously you link these racks using lasers traveling through vacuum which are already on every Starlink.
将无关紧要。你将销售你能制造的尽可能多的轨道计算能力。然后显然，你将使用激光在真空中连接这些机架，这些激光已经存在于每个星链上。

[18:01] much orbital compute as you can make. And then obviously you link these racks using lasers traveling through vacuum which are already on every Starlink.
尽可能多的轨道计算能力。然后显然，你将使用激光在真空中连接这些机架，这些激光已经存在于每个星链上。

[18:04] And then obviously you link these racks using lasers traveling through vacuum which are already on every Starlink. And it's just it's just mindblowing to me that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit.
然后显然，你将使用激光在真空中连接这些机架，这些激光已经存在于每个星链上。对我来说，这简直令人难以置信，SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。

[18:07] using lasers traveling through vacuum which are already on every Starlink. And it's just it's just mindblowing to me that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit.
使用激光在真空中旅行，这些激光已经存在于每个星链上。对我来说，这简直令人难以置信，SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。

[18:10] which are already on every Starlink. And it's just it's just mindblowing to me that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit.
已经存在于每个星链上。对我来说，这简直令人难以置信，SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。

[18:11] And it's just it's just mindblowing to me that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit.
对我来说，这简直令人难以置信，SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。

[18:15] it's just it's just mindblowing to me that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit.
简直令人难以置信，SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。

[18:17] that SpaceX operates the world's largest satellite fleet which is like 98 or 99% of all satellites in orbit. Every
SpaceX运营着世界上最大的卫星舰队，占所有在轨卫星的98%或99%。每一个

[18:24] of all satellites in orbit.
在轨卫星的总数。

[18:24] Every Starlink they're cooling it today.
每一个星链，他们今天都在降温。

[18:28] Starlink they're cooling it today.
星链，他们今天都在降温。

[18:28] And you know, I think Starlink V3 is going to operate at 20 kW.
而且你知道，我认为星链V3将运行在20千瓦。

[18:33] A Blackwell rack is only 100 kows.
一个 Blackwell 机架只有100千瓦。

[18:37] And people talk a lot about density.
人们谈论了很多关于密度的问题。

[18:39] Well, if you're connecting the racks with lasers through vacuum, you know, you can make the rack bigger physically.
嗯，如果你通过真空用激光连接机架，你知道，你可以让机架在物理上更大。

[18:45] You're focused on weight, not size.
你关注的是重量，而不是大小。

[18:49] In a data center on Earth where you're trying to connect racks, ideally using copper, minimize lengths, etc., etc.
在地球上的数据中心，你试图连接机架，理想情况下使用铜线，尽量缩短长度等等。

[18:53] Cabling is a big cost.
布线是一笔巨大的开销。

[18:56] um you do want that rack to be small because you know copper when you can, optics when you must.
嗯，你确实希望那个机架小一些，因为你知道，在可能的情况下使用铜线，在必须的情况下使用光纤。

[19:01] But in space, you know, there's all sorts of things that SpaceX can do that I think maybe some of these naysayers are not contemplating, but it's just they operate more satellites than you want.
但在太空，你知道，SpaceX 可以做各种各样的事情，我认为也许一些反对者没有考虑到，但他们只是运营的卫星比你想要的要多。

[19:12] They have a 20 kow satellite today, so maybe you just scale that up to 60 kilowatts to start.
他们今天有一个20千瓦的卫星，所以也许你一开始就将其扩展到60千瓦。

[19:18] They seem very confident they're going to go right to 100 to 120.
他们似乎非常有信心，将直接达到100到120。

[19:23] And they also the same company now also operates the largest
而且他们也是同一家公司，现在也运营着最大的

[19:26] company now also operates the largest data center on earth.
公司现在还运营着地球上最大的数据中心。

[19:28] They have the data center on earth.
他们拥有地球上的数据中心。

[19:31] They have the world's best hardware engineers and all sorts of people almost all of whom are not smart enough or practical enough to work at SpaceX are these armchair skeptics.
他们拥有世界上最好的硬件工程师和各种各样的人，几乎所有这些人都不够聪明或不够实用，无法在SpaceX工作，他们是这些纸上谈兵的怀疑论者。

[19:37] not smart enough or practical enough to work at SpaceX are these armchair skeptics.
不够聪明或不够实用，无法在SpaceX工作，他们是这些纸上谈兵的怀疑论者。

[19:40] work at SpaceX are these armchair skeptics.
在SpaceX工作，他们是这些纸上谈兵的怀疑论者。

[19:42] skeptics.
怀疑论者。

[19:43] You know, I don't want to quote Larry Ellison, but somebody was, you know, being skeptical and Larry and Larry was just like, "Listen, he's out there landing rockets.
你知道，我不想引用拉里·埃里森的话，但有人，你知道，持怀疑态度，而拉里，拉里就像说，“听着，他正在那里着陆火箭。

[19:47] being skeptical and Larry and Larry was just like, "Listen, he's out there landing rockets.
持怀疑态度，而拉里，拉里就像说，“听着，他正在那里着陆火箭。

[19:49] just like, "Listen, he's out there landing rockets.
就像说，“听着，他正在那里着陆火箭。

[19:51] landing rockets.
着陆火箭。

[19:53] I don't see anybody else landing rockets.
我没看到其他人着陆火箭。

[19:56] And the reality is is that 10 years later, no other company is consistently capable of landing and fully reusing an orbital rocket.
而现实是，10年后，没有其他公司能够持续地着陆和完全重复使用轨道火箭。

[19:59] is consistently capable of landing and fully reusing an orbital rocket.
能够持续地着陆和完全重复使用轨道火箭。

[20:03] And fully reusing an orbital rocket.
并完全重复使用轨道火箭。

[20:05] And none of this works makes sense without reusability.
而这一切的运作如果没有可重用性就没有意义。

[20:06] That means you have to land it.
这意味着你必须着陆它。

[20:08] I would like to redefine orbital compute has racks in space, not giant floating pentagoniz data centers in space, which is just, you know, that's silly.
我想重新定义轨道计算，它在太空中拥有机架，而不是在太空中拥有巨大的漂浮的五角形数据中心，这简直是，你知道，这很愚蠢。

[20:14] floating pentagoniz data centers in space, which is just, you know, that's silly.
在太空中漂浮的五角形数据中心，这简直是，你知道，这很愚蠢。

[20:17] data centers in space, which is just, you know, that's silly.
在太空中数据中心，这简直是，你知道，这很愚蠢。

[20:19] But you can, you know, what makes a data center is you're connecting these racks with lasers.
但你可以，你知道，使数据中心成为数据中心的是你用激光连接这些机架。

[20:23] connecting these racks with lasers.
用激光连接这些机架。

[20:24] So it'll be racks in space that are connected with lasers into a virtual data center.
所以它将是在太空中用激光连接成虚拟数据中心的机架。

[20:26] connected with lasers into a virtual data center.
用激光连接成虚拟数据中心。

[20:26] And and if you think about
而如果你想到

[20:28] data center. And and if you think about that state of the world, let's say that

[20:31] that state of the world, let's say that all happens and we're really good at

[20:32] all happens and we're really good at getting these things up economically,

[20:34] getting these things up economically, running matrix multiplication all over

[20:36] running matrix multiplication all over space. What does that mean for

[20:37] space. What does that mean for terrestrial data centers? Someone once

[20:40] terrestrial data centers? Someone once said, um, you know, America was going to

[20:44] said, um, you know, America was going to suck as hard as it can on every energy

[20:47] suck as hard as it can on every energy source it can get. And I just think the

[20:49] source it can get. And I just think the same is true of compute.

[20:51] same is true of compute. >> It's why I'm probably less worried about

[20:53] >> It's why I'm probably less worried about like an edge AI barecase than I was.

[20:57] like an edge AI barecase than I was. >> We're going to consume as much compute

[21:02] >> We're going to consume as much compute as we can. And inference I think is very

[21:06] as we can. And inference I think is very sensible for orbital compute. Training

[21:09] sensible for orbital compute. Training will be done on Earth for a long time.

[21:12] will be done on Earth for a long time. So I don't think that this is super

[21:14] So I don't think that this is super bearish for terrestrial data centers. I

[21:16] bearish for terrestrial data centers. I think those are going to be valuable for

[21:18] think those are going to be valuable for my lifetime.

[21:21] my lifetime. But I do think if you are in this

[21:23] But I do think if you are in this ecosystem of power production and

[21:26] ecosystem of power production and cooling and you are massively ramping

[21:31] cooling and you are massively ramping capacity and you know a lot of these

[21:33] capacity and you know a lot of these capacity ramps are going to be hitting

[21:36] capacity ramps are going to be hitting just as I think you know all of the

[21:38] just as I think you know all of the silly skeptics start to understand that

[21:40] silly skeptics start to understand that orbital compute is very real like I

[21:42] orbital compute is very real like I think it's worth thinking long and hard

[21:44] think it's worth thinking long and hard about that if you're one of those

[21:46] about that if you're one of those companies and then all sorts of cool

[21:48] companies and then all sorts of cool stuff is happening in the interim you

[21:49] stuff is happening in the interim you know we're getting really good at

[21:51] know we're getting really good at repurposing jet engines. You know,

[21:53] repurposing jet engines. You know, there's that boom aerospace that is

[21:54] there's that boom aerospace that is doing this.

[21:55] doing this. >> So, there's a lot like capitalism is

[21:57] >> So, there's a lot like capitalism is hard at work

[21:59] hard at work >> on on watts. On wafers, though, it's

[22:03] >> on on watts. On wafers, though, it's just this group of, you know, flinty

[22:08] just this group of, you know, flinty older humans in Taiwan who are the most

[22:11] older humans in Taiwan who are the most important humans in Taiwan. They are the

[22:14] important humans in Taiwan. They are the overwhelming fraction of the country's

[22:15] overwhelming fraction of the country's GDP, water usage, electricity usage.

[22:18] GDP, water usage, electricity usage. They talk about the Silicon Shield. They

[22:21] They talk about the Silicon Shield. They all view themselves as inheritors of,

[22:25] all view themselves as inheritors of, you know, Morris Chain's sacred legacy.

[22:27] you know, Morris Chain's sacred legacy. I vividly remember like visiting Science

[22:29] I vividly remember like visiting Science Park more than 20 years ago and, you

[22:34] Park more than 20 years ago and, you know, talking to them. Do you think you

[22:35] know, talking to them. Do you think you could catch Intel? And they said, "This

[22:38] could catch Intel? And they said, "This is such a beautiful dream, but it's a

[22:40] is such a beautiful dream, but it's a dream for our grandchildren."

[22:43] dream for our grandchildren." >> And they did it. partly because of

[22:45] >> And they did it. partly because of Intel's self-inflicted wounds, but just

[22:48] Intel's self-inflicted wounds, but just they don't they think very differently.

[22:51] they don't they think very differently. You know, one reason, you know, Jensen

[22:53] You know, one reason, you know, Jensen flies over there so much is he wants

[22:55] flies over there so much is he wants them to expand capacity. I do think it's

[22:57] them to expand capacity. I do think it's wild that Jensen has never had a

[22:59] wild that Jensen has never had a contract with Taiwan Semi. They do

[23:01] contract with Taiwan Semi. They do business on what seems fair in

[23:03] business on what seems fair in handshakes. Just fascinating. No

[23:05] handshakes. Just fascinating. No contract. It's going to be fair over

[23:07] contract. It's going to be fair over time. We're partners. We're going to be

[23:09] time. We're partners. We're going to be fair to each other. And the truth is,

[23:11] fair to each other. And the truth is, you know, based on every every prior

[23:15] you know, based on every every prior market precedent for a foundational new

[23:17] market precedent for a foundational new technology like AI, you've always had a

[23:19] technology like AI, you've always had a bubble. You know, Carleta Perez wrote

[23:21] bubble. You know, Carleta Perez wrote this great book about this. And

[23:23] this great book about this. And basically, markets are efficient. They

[23:25] basically, markets are efficient. They correctly understand that this is a

[23:27] correctly understand that this is a foundational technology.

[23:29] foundational technology. There's what Mobison calls a breakdown

[23:31] There's what Mobison calls a breakdown in diversity.

[23:33] in diversity. Everyone becomes bullish on this new

[23:35] Everyone becomes bullish on this new technology. And I am beginning to worry

[23:38] technology. And I am beginning to worry a little bit about a diversity

[23:39] a little bit about a diversity breakdown. And then you get a bubble.

[23:43] breakdown. And then you get a bubble. That bubble funds the buildout of this

[23:45] That bubble funds the buildout of this new technology, but supply gets ahead of

[23:48] new technology, but supply gets ahead of demand. And you get a crash and it's a

[23:51] demand. And you get a crash and it's a particularly severe crash if it's a

[23:53] particularly severe crash if it's a debtfueled buildout like the year 2000.

[23:56] debtfueled buildout like the year 2000. And one thing really happy about really

[23:59] And one thing really happy about really good about the current buildout is it's

[24:00] good about the current buildout is it's still overwhelmingly funded out of

[24:02] still overwhelmingly funded out of operating cash flows which is a a really

[24:05] operating cash flows which is a a really important fundamental difference versus

[24:06] important fundamental difference versus the year 2000 has is valuation has is

[24:09] the year 2000 has is valuation has is the fact that every GPU is running at

[24:11] the fact that every GPU is running at 100% utilization when 99% of fiber was

[24:14] 100% utilization when 99% of fiber was unutilized. So there's all these

[24:15] unutilized. So there's all these fundamental differences, but we do have

[24:17] fundamental differences, but we do have to history doesn't repeat, but it rhymes

[24:19] to history doesn't repeat, but it rhymes and and as investor, we have to be very

[24:21] and and as investor, we have to be very cognizant of it

[24:23] cognizant of it and recognize that based on the last two

[24:26] and recognize that based on the last two or 30 hundred years, you know, forget

[24:28] or 30 hundred years, you know, forget the internet bubble. We had a railroad

[24:29] the internet bubble. We had a railroad bubble, a canal bubble, we should expect

[24:31] bubble, a canal bubble, we should expect a bubble. And that's terrifying. Like

[24:35] a bubble. And that's terrifying. Like nobody wants a bubble. A bubble is

[24:37] nobody wants a bubble. A bubble is terrible. reason it's terrible is if

[24:39] terrible. reason it's terrible is if you're valuation sensitive, you like

[24:41] you're valuation sensitive, you like massively underperform. You get fired by

[24:44] massively underperform. You get fired by probably all your clients. George

[24:46] probably all your clients. George Vanderhiden, who um is is is no longer

[24:49] Vanderhiden, who um is is is no longer with us, great port uh fidelity

[24:51] with us, great port uh fidelity portfolio manager, he fought the bubble

[24:54] portfolio manager, he fought the bubble in 99 and he retired in two in early

[24:58] in 99 and he retired in two in early 2000 because I think he just couldn't

[24:59] 2000 because I think he just couldn't couldn't take it.

[25:00] couldn't take it. >> He knew it was wrong and you know his

[25:03] >> He knew it was wrong and you know his his clients were deeply skeptical.

[25:05] his clients were deeply skeptical. George, you're out of step. you know, he

[25:07] George, you're out of step. you know, he had he had white hair. He's truly great

[25:09] had he had white hair. He's truly great man. I I only overlapped with him

[25:11] man. I I only overlapped with him briefly, but he was a very important

[25:13] briefly, but he was a very important mentor and friend to my good friend and

[25:16] mentor and friend to my good friend and mentor Jennifer Urick. So, I have a lot

[25:18] mentor Jennifer Urick. So, I have a lot of Vanderhiden DNA through her. He was

[25:21] of Vanderhiden DNA through her. He was the same person who said being early is

[25:23] the same person who said being early is the same thing as being wrong. George

[25:24] the same thing as being wrong. George retires because he can't take the

[25:27] retires because he can't take the underperformance and he can't take

[25:29] underperformance and he can't take clients saying what's wrong with you?

[25:31] clients saying what's wrong with you? You don't get it. and he has like 40% of

[25:34] You don't get it. and he has like 40% of his funded tobacco, 40% did homebuilders

[25:38] his funded tobacco, 40% did homebuilders and literally he underperfor he probably

[25:40] and literally he underperfor he probably outperformed the NASDAQ

[25:43] outperformed the NASDAQ by like 20 or 30x over the next three

[25:47] by like 20 or 30x over the next three years. Okay. And I have been optimistic

[25:50] years. Okay. And I have been optimistic that this fundamental shortage of wafers

[25:53] that this fundamental shortage of wafers which really today is controlled by

[25:56] which really today is controlled by Taiwan Semi will prevent one. If Taiwan

[25:58] Taiwan Semi will prevent one. If Taiwan Semi did what Jensen wanted, I think

[26:00] Semi did what Jensen wanted, I think Nvidia could sell two trillion dollars

[26:02] Nvidia could sell two trillion dollars of GPUs in 26 in 26 or 27, maybe two.5

[26:07] of GPUs in 26 in 26 or 27, maybe two.5 trillion, maybe three trillion, but

[26:09] trillion, maybe three trillion, but there is a limit where consumers would

[26:11] there is a limit where consumers would consume so much that you probably would

[26:14] consume so much that you probably would be in an overbuild. And so Taiwan Smi,

[26:17] be in an overbuild. And so Taiwan Smi, if we don't get a bubble, like we need

[26:18] if we don't get a bubble, like we need to throw a party for them because they

[26:20] to throw a party for them because they will have single-handedly prevented a

[26:22] will have single-handedly prevented a bubble. Okay, you are starting to see

[26:25] bubble. Okay, you are starting to see companies go to Intel

[26:28] companies go to Intel and Samsung.

[26:29] and Samsung. >> Let's just assume TSM stays super supply

[26:31] >> Let's just assume TSM stays super supply constraint versus you know the latent

[26:33] constraint versus you know the latent demand like what what happens?

[26:35] demand like what what happens? >> Well, one of you know the history

[26:38] >> Well, one of you know the history markets is I don't know who but one of

[26:40] markets is I don't know who but one of Intel and Samsung they're not going to

[26:42] Intel and Samsung they're not going to stay disciplined. They will break and

[26:44] stay disciplined. They will break and then at some level that will force

[26:47] then at some level that will force everyone else to break.

[26:50] everyone else to break. So like I think a lot of this may come

[26:52] So like I think a lot of this may come down to the degree to which Taiwan SIM

[26:55] down to the degree to which Taiwan SIM can maintain a lead over Intel and

[26:59] can maintain a lead over Intel and Samsung. You got to remember it's

[27:00] Samsung. You got to remember it's whatever it is it's 9 12 15 months.

[27:02] whatever it is it's 9 12 15 months. >> Sort of like the leading node edge. You

[27:04] >> Sort of like the leading node edge. You mean

[27:04] mean >> exactly you know the pace at which they

[27:07] >> exactly you know the pace at which they expand capacity. Like if I were to watch

[27:10] expand capacity. Like if I were to watch one thing to understand whe there's a

[27:11] one thing to understand whe there's a bubble it's Taiwan Simmy's capacity

[27:13] bubble it's Taiwan Simmy's capacity decisions. And I think there's a

[27:15] decisions. And I think there's a Goldilocks zone where they expand enough

[27:21] Goldilocks zone where they expand enough they make it hard for Intel or Samsung

[27:24] they make it hard for Intel or Samsung to really truly emerge as like a um at

[27:29] to really truly emerge as like a um at scale second source with something you

[27:31] scale second source with something you know well north of 30% market share. And

[27:35] know well north of 30% market share. And yet they also keep this fundamental

[27:38] yet they also keep this fundamental constraint on wafers

[27:41] constraint on wafers that you know helps us avoid a bubble.

[27:44] that you know helps us avoid a bubble. And then obviously I think the terapab

[27:47] And then obviously I think the terapab um is going to play into this too.

[27:48] um is going to play into this too. >> Say more about that for people that

[27:51] >> Say more about that for people that >> the turfab it's a SpaceX I believe

[27:53] >> the turfab it's a SpaceX I believe Tesla's involved as well um joint

[27:56] Tesla's involved as well um joint venture to build the world's largest fab

[27:59] venture to build the world's largest fab here in America and I'm I think they're

[28:03] here in America and I'm I think they're going to be successful. on they have a

[28:05] going to be successful. on they have a partnership with Intel which is very

[28:06] partnership with Intel which is very important um because they're getting

[28:08] important um because they're getting access to 50 years of institutional

[28:11] access to 50 years of institutional knowledge that's just you know a nine

[28:13] knowledge that's just you know a nine months a few quarters 12 months 3 to

[28:16] months a few quarters 12 months 3 to five quarters behind the front that's an

[28:17] five quarters behind the front that's an advantage it's also an advantage that I

[28:21] advantage it's also an advantage that I believe that terafab is going to get

[28:23] believe that terafab is going to get attention from the a teams at all the

[28:25] attention from the a teams at all the semicap equipment companies like one big

[28:27] semicap equipment companies like one big reason Taiwan semicought up is ASML and

[28:31] reason Taiwan semicought up is ASML and KA tenor and lamb research and applied

[28:33] KA tenor and lamb research and applied material materials. They wanted them to

[28:35] material materials. They wanted them to catch up. They didn't they don't like

[28:37] catch up. They didn't they don't like having a monopsiny and so the A teams

[28:40] having a monopsiny and so the A teams were in Taiwan working. Intel made some

[28:42] were in Taiwan working. Intel made some mistakes and presto. And so the A teams

[28:46] mistakes and presto. And so the A teams will will be here because of Elon's

[28:48] will will be here because of Elon's reputation in in hardware engineering.

[28:51] reputation in in hardware engineering. And then just to a degree that I think

[28:54] And then just to a degree that I think is u maybe hard for people to imagine in

[28:58] is u maybe hard for people to imagine in America um where you know politics has

[29:00] America um where you know politics has replaced religion because Elon had his

[29:02] replaced religion because Elon had his fora into politics that makes it hard

[29:04] fora into politics that makes it hard for some people in America to see him

[29:07] for some people in America to see him clearly which is sad because I do think

[29:10] clearly which is sad because I do think you know he's probably doing more for

[29:11] you know he's probably doing more for America than any other American. You

[29:14] America than any other American. You know he's single-handedly bringing

[29:16] know he's single-handedly bringing manufacturing back to America. He's

[29:18] manufacturing back to America. He's revived Dince Tech. SpaceX is in some

[29:21] revived Dince Tech. SpaceX is in some ways the most important defense

[29:22] ways the most important defense contractor in America. You know, what

[29:24] contractor in America. You know, what he's doing with Starlink is amazing for

[29:26] he's doing with Starlink is amazing for the world. He's creating all these blue

[29:29] the world. He's creating all these blue collar manufacturing jobs, which is like

[29:30] collar manufacturing jobs, which is like a goal, I think, of a lot of liberals

[29:32] a goal, I think, of a lot of liberals and good for America. He's done more

[29:34] and good for America. He's done more than any living human to decarbonize the

[29:37] than any living human to decarbonize the world. And if you are upset about data

[29:39] world. And if you are upset about data sitters on Earth for environmental

[29:41] sitters on Earth for environmental reasons, well, here you go. You know, so

[29:45] reasons, well, here you go. You know, so it's it's sad, but he is a living deity.

[29:49] it's it's sad, but he is a living deity. in China, Taiwan, South Korea, and

[29:53] in China, Taiwan, South Korea, and Japan.

[29:55] Japan. And having watched him for a long time,

[29:59] And having watched him for a long time, what he's going to do is they're going

[30:00] what he's going to do is they're going to recruit the best people because the

[30:04] to recruit the best people because the best engineers want to work for Elon,

[30:08] best engineers want to work for Elon, especially in hardware engineering. He's

[30:10] especially in hardware engineering. He's going to recruit incredible engineers.

[30:12] going to recruit incredible engineers. And then they'll be next to next to

[30:14] And then they'll be next to next to Turfab, they'll be a Taiwan town. Oh,

[30:17] Turfab, they'll be a Taiwan town. Oh, these are your favorite restaurants. I'm

[30:19] these are your favorite restaurants. I'm going to move them and their whole staff

[30:21] going to move them and their whole staff from Taiwan to Texas and we're going to

[30:24] from Taiwan to Texas and we're going to make everything the way they like it.

[30:26] make everything the way they like it. And then we'll have Japan Town. Same

[30:28] And then we'll have Japan Town. Same thing. Then we're going to have Korea

[30:29] thing. Then we're going to have Korea Town. We're going to have all these

[30:30] Town. We're going to have all these things exactly but dialed to recruit the

[30:35] things exactly but dialed to recruit the best engineers. And that's just not the

[30:39] best engineers. And that's just not the way that the people who run Intel at

[30:42] way that the people who run Intel at Seung think. So he's going to have the

[30:44] Seung think. So he's going to have the best talent. He's going to have the A

[30:46] best talent. He's going to have the A teams at the wafer fab equipment

[30:48] teams at the wafer fab equipment companies. He's he has intel which is

[30:51] companies. He's he has intel which is important. It's so good for all of any

[30:54] important. It's so good for all of any administration's political goals. And I

[30:56] administration's political goals. And I think it's different enough that it will

[30:58] think it's different enough that it will not alienate Taiwan SMI.

[31:00] not alienate Taiwan SMI. >> And these have long lead times, right?

[31:02] >> And these have long lead times, right? So like Terrafab is going to be pumping

[31:04] So like Terrafab is going to be pumping out Nvidia G or whatever GPUs, whatever

[31:06] out Nvidia G or whatever GPUs, whatever chips like quite quite a long time from

[31:08] chips like quite quite a long time from now.

[31:09] now. >> Elon tends to do things differently.

[31:10] >> Elon tends to do things differently. Everybody else has taken three years to

[31:12] Everybody else has taken three years to build a data center. He built one in 122

[31:14] build a data center. He built one in 122 days. You know, Samsung had to give him

[31:17] days. You know, Samsung had to give him an office in their fab in Texas because

[31:20] an office in their fab in Texas because he was so unhappy about like the pace at

[31:22] he was so unhappy about like the pace at which they're expanding a building.

[31:24] which they're expanding a building. We'll see. Are you surprised by you

[31:27] We'll see. Are you surprised by you mentioned Deep Seek earlier? The simple

[31:29] mentioned Deep Seek earlier? The simple reaction to that was okay, these models

[31:31] reaction to that was okay, these models are just going to get 95% as effective

[31:34] are just going to get 95% as effective for some tiny fraction of the cost to

[31:36] for some tiny fraction of the cost to still Chinese open source models. Like

[31:38] still Chinese open source models. Like we'll be able to use these for most of

[31:39] we'll be able to use these for most of what we want to do. Fast forwarded a

[31:41] what we want to do. Fast forwarded a little bit of time, you know, two years

[31:43] little bit of time, you know, two years from now, there's no reason I have to

[31:45] from now, there's no reason I have to spend a million dollars a year in my

[31:46] spend a million dollars a year in my small little firm on on tokens or

[31:48] small little firm on on tokens or something. But then the actual reality

[31:50] something. But then the actual reality seems quite different than this. And I'm

[31:52] seems quite different than this. And I'm curious why there's that dissonance in

[31:54] curious why there's that dissonance in your mind.

[31:54] your mind. >> I do think it's the fascinating the

[31:56] >> I do think it's the fascinating the returns to the frontier, all the

[31:58] returns to the frontier, all the economic returns to AI at the model

[32:01] economic returns to AI at the model layer, not all of them, but an

[32:04] layer, not all of them, but an overwhelming amount of them have been at

[32:05] overwhelming amount of them have been at the frontier, which is surprising to me.

[32:09] the frontier, which is surprising to me. And I think it's been surprising to a

[32:11] And I think it's been surprising to a lot of people and I think this is one of

[32:15] lot of people and I think this is one of the most important questions to be

[32:18] the most important questions to be answered and you need to have a

[32:19] answered and you need to have a hypothesis on it as an investor. Are

[32:21] hypothesis on it as an investor. Are frontier tokens going to continue

[32:24] frontier tokens going to continue capturing the overwhelming majority of

[32:27] capturing the overwhelming majority of economic value created at the model

[32:29] economic value created at the model layer? And it is surprising like I just

[32:31] layer? And it is surprising like I just I remember when Gemini 3.1 Pro came out

[32:35] I remember when Gemini 3.1 Pro came out and it was it was mind-blowing to me. It

[32:37] and it was it was mind-blowing to me. It was so good. And today it's intolerable.

[32:42] was so good. And today it's intolerable. >> Intolerable.

[32:43] >> Intolerable. And you know there's probably a little

[32:45] And you know there's probably a little bit of a dynamic where companies

[32:46] bit of a dynamic where companies prototype with Frontiers then when they

[32:48] prototype with Frontiers then when they put something into production you're

[32:50] put something into production you're hearing a lot of people do use Vertex or

[32:52] hearing a lot of people do use Vertex or you know open source. But still it is it

[32:56] you know open source. But still it is it is a fact today that the overwhelming

[32:58] is a fact today that the overwhelming majority of these economic turns come

[33:00] majority of these economic turns come from Frontier tokens. And that's

[33:02] from Frontier tokens. And that's surprising and whether or not it

[33:04] surprising and whether or not it continues I think is a very interesting

[33:07] continues I think is a very interesting question. And I'm much more open-minded

[33:09] question. And I'm much more open-minded to that having had the experience I've

[33:11] to that having had the experience I've had with Gemini 3.1

[33:14] had with Gemini 3.1 and then Opus. Um, and then I do use Gro

[33:18] and then Opus. Um, and then I do use Gro 4.3. It is on the paro frontier. like

[33:21] 4.3. It is on the paro frontier. like the companies that are on the paro

[33:22] the companies that are on the paro frontier are and this is by the way a

[33:24] frontier are and this is by the way a big change in a a consequence of what we

[33:26] big change in a a consequence of what we talked about last time. Google losing

[33:28] talked about last time. Google losing their percost token leadership as a

[33:31] their percost token leadership as a result of making very conservative

[33:32] result of making very conservative design decisions with TPU V8 to try and

[33:35] design decisions with TPU V8 to try and take it away partially from Broadcom and

[33:37] take it away partially from Broadcom and Nvidia um continuing to make aggressive

[33:40] Nvidia um continuing to make aggressive choices. Uh but Google dominated the

[33:43] choices. Uh but Google dominated the prao frontier. The prao frontier being

[33:45] prao frontier. The prao frontier being intelligence versus cost. And I think

[33:47] intelligence versus cost. And I think this is the most important thing to look

[33:48] this is the most important thing to look at to analyze AI labs. Google dominated

[33:51] at to analyze AI labs. Google dominated that nine months ago. They at every

[33:53] that nine months ago. They at every point on the paro frontier. OpenAI, XAI

[33:58] point on the paro frontier. OpenAI, XAI and Anthropic were inside of them. Now

[34:01] and Anthropic were inside of them. Now the Paro frontier is dominated by

[34:03] the Paro frontier is dominated by Enthropic, OpenAI. And then Grock 4.3 is

[34:06] Enthropic, OpenAI. And then Grock 4.3 is on the paro frontier. It's clearly like

[34:09] on the paro frontier. It's clearly like the, you know, the best lowest cost 500

[34:11] the, you know, the best lowest cost 500 billion parameter model. And then Gemini

[34:14] billion parameter model. And then Gemini 3.1 is like hanging on to the paro

[34:17] 3.1 is like hanging on to the paro frontier. And if I were to bet or bet

[34:19] frontier. And if I were to bet or bet that they're subsidizing that out of

[34:20] that they're subsidizing that out of pride, I would just say one a violation

[34:23] pride, I would just say one a violation of Richard Sutton's bitter lesson is for

[34:25] of Richard Sutton's bitter lesson is for sure the biggest risk to this trade

[34:27] sure the biggest risk to this trade >> to all of AI. Now the closer someone is

[34:30] >> to all of AI. Now the closer someone is to AI, the more skeptical they are this

[34:32] to AI, the more skeptical they are this will occur. One thing I think

[34:34] will occur. One thing I think contributed to weakness in March was,

[34:36] contributed to weakness in March was, you know, a much more stupid version of

[34:39] you know, a much more stupid version of DeepSeek, which was this thing called

[34:40] DeepSeek, which was this thing called Turboquant. and Turbo Quan is some

[34:42] Turboquant. and Turbo Quan is some Google memory optimization that was

[34:44] Google memory optimization that was written up in a paper a year ago. And

[34:46] written up in a paper a year ago. And then during the middle of an agreement

[34:49] then during the middle of an agreement while Google was negotiating with

[34:51] while Google was negotiating with Micron, Samsung and Highex to sign, you

[34:53] Micron, Samsung and Highex to sign, you know, some LTA that would lock in really

[34:55] know, some LTA that would lock in really high prices for a long time. They

[34:57] high prices for a long time. They released this. You know, what people do

[34:59] released this. You know, what people do is always more important than they say.

[35:00] is always more important than they say. And they just kind of publicize it on X

[35:03] And they just kind of publicize it on X and it goes viral like, "Oh my god, DRM

[35:06] and it goes viral like, "Oh my god, DRM is cooked. Here's this DRAM

[35:08] is cooked. Here's this DRAM optimization." I was unable to find a

[35:10] optimization." I was unable to find a single AI engineer on planet earth who

[35:13] single AI engineer on planet earth who believed that turbo quant would have any

[35:15] believed that turbo quant would have any impact on DRAM demand but nonetheless a

[35:18] impact on DRAM demand but nonetheless a violation of Richard Sutton's bitter

[35:20] violation of Richard Sutton's bitter lesson you know more compute will always

[35:21] lesson you know more compute will always outperform human algorithmic ingenuity

[35:23] outperform human algorithmic ingenuity more computing data or chin beyond

[35:25] more computing data or chin beyond chinchilla optimal I guess what what

[35:27] chinchilla optimal I guess what what people increasingly do today that's a

[35:30] people increasingly do today that's a real risk man and I think the people who

[35:33] real risk man and I think the people who are building these models are skeptical

[35:35] are building these models are skeptical of that risk the reason I am a little

[35:38] of that risk the reason I am a little less skep skeptical is I think we are

[35:40] less skep skeptical is I think we are very close to ASI and who knows if the

[35:43] very close to ASI and who knows if the bitter lesson holds for 400 IQ models

[35:47] bitter lesson holds for 400 IQ models just you know or maybe we get a

[35:49] just you know or maybe we get a temporary

[35:51] temporary period where these you know if you get

[35:52] period where these you know if you get to ASI the first thing it wants is

[35:55] to ASI the first thing it wants is probably to be smarter and have more

[35:57] probably to be smarter and have more resources. How does it do that? It makes

[35:58] resources. How does it do that? It makes itself more efficient. I think that is

[36:02] itself more efficient. I think that is an actual risk that humans the bitter

[36:06] an actual risk that humans the bitter lesson literally I believe includes

[36:08] lesson literally I believe includes humans in it. So we're about to find out

[36:11] humans in it. So we're about to find out whether the bitter lesson we'll find out

[36:12] whether the bitter lesson we'll find out if it applies to 300 IQ ahis then 400

[36:16] if it applies to 300 IQ ahis then 400 then 500 and 600 and at some point we

[36:20] then 500 and 600 and at some point we may have like a temporary violation of

[36:22] may have like a temporary violation of the bitter lesson based upon AI and ASI.

[36:27] the bitter lesson based upon AI and ASI. So I'm curious how you think about some

[36:29] So I'm curious how you think about some other parts of the innovation around the

[36:32] other parts of the innovation around the model continual learning and memory

[36:34] model continual learning and memory being two that see people seem to be

[36:36] being two that see people seem to be most focused on as things that might

[36:37] most focused on as things that might create yet another you know new paradigm

[36:39] create yet another you know new paradigm that we would enter. What do you think

[36:40] that we would enter. What do you think about the role of those two things?

[36:42] about the role of those two things? Yeah. Well, I think we've done a lot

[36:43] Yeah. Well, I think we've done a lot with memory through these harnesses. And

[36:46] with memory through these harnesses. And it turns out that harness engineering is

[36:49] it turns out that harness engineering is not as important as the model, but it

[36:52] not as important as the model, but it really matters. And these harnesses in

[36:54] really matters. And these harnesses in these models are increasingly being

[36:56] these models are increasingly being co-developed. One of the big things a

[36:58] co-developed. One of the big things a harness does, which you just think of as

[36:59] harness does, which you just think of as like a a runtime that the model operates

[37:03] like a a runtime that the model operates in, knows where the pool tools are. It

[37:06] in, knows where the pool tools are. It like creates context, memory, state, um,

[37:11] like creates context, memory, state, um, you know, has very specific,

[37:13] you know, has very specific, you know, prompts or instructions and

[37:16] you know, prompts or instructions and just

[37:16] just >> makes a huge difference. Even simple

[37:18] >> makes a huge difference. Even simple versions,

[37:19] versions, >> it makes an incredible difference. I

[37:20] >> it makes an incredible difference. I think the last time I was on here or one

[37:22] think the last time I was on here or one of the other times I just said like,

[37:23] of the other times I just said like, "Hey, as an investor, it's very

[37:26] "Hey, as an investor, it's very important that you pay for the $250 a

[37:29] important that you pay for the $250 a month version to get like your own

[37:31] month version to get like your own intuitive sense." that's no longer

[37:33] intuitive sense." that's no longer possible to understand what frontier AI

[37:35] possible to understand what frontier AI is capable of today even for like a

[37:39] is capable of today even for like a non-coding use case you need to have

[37:40] non-coding use case you need to have cloud code or codeex and you need to be

[37:43] cloud code or codeex and you need to be on an enterprise plan and the reason for

[37:45] on an enterprise plan and the reason for this is and this is another I think this

[37:49] this is and this is another I think this is another dynamic that's enabled by

[37:51] is another dynamic that's enabled by Google losing their cost leadership is

[37:54] Google losing their cost leadership is these AI models just shifted to

[37:56] these AI models just shifted to usagebased pricing and if you're on that

[37:59] usagebased pricing and if you're on that $250 or$300 or $280 month plan or

[38:01] $250 or$300 or $280 month plan or whatever it is you are getting severely

[38:04] whatever it is you are getting severely rate limited. You are getting a

[38:06] rate limited. You are getting a labbotomized version of the AI because

[38:09] labbotomized version of the AI because like we talked about Claude now produces

[38:11] like we talked about Claude now produces 70% less tokens. You want the tokens

[38:13] 70% less tokens. You want the tokens that Claude and its harness really think

[38:16] that Claude and its harness really think it needs to produce to get you a good

[38:17] it needs to produce to get you a good answer, you need to be on a usage based

[38:19] answer, you need to be on a usage based plan. And by the way, this is so bullish

[38:23] plan. And by the way, this is so bullish for AI. I was a telecom analyst in ' 05

[38:25] for AI. I was a telecom analyst in ' 05 to07 and cellular had been a great

[38:28] to07 and cellular had been a great growth industry really for the last 10

[38:30] growth industry really for the last 10 years and the reason was you had a

[38:32] years and the reason was you had a combination of fixed pricing you had 900

[38:35] combination of fixed pricing you had 900 minutes for whatever it was and then

[38:37] minutes for whatever it was and then usage based pricing over that and when

[38:40] usage based pricing over that and when did cellular stop being a great growth

[38:42] did cellular stop being a great growth industry when everybody just went to all

[38:44] industry when everybody just went to all you can eat. And and by the way long

[38:47] you can eat. And and by the way long distance was the same thing. AI is just

[38:48] distance was the same thing. AI is just shifting from all you can eat to pay by

[38:51] shifting from all you can eat to pay by the drink. And it turns out people

[38:52] the drink. And it turns out people really like to talk to their friends

[38:54] really like to talk to their friends long distance. They really like to talk

[38:56] long distance. They really like to talk to their friends on the phone. And

[38:57] to their friends on the phone. And people really like to use AI and

[39:00] people really like to use AI and particularly now that one person can

[39:02] particularly now that one person can have a 100 agents working. So I think

[39:03] have a 100 agents working. So I think the shift to usage based pricing is

[39:07] the shift to usage based pricing is probably why you will see OpenAI and

[39:10] probably why you will see OpenAI and Anthropic exceed well over $200 billion

[39:13] Anthropic exceed well over $200 billion in ARR this year. because not only is

[39:15] in ARR this year. because not only is more compute going to become online, but

[39:17] more compute going to become online, but they're going to be able to push

[39:19] they're going to be able to push frontier token pricing with these usage

[39:21] frontier token pricing with these usage enterprise models, but it's it's sad.

[39:24] enterprise models, but it's it's sad. It's sad for the world and because it

[39:26] It's sad for the world and because it just means if you can't afford that,

[39:27] just means if you can't afford that, you're not at the frontier. But yeah,

[39:30] you're not at the frontier. But yeah, continual learning, man, I mean, if we

[39:32] continual learning, man, I mean, if we solve that,

[39:32] solve that, >> how do you conceptualize that? Like

[39:34] >> how do you conceptualize that? Like >> there's so many mysteries about the

[39:35] >> there's so many mysteries about the human mind, like we're such sample

[39:38] human mind, like we're such sample efficient learners relative to AI. Like

[39:42] efficient learners relative to AI. Like I forget what it is, but like an AI

[39:44] I forget what it is, but like an AI needs orders of magnitude.

[39:45] needs orders of magnitude. >> Yeah. Many orders of magnitude. Now we

[39:47] >> Yeah. Many orders of magnitude. Now we have a crude variant of continual

[39:49] have a crude variant of continual learning today when something is

[39:51] learning today when something is verifiable and that's just, you know,

[39:53] verifiable and that's just, you know, reinforcement learning during

[39:54] reinforcement learning during mid-training. But yeah, continual

[39:56] mid-training. But yeah, continual learning is a model that dynamically

[39:58] learning is a model that dynamically adjusts its weights or adjusts in some

[40:02] adjusts its weights or adjusts in some way in real time. Like as a human,

[40:05] way in real time. Like as a human, >> that's what you do.

[40:06] >> that's what you do. >> Yeah. Like if I the first time I touch

[40:08] >> Yeah. Like if I the first time I touch or you know put my hand in a fire, I've

[40:11] or you know put my hand in a fire, I've learned I never put it in there before.

[40:13] learned I never put it in there before. That model today needs to put its hand

[40:15] That model today needs to put its hand in the fire a million times and then

[40:18] in the fire a million times and then have, you know, the designers

[40:20] have, you know, the designers effectively put a fire in the next

[40:23] effectively put a fire in the next training run or an RL gym for it to

[40:26] training run or an RL gym for it to learn. I think it has to be dynamically

[40:29] learn. I think it has to be dynamically updating the weights, but I think people

[40:31] updating the weights, but I think people are working on really smart techniques

[40:33] are working on really smart techniques beyond this. But if we get that then we

[40:37] beyond this. But if we get that then we have a really fast takeoff and people

[40:40] have a really fast takeoff and people seem

[40:42] seem confident that continual learning is

[40:45] confident that continual learning is kind of just around the corner. And I do

[40:47] kind of just around the corner. And I do think this is like the third big

[40:50] think this is like the third big question. Bitter le violation as a

[40:52] question. Bitter le violation as a result of ASI or less likely human

[40:55] result of ASI or less likely human ingenuity. Will Frontier tokens still

[40:57] ingenuity. Will Frontier tokens still command the premium they do? And will we

[41:00] command the premium they do? And will we get continual learning? And if so, when?

[41:02] get continual learning? And if so, when? As your business scales up, everything

[41:04] As your business scales up, everything gets more complex, especially your

[41:05] gets more complex, especially your compliance and security needs. With so

[41:07] compliance and security needs. With so many tools offering band-aids and

[41:09] many tools offering band-aids and patches, it's unfortunately far too easy

[41:11] patches, it's unfortunately far too easy for something to slip through the

[41:12] for something to slip through the cracks. Fortunately, Vanta is a powerful

[41:14] cracks. Fortunately, Vanta is a powerful tool designed to simplify and automate

[41:16] tool designed to simplify and automate your security work and deliver a single

[41:18] your security work and deliver a single source of truth for compliance and risk.

[41:20] source of truth for compliance and risk. There's a reason that Ramp, Cursor, and

[41:22] There's a reason that Ramp, Cursor, and Snowflake all use Vanta. It frees them

[41:24] Snowflake all use Vanta. It frees them to focus on building amazing

[41:26] to focus on building amazing differentiated products knowing that

[41:27] differentiated products knowing that compliance and security are under

[41:29] compliance and security are under control. Invest like the best listeners.

[41:31] control. Invest like the best listeners. Get a special offer of $1,000 off Vanta

[41:33] Get a special offer of $1,000 off Vanta when you go to vanta.com/invest.

[41:36] when you go to vanta.com/invest. I know firsthand how complex the tech

[41:38] I know firsthand how complex the tech stack is for asset management firms. And

[41:40] stack is for asset management firms. And seemingly every new tool and data source

[41:42] seemingly every new tool and data source makes the problem even worse, adding

[41:44] makes the problem even worse, adding more complexity, more headcount, and

[41:46] more complexity, more headcount, and more risk. Ridgeline offers a better way

[41:48] more risk. Ridgeline offers a better way forward. One unified platform that

[41:49] forward. One unified platform that automates away that complexity across

[41:51] automates away that complexity across portfolio accounting, reconciliation,

[41:53] portfolio accounting, reconciliation, reporting, trading, compliance, and

[41:55] reporting, trading, compliance, and more. All at scale. Ridgeline is

[41:58] more. All at scale. Ridgeline is revolutionizing investment management.

[41:59] revolutionizing investment management. helping ambitious firms scale faster,

[42:01] helping ambitious firms scale faster, operate smarter, and stay ahead of the

[42:02] operate smarter, and stay ahead of the curve. See what Ridgeline can unlock for

[42:04] curve. See what Ridgeline can unlock for your firm. Schedule a demo at

[42:06] your firm. Schedule a demo at ridgeline.ai.

[42:07] ridgeline.ai. What is the role of new chip companies

[42:10] What is the role of new chip companies in all of this? Like we talked a lot

[42:11] in all of this? Like we talked a lot about Nvidia and you know their their

[42:13] about Nvidia and you know their their sort of relationship with TSMC and Intel

[42:15] sort of relationship with TSMC and Intel and all these sorts of things. There's a

[42:17] and all these sorts of things. There's a thousand flowers blooming. I think

[42:19] thousand flowers blooming. I think literally probably a thousand flowers

[42:20] literally probably a thousand flowers blooming trying to create a new chip to

[42:23] blooming trying to create a new chip to address some part of this bottleneck.

[42:26] address some part of this bottleneck. I'm curious how you process this space,

[42:28] I'm curious how you process this space, this opportunity, what role it will

[42:29] this opportunity, what role it will play, what role they'll play.

[42:31] play, what role they'll play. >> So, I think this is good and healthy for

[42:32] >> So, I think this is good and healthy for the world. It's good for Jensen too. Um,

[42:35] the world. It's good for Jensen too. Um, you know, because a different

[42:36] you know, because a different administration might take a different

[42:39] administration might take a different view. Competition, I think, is good for

[42:41] view. Competition, I think, is good for everyone. In in tank design, they talk

[42:42] everyone. In in tank design, they talk about the iron triangle. The iron

[42:44] about the iron triangle. The iron triangle of tank design is that all

[42:46] triangle of tank design is that all designers of a tank, they have to make

[42:48] designers of a tank, they have to make trade-offs between attack, defense, and

[42:50] trade-offs between attack, defense, and mobility. And you know, for obvious

[42:51] mobility. And you know, for obvious reasons, the more defense you have,

[42:53] reasons, the more defense you have, which is just armor, the heavier the

[42:55] which is just armor, the heavier the tank is, the less mobile it is. So you

[42:58] tank is, the less mobile it is. So you have to live in this triangle and make

[43:00] have to live in this triangle and make tradeoffs. Okay? Like the marava in

[43:03] tradeoffs. Okay? Like the marava in Israel, it's optimized for defense.

[43:05] Israel, it's optimized for defense. Russian tanks and like the Leopard are

[43:08] Russian tanks and like the Leopard are generally more optimized for mobility.

[43:10] generally more optimized for mobility. Chip design is the same. And you there

[43:13] Chip design is the same. And you there there are these fundamental constraints

[43:16] there are these fundamental constraints imposed by the laws of physics has

[43:19] imposed by the laws of physics has embedded in the Taiwan semi design rules

[43:22] embedded in the Taiwan semi design rules that you need to live within and you

[43:26] that you need to live within and you have TPU tranium and AMD which are all

[43:30] have TPU tranium and AMD which are all um you know essentially trying to be a

[43:34] um you know essentially trying to be a better GPU and today I think probably

[43:37] better GPU and today I think probably Tranium is doing the best. Nobody's a

[43:39] Tranium is doing the best. Nobody's a better GPU, but Trrenium is is I think

[43:42] better GPU, but Trrenium is is I think their, you know, they're they're tugging

[43:44] their, you know, they're they're tugging on Superman's cape

[43:46] on Superman's cape >> and and this hadn't started yet. The

[43:48] >> and and this hadn't started yet. The Tranium 3 needs to ramp into production

[43:50] Tranium 3 needs to ramp into production because it has a switch scale up

[43:52] because it has a switch scale up network, which you really need to

[43:53] network, which you really need to economically inference models. You know,

[43:56] economically inference models. You know, a lot of companies have a Taurus

[43:58] a lot of companies have a Taurus architecture. Um that that's where

[44:00] architecture. Um that that's where Google was and AMD. We'll see. The

[44:03] Google was and AMD. We'll see. The MI450, we don't know yet. We'll see. We

[44:06] MI450, we don't know yet. We'll see. We probably know more about Trinidium 3

[44:07] probably know more about Trinidium 3 than the MI450, but that's a hard game

[44:10] than the MI450, but that's a hard game to play. So you have to do something

[44:14] to play. So you have to do something different and you have to do something

[44:16] different and you have to do something different that is also hard to do. So I

[44:20] different that is also hard to do. So I think the best path for these startups

[44:22] think the best path for these startups like my rule of thumb is 1% market share

[44:24] like my rule of thumb is 1% market share is going to be worth 100 billion. 100

[44:26] is going to be worth 100 billion. 100 billion is a pretty good venture

[44:27] billion is a pretty good venture outcome. I think what Jensen would say

[44:29] outcome. I think what Jensen would say is like, okay, if something somebody

[44:30] is like, okay, if something somebody does something different and it gets to

[44:33] does something different and it gets to one or two or 3% share, we'll make that

[44:36] one or two or 3% share, we'll make that chip and that's that's coming for

[44:39] chip and that's that's coming for everyone. But if you're trying to make a

[44:41] everyone. But if you're trying to make a better GPU, good luck. If you were doing

[44:43] better GPU, good luck. If you were doing something different, it also needs to be

[44:47] something different, it also needs to be hard to do. And you can make different

[44:49] hard to do. And you can make different trade-offs. you know, the disagregation

[44:51] trade-offs. you know, the disagregation of prefill and inference really have

[44:53] of prefill and inference really have opened the aperture um for making these

[44:56] opened the aperture um for making these different trade-offs because you can

[44:58] different trade-offs because you can make very aggressive trade-offs for

[44:59] make very aggressive trade-offs for decode, aggressive trade-offs for

[45:01] decode, aggressive trade-offs for prefill.

[45:02] prefill. >> Prefill being taking in the context,

[45:03] >> Prefill being taking in the context, decode being, you know, write the

[45:05] decode being, you know, write the output.

[45:06] output. >> Yeah. I have a great colleague named

[45:07] >> Yeah. I have a great colleague named Andrew Fox who said, "Picture, you know,

[45:09] Andrew Fox who said, "Picture, you know, a British naval ship from the 18th

[45:11] a British naval ship from the 18th century. Prefill is loading the cannon,

[45:13] century. Prefill is loading the cannon, decode is firing it." And what prefill

[45:15] decode is firing it." And what prefill literally is is just the model

[45:17] literally is is just the model understanding the question, the prompt

[45:19] understanding the question, the prompt and then kind of keeping track of its

[45:20] and then kind of keeping track of its own dec.

[45:22] own dec. And that is fundamentally a memory

[45:24] And that is fundamentally a memory capacity bound problem. Decode is a

[45:27] capacity bound problem. Decode is a process of generating new tokens and

[45:28] process of generating new tokens and that is memory bandwidth constraint. And

[45:31] that is memory bandwidth constraint. And so if you're a chip designer, this gives

[45:33] so if you're a chip designer, this gives you a richer canvas to to paint on. But

[45:36] you a richer canvas to to paint on. But even so, it needs to be hard because if

[45:39] even so, it needs to be hard because if you make different trade-offs in that

[45:40] you make different trade-offs in that iron triangle to optimize for memory

[45:43] iron triangle to optimize for memory capacity, and they're not hard

[45:44] capacity, and they're not hard trade-offs to make, well then Nvidia is

[45:47] trade-offs to make, well then Nvidia is going to make those same trade-offs.

[45:49] going to make those same trade-offs. They get better prices from Taiwan Semi

[45:52] They get better prices from Taiwan Semi than you're ever going to get. Um, and

[45:54] than you're ever going to get. Um, and good luck. Good luck. And they have the

[45:56] good luck. Good luck. And they have the advantage of working with every model

[45:58] advantage of working with every model company and optimizing their designs.

[46:00] company and optimizing their designs. And by the way, another very funny thing

[46:02] And by the way, another very funny thing is if you're a VC

[46:04] is if you're a VC and you're investing in semiconductor

[46:06] and you're investing in semiconductor company that is telling you they are

[46:08] company that is telling you they are going to have an advantage because of a

[46:10] going to have an advantage because of a Taiwan semi process that they have

[46:12] Taiwan semi process that they have special access to. I promise you that

[46:15] special access to. I promise you that Jensen saw that process when it was a

[46:19] Jensen saw that process when it was a twinkle in Taiwan Simmy's eyes and it

[46:22] twinkle in Taiwan Simmy's eyes and it they know more about it than this little

[46:25] they know more about it than this little company with 200 people can imagine.

[46:28] company with 200 people can imagine. Taiwan, CMI, everybody in the supply

[46:30] Taiwan, CMI, everybody in the supply chain is showing Jensen everything the

[46:32] chain is showing Jensen everything the same way they're showing Amazon

[46:34] same way they're showing Amazon everything, AMD everything, TPU

[46:37] everything, AMD everything, TPU everything. And that's another reason

[46:38] everything. And that's another reason don't go try to make a better GPU. So

[46:41] don't go try to make a better GPU. So you can do something different. You can

[46:42] you can do something different. You can paint in the pre-filled canvas. You can

[46:44] paint in the pre-filled canvas. You can paint in the decode canvas, but you also

[46:46] paint in the decode canvas, but you also have to do something hard because if it

[46:48] have to do something hard because if it gets to scale, you're going to have

[46:51] gets to scale, you're going to have those four companies has very fast

[46:53] those four companies has very fast followers. My firm was a was a um

[46:56] followers. My firm was a was a um venture investor in Cerebras. What

[46:59] venture investor in Cerebras. What Cerebrris has done is something hard and

[47:00] Cerebrris has done is something hard and fundamentally different way for scale

[47:02] fundamentally different way for scale computing. And it it comes with a set of

[47:05] computing. And it it comes with a set of trade-offs, but that architectural

[47:08] trade-offs, but that architectural decision they made was hard and lets

[47:11] decision they made was hard and lets them do something that no one else can

[47:13] them do something that no one else can do. And we'll find out how big that is.

[47:17] do. And we'll find out how big that is. And you know, they're working on really

[47:18] And you know, they're working on really cool things like um one of the problems

[47:21] cool things like um one of the problems Cerebras has. Once you start needing to

[47:22] Cerebras has. Once you start needing to glue a lot of chips together and scale

[47:24] glue a lot of chips together and scale up networks or scale out networks, you

[47:27] up networks or scale out networks, you need a lot of IO and IO is bound by

[47:31] need a lot of IO and IO is bound by what's called the shoreline, the sides

[47:32] what's called the shoreline, the sides of the chip. And so Cerebris has an

[47:35] of the chip. And so Cerebris has an overwhelming ratio of onchip computed

[47:37] overwhelming ratio of onchip computed memory relative to shoreline IO. Well,

[47:40] memory relative to shoreline IO. Well, they're really smart people. They did

[47:41] they're really smart people. They did something really hard. They're trying to

[47:43] something really hard. They're trying to see if they can put an optical wafer

[47:45] see if they can put an optical wafer right on top of that and then that

[47:46] right on top of that and then that solves that problem. Um, I'm sure

[47:48] solves that problem. Um, I'm sure they're looking at hybrid bonding of

[47:50] they're looking at hybrid bonding of DRAM, you know, to get around these

[47:52] DRAM, you know, to get around these alleged limitations that are not true. A

[47:54] alleged limitations that are not true. A cerebrus machine can theoretically run

[47:56] cerebrus machine can theoretically run any size model. There are sizes of

[47:58] any size model. There are sizes of models where they're much better than

[47:59] models where they're much better than other sizes. So, Cerebras, what I think

[48:02] other sizes. So, Cerebras, what I think is interesting is they did something

[48:03] is interesting is they did something different that's hard to do, really hard

[48:05] different that's hard to do, really hard to do wafer scale computing. So, I do

[48:08] to do wafer scale computing. So, I do think there's a role for these and you

[48:10] think there's a role for these and you know, I would just encourage them all

[48:12] know, I would just encourage them all make a different trade-off

[48:14] make a different trade-off and try and do something hard. because

[48:18] and try and do something hard. because everybody's going to get funded after

[48:20] everybody's going to get funded after the Cerebrus IPO. It's not going to be a

[48:22] the Cerebrus IPO. It's not going to be a problem. But it took it took Cerebras

[48:25] problem. But it took it took Cerebras three generations of chips to get it

[48:28] three generations of chips to get it right. And it's really hard. Like Andrew

[48:31] right. And it's really hard. Like Andrew Feldman, the CEO, you can just see

[48:36] Feldman, the CEO, you can just see >> how hard it was

[48:39] >> how hard it was and that whole team did

[48:42] and that whole team did >> to get where they are today. And they

[48:44] >> to get where they are today. And they need to have the grit to do that, the

[48:45] need to have the grit to do that, the resilience. This first chip is a

[48:47] resilience. This first chip is a failure. It happens. Can you come back

[48:48] failure. It happens. Can you come back and make a second chip? But the one last

[48:50] and make a second chip? But the one last thing on this topic, this is going to be

[48:52] thing on this topic, this is going to be amazing for the useful lives of GPUs and

[48:55] amazing for the useful lives of GPUs and may single-handedly save private credit.

[48:57] may single-handedly save private credit. >> Say more about that. What what do you

[48:58] >> Say more about that. What what do you mean by the private credit? Well, just,

[49:00] mean by the private credit? Well, just, you know, private credit, they're in

[49:01] you know, private credit, they're in pain from these SAS loans and however

[49:03] pain from these SAS loans and however much they're marked down, they probably

[49:04] much they're marked down, they probably need to be marked down more because if

[49:06] need to be marked down more because if the public companies are struggling to

[49:07] the public companies are struggling to adapt, how's like a debtleaden company

[49:10] adapt, how's like a debtleaden company going going to adapt um and invest in

[49:13] going going to adapt um and invest in what is a very different margin

[49:15] what is a very different margin structure business? But there's a lot of

[49:17] structure business? But there's a lot of private credit and GPUs too. They were

[49:19] private credit and GPUs too. They were underwriting that to I think three or

[49:21] underwriting that to I think three or four years. And the disagregation of

[49:24] four years. And the disagregation of inference means that I think these GPUs

[49:28] inference means that I think these GPUs are going to have 10 or 15 year lives.

[49:30] are going to have 10 or 15 year lives. The AI skeptics are like, oh, these

[49:32] The AI skeptics are like, oh, these companies are all cooking their books.

[49:33] companies are all cooking their books. You know, the useful life of GU GPU is

[49:35] You know, the useful life of GU GPU is only a year or two. The useful life of a

[49:37] only a year or two. The useful life of a CPU is only four years because the rapid

[49:39] CPU is only four years because the rapid technological change. No. What rapid

[49:42] technological change. No. What rapid technological change has done with the

[49:44] technological change has done with the disagregation of pre-fill and inference

[49:46] disagregation of pre-fill and inference is mean that you you know you can put a

[49:48] is mean that you you know you can put a cerebra system or grock LPUs that Nvidia

[49:51] cerebra system or grock LPUs that Nvidia acquired effectively in front of a

[49:53] acquired effectively in front of a hopper or even an ampier use that hopper

[49:56] hopper or even an ampier use that hopper and ampear for prefill and extend the

[49:58] and ampear for prefill and extend the useful life of that GPU

[50:00] useful life of that GPU until it melts. Now they do melts they

[50:02] until it melts. Now they do melts they do melt so they have a time but you know

[50:04] do melt so they have a time but you know maybe you you don't have to run them as

[50:06] maybe you you don't have to run them as fast. This is going to be really good

[50:08] fast. This is going to be really good for the whole private credit industry.

[50:10] for the whole private credit industry. It's going to help finance the AI

[50:11] It's going to help finance the AI buildout because if you can start to

[50:13] buildout because if you can start to finance GPUs at more like you know 5% or

[50:17] finance GPUs at more like you know 5% or 6% instead of I think Corw's lowest

[50:19] 6% instead of I think Corw's lowest financing was like low sevens that

[50:21] financing was like low sevens that actually mathematically changes the cost

[50:23] actually mathematically changes the cost to finance this buildout. We had this

[50:25] to finance this buildout. We had this technological innovation that it's going

[50:27] technological innovation that it's going to lower the cost of financing extend

[50:29] to lower the cost of financing extend the useful life of compute on Earth. And

[50:31] the useful life of compute on Earth. And then I do think the one last thing

[50:32] then I do think the one last thing that's interesting about that is um my

[50:35] that's interesting about that is um my friend Jamon from Kotu just did a

[50:37] friend Jamon from Kotu just did a podcast and Cotu had a deck and they

[50:39] podcast and Cotu had a deck and they talked about hey you know the sellers of

[50:42] talked about hey you know the sellers of shortage are doing so much better than

[50:43] shortage are doing so much better than the buyers of shortage. Buyers of

[50:44] the buyers of shortage. Buyers of shortage being you know the the

[50:46] shortage being you know the the hyperscalers

[50:48] hyperscalers but if you own a giant installed base of

[50:52] but if you own a giant installed base of what is currently in shortage that's

[50:55] what is currently in shortage that's also a very very good place to be. And

[50:57] also a very very good place to be. And we're hearing, you know, CPUs are way

[50:59] we're hearing, you know, CPUs are way more important than they were in an

[51:00] more important than they were in an agentic world. They do all these things

[51:02] agentic world. They do all these things around orchestration, tool calls, etc.,

[51:03] around orchestration, tool calls, etc., etc., etc. The biggest CPU fleets in the

[51:06] etc., etc. The biggest CPU fleets in the world sit at the hyperscalers. So, I

[51:08] world sit at the hyperscalers. So, I think some of these hyperscalers may

[51:09] think some of these hyperscalers may have,

[51:10] have, >> you know, may may catch up a little bit

[51:12] >> you know, may may catch up a little bit to the sellers of shortage.

[51:13] to the sellers of shortage. >> I want to talk about this idea of

[51:15] >> I want to talk about this idea of different and hard applied outside of

[51:17] different and hard applied outside of the infrastructure piece of this. So,

[51:20] the infrastructure piece of this. So, now you're starting to interact with new

[51:21] now you're starting to interact with new founders, um, existing CEOs and founders

[51:24] founders, um, existing CEOs and founders that have to adjust to this new world.

[51:26] that have to adjust to this new world. What are you seeing like the most AI

[51:28] What are you seeing like the most AI native founders that aren't building

[51:30] native founders that aren't building chips or infrastructure or models, but

[51:32] chips or infrastructure or models, but just people using this technology to

[51:34] just people using this technology to build other stuff? How do they feel the

[51:36] build other stuff? How do they feel the most different to you if if you've

[51:38] most different to you if if you've observed differences?

[51:39] observed differences? >> Well, one, I do think this is just for

[51:40] >> Well, one, I do think this is just for chip design. To me, it's always been a

[51:42] chip design. To me, it's always been a fundamental question for venture. So,

[51:44] fundamental question for venture. So, there are different ideas that are

[51:47] there are different ideas that are obvious to everyone on planet Earth as

[51:48] obvious to everyone on planet Earth as soon as they hear it. And if that's

[51:50] soon as they hear it. And if that's where you are in venture, if it's not

[51:52] where you are in venture, if it's not hard to do, if it becomes obvious to the

[51:55] hard to do, if it becomes obvious to the world before you have built um scale,

[51:59] world before you have built um scale, scale is the ultimate advantage, you're

[52:00] scale is the ultimate advantage, you're in trouble. And the great thing Amazon

[52:03] in trouble. And the great thing Amazon had was um you know, I think it was

[52:06] had was um you know, I think it was obvious to a lot of people, but it

[52:08] obvious to a lot of people, but it wasn't obvious to the retail CEOs. And

[52:10] wasn't obvious to the retail CEOs. And Amazon, they were very smart.

[52:13] Amazon, they were very smart. Any e-commerce company that VCs invested

[52:16] Any e-commerce company that VCs invested in, they would destroy. They'd be like,

[52:18] in, they would destroy. They'd be like, "Oh, that's so cute. We're gonna we're

[52:20] "Oh, that's so cute. We're gonna we're gonna take our margins of that to

[52:22] gonna take our margins of that to negative 10,000%."

[52:24] negative 10,000%." And that's what like like the guys at

[52:26] And that's what like like the guys at Wayfair, they did something hard and

[52:27] Wayfair, they did something hard and Amazon tried to kill them and they

[52:28] Amazon tried to kill them and they failed. Those are like tough

[52:30] failed. Those are like tough operationally

[52:31] operationally like really competent CEOs. For me in

[52:34] like really competent CEOs. For me in venture, I always look, is this going to

[52:36] venture, I always look, is this going to be obvious to the world before this

[52:39] be obvious to the world before this company could build scale

[52:42] company could build scale or is this both not obvious, different,

[52:45] or is this both not obvious, different, and really hard to do? I think a lot of

[52:48] and really hard to do? I think a lot of founders are really struggling with this

[52:51] founders are really struggling with this >> in AI like I think people are

[52:56] >> in AI like I think people are becoming worried you know today in that

[52:59] becoming worried you know today in that in Jensen's five layer cake of AI

[53:02] in Jensen's five layer cake of AI and the profits they're acrewing to

[53:04] and the profits they're acrewing to energy they're crewing to data centers

[53:06] energy they're crewing to data centers they're crewing to chips they're

[53:07] they're crewing to chips they're acrewing to models they're not really

[53:09] acrewing to models they're not really acrewing to the applications cursor and

[53:12] acrewing to the applications cursor and cognition you know got to a scale you

[53:15] cognition you know got to a scale you know they focused on coding

[53:17] know they focused on coding you know 18 months ago the people were

[53:19] you know 18 months ago the people were focusing on coding. OpenAI was doing

[53:20] focusing on coding. OpenAI was doing everything under the sun. The people

[53:22] everything under the sun. The people focused on coding were cursor cognition

[53:24] focused on coding were cursor cognition and um anthropic and it was really right

[53:27] and um anthropic and it was really right to focus on code. Um I'm MSAD the

[53:30] to focus on code. Um I'm MSAD the founder of Replet tweeted something that

[53:32] founder of Replet tweeted something that I thought was so smart just it was

[53:34] I thought was so smart just it was something like you know bitter lesson

[53:36] something like you know bitter lesson adjacent is the fact that coding might

[53:39] adjacent is the fact that coding might be the shortest path to ASI and useful

[53:41] be the shortest path to ASI and useful AI because if you're really good at

[53:43] AI because if you're really good at coding you can write yourself code to do

[53:45] coding you can write yourself code to do anything. So I think it was really smart

[53:46] anything. So I think it was really smart of those companies to focus intensely on

[53:48] of those companies to focus intensely on coding and I think they all probably got

[53:51] coding and I think they all probably got to a scale where they they have a place.

[53:54] to a scale where they they have a place. I think cognition is doing something

[53:55] I think cognition is doing something really really different but I think a

[53:57] really really different but I think a lot of founders are really struggling

[53:59] lot of founders are really struggling man they're really struggling

[54:03] man they're really struggling and you know I think they're trying to

[54:04] and you know I think they're trying to get confidence that in nichier areas

[54:08] get confidence that in nichier areas >> that they can get to them and get like a

[54:12] >> that they can get to them and get like a you know a data moat

[54:14] you know a data moat >> before the model companies get to that

[54:16] >> before the model companies get to that niche or that it's a small enough niche

[54:18] niche or that it's a small enough niche that the model companies won't do it

[54:20] that the model companies won't do it themselves but it can still produce a

[54:22] themselves but it can still produce a venture outcome. Is this related to what

[54:23] venture outcome. Is this related to what you would call like the token path? I

[54:24] you would call like the token path? I know you've used that phrase with me

[54:26] know you've used that phrase with me before.

[54:26] before. >> Yeah, I he comes from a guy um at

[54:28] >> Yeah, I he comes from a guy um at alttimeter, Jamon Ball. He just said if

[54:30] alttimeter, Jamon Ball. He just said if you're a software company or an AI

[54:32] you're a software company or an AI company of any kind, you have to be in

[54:33] company of any kind, you have to be in the token path. So, data bricks that's

[54:35] the token path. So, data bricks that's in the token path. Comparable companies

[54:37] in the token path. Comparable companies are in the token path. If you're not in

[54:39] are in the token path. If you're not in the token path

[54:41] the token path and you're not in some really niche

[54:45] and you're not in some really niche thing, life may be hard. And even for

[54:49] thing, life may be hard. And even for these vertical niches, I think if you

[54:51] these vertical niches, I think if you talk to the people at the model

[54:54] talk to the people at the model companies, they're even skeptical of

[54:56] companies, they're even skeptical of some of these because all of the data

[54:59] some of these because all of the data that's, you know, being generated in

[55:01] that's, you know, being generated in these niches come from humans. But then

[55:03] these niches come from humans. But then you're betting that you're able to use

[55:05] you're betting that you're able to use that proprietary data in this narrow