# Why the AI Boom Is Just Getting Started

https://www.youtube.com/watch?v=DZt1DDmMNGk

[00:00] When you get the right part of the
[00:01] S-curve, you get exponential unit
[00:03] growth. If you have a very strong
[00:05] business model, your earnings don't grow
[00:07] linearly, they grow exponentially. You
[00:09] know, the world doesn't think
[00:11] exponentially. Very few people believe
[00:15] you can accurately predict 2 3 4 years
[00:18] out. But if you follow and understand
[00:20] the Scurve and you you know the moes and
[00:23] you know how to model, you really can uh
[00:26] predict these these great things. the
[00:28] enterprise AI or enterprise application
[00:31] AI market is less than 1% penetrated and
[00:35] we've never seen, you know, we talk
[00:37] about S-curves, we call this an L curve,
[00:38] just straight up.
[00:53] Alex, you were saying that your highest
[00:54] conviction position is anthropic right
[00:56] now. Can you tell the story of
[00:58] discovering it, making the investment,
[01:00] using this anecdote as an excuse to talk
[01:02] about all the things that I think you
[01:04] and I are mutually interested right now,
[01:06] investors like you, investing in private
[01:07] markets, anthropic, the business, AI,
[01:09] everything. It's a great great way to
[01:10] zoom in. Why is it your highest
[01:12] conviction? And how did you get started?
[01:13] >> Yeah. Well, when when the gun went off
[01:16] with OpenAI chat GPT in November 2022,
[01:21] we immediately took the firm and did a
[01:23] massive deep dive with our 10 person
[01:26] team. And we anytime you have a new
[01:29] compute paradigm, there's a new stack
[01:32] and on the and and that creates new
[01:34] winners and losers on the old stack. And
[01:36] in this stack, you know, it's now Jensen
[01:39] talks a lot about it, but it's power at
[01:40] the bottom, chips at the bottom, the
[01:43] clouds, and then the foundational
[01:45] models, and then the applications on
[01:48] top. And at that time, this was 2023
[01:52] early, we said, we want to be in the
[01:55] chips and the infrastructure first. And
[01:58] not only do they get the uh demand
[02:01] first, but we know who the winners are.
[02:03] And no matter who wins above, which we
[02:06] weren't sure at the time, we know we're
[02:08] going to need tremendous amounts of
[02:10] compute. And we did a deep dive into
[02:11] that, which we can talk about later, but
[02:14] over the next 2 or 3 years, we started
[02:17] to get more clarity on how the
[02:20] foundational model, the layer would
[02:23] evolve. And at the time, two or three
[02:25] years ago, there were 60 different
[02:28] companies going after it. OpenAI was
[02:30] kind of in the lead. And we did a
[02:32] webinar in April 2023. We said, look,
[02:36] this might be a winner take all. It
[02:38] might be a total commodity because
[02:40] there's open-source players. It might be
[02:42] a race to zero or it might be an
[02:45] oligopoly where there's three or four
[02:48] leading players. And what we saw over
[02:51] the following, you know, 3 years was
[02:54] that almost all the startups
[02:57] fell away and died. And then some of the
[03:00] largest companies in the world including
[03:02] Amazon and and Meta. Amazon really never
[03:05] really showed up. We'll see what happens
[03:07] with Meta, but they were they came in
[03:10] strong and then basically their effort
[03:13] faltered and they had to do a total
[03:14] reboot. In the meantime, Anthropic kind
[03:18] of was this dark horse candidate, the
[03:20] startup and um they focused uh really
[03:26] purely on the enterprise and OpenAI had
[03:30] kind of won the consumer and then Gemini
[03:32] can never be counted out. We we love
[03:34] Google as well. It's one of our largest
[03:37] positions. So it really started to look
[03:40] like a three-horse race and somewhat of
[03:42] an oligopoly
[03:44] very similar to how the uh cloud market
[03:49] evolved where three companies underpin
[03:53] the entire SAS cloud world and and have
[03:57] really excellent businesses and then we
[04:00] also were aware of the open- source risk
[04:03] um from China and we started to get
[04:06] comfortable that the quality of the
[04:08] tokens from the leading edge were
[04:11] superior because if you're 80%
[04:15] close to the top of the benchmarks going
[04:18] from 80 to 85 is a huge unlock and the
[04:22] um the open- source guys they don't have
[04:24] as much compute so they can come close
[04:26] to the leading edge but they can't
[04:28] leapfrog it and then they kind of
[04:30] falter. Meanwhile, the scaling laws and
[04:33] other
[04:35] means of improving the models, the
[04:37] feedback loops, etc. Uh we saw that
[04:40] there was a very strong runway and
[04:42] everyone we talked to close to the
[04:44] industry saw that the scaling laws would
[04:46] continue. So we developed this thesis
[04:49] that it would be a three-horse race. And
[04:51] then the big kicker was code. And this
[04:55] is the true unlock of AI. In the first
[04:59] few years, we knew AI would would be
[05:01] big, but we were skeptical. Also, we
[05:04] made large investments because we knew
[05:05] the training was would be there, but we
[05:07] weren't sure how much revenue might come
[05:10] and if it could truly replace labor
[05:12] because if you remember the early
[05:14] versions of the models were good, but it
[05:17] there was a lot of uh some negative
[05:19] feedback from corporates and could they
[05:22] be truly agentic? We realized in 2025,
[05:27] the first cloud code and and the coding
[05:30] tools really began to explode and you
[05:34] saw
[05:36] the first gen was like Microsoft C-Pilot
[05:38] which is like $20 a month and then and
[05:42] then it started and that could sort of
[05:44] improve your grammar of coding, maybe
[05:47] find a bug, maybe make a block of code
[05:49] like a paragraph and then Anthropic came
[05:53] out sometime in in the middle of the
[05:54] year and it could do so much more. Um,
[05:58] and it started to get to this point
[06:00] where it could run agentically and we
[06:02] kind of saw that happening and the
[06:03] coding market just exploded and then we
[06:06] started hearing that people who could
[06:09] use it unfettered. We heard we heard
[06:11] that you know even within Anthropic at
[06:14] that time people were spending $100 a
[06:17] day on tokens which if you do the math
[06:19] comes out to 20 or $30,000 a year. And
[06:22] if you think about how many coders there
[06:24] are in the world, 20 million, you've got
[06:25] a half a trillion dollar market just
[06:28] from coding alone. And mind you, that
[06:31] was on 7 8 9 month old technology. We
[06:34] could see just on the coding market
[06:36] alone that Anthropic had a tremendous
[06:40] opportunity ahead of it. So I think at
[06:43] the time, this is pretty funny, we wrote
[06:44] in our letter, you know, we made the
[06:47] investment um at the 180 valuation. And
[06:50] we said, and I think they were
[06:54] hoping to get to a nine billion
[06:57] >> one to nine. Yeah. And and then the
[06:58] numbers were like nothing we'd ever seen
[07:00] before. 100 to a billion on the way to
[07:04] 9. But when we did it in August of 2025,
[07:07] we nobody had any idea what 2026
[07:12] could be. the the the second big unlock
[07:14] lately which is that you know claude
[07:17] code has gone to almost completely
[07:20] agentic um where you had Andre Carpathy
[07:24] and Lionus Torvalds last year saying two
[07:27] of the smartest people in coding and
[07:30] they completely flip-fpped and Karpathy
[07:33] said you know last year's code tools
[07:36] could write 20%
[07:38] and 80% would be handwritten that
[07:41] flipped when the the latest model came
[07:43] out and now he hasn't written a line of
[07:45] code not except in English and not to
[07:48] mention the pure unlocked that we're
[07:51] going to get for the people that never
[07:53] knew how to code. So just coding alone
[07:57] has completely taken off. Anthropic has
[08:00] been able to stay ahead in coding. And
[08:04] so one difference between the cloud,
[08:08] GCP, AWS, and the AI companies is the
[08:13] cloud's generally it's commodity.
[08:15] They're they're selling you servers and
[08:17] storage. You know, they have a lot of
[08:19] software on top and there is stickiness
[08:21] to it. But in the AI models, everyone
[08:24] thought it would be pure commodity. But
[08:26] there's tremendous differentiation with
[08:29] within. There's different training
[08:31] methods and different skills that
[08:33] they're good at. And a lot of people
[08:34] have routers that switch in between,
[08:37] which sort of makes it sound like
[08:39] they're commodity, but anthropic,
[08:40] they're very good for anything that has
[08:42] to do with private equity and finance.
[08:45] Google's very good for ingesting PDF.
[08:48] And so there's a lot of like
[08:49] differentiation critical IP, which is a
[08:53] great competitive advantage.
[08:55] and companies many companies have come
[08:58] after the coding franchise and Anthropic
[09:01] has been able to keep ahead.
[09:04] The other thing that's good about the
[09:06] foundational models and anthropic is
[09:08] it's not just the API or the model.
[09:11] They're building a whole monopoly or
[09:14] whole ecosystem of products around the
[09:17] API. So we've got the SDK claude for
[09:20] co-work uh orchestration layer and and
[09:24] all the tools and and they call it sort
[09:26] of a harness which is the software
[09:30] around the API that gets the most out of
[09:33] the model. This was one of the things we
[09:35] saw with AWS really early on in 2013 was
[09:38] oh people thought it was a commodity
[09:40] server up in a warehouse big deal and
[09:43] what they they saw this was a new way of
[09:46] do doing computing. So they had they
[09:48] invented all these products that they
[09:51] could see before everybody else that
[09:53] slowly built lock in. The other way we
[09:56] think about this is where are we on this
[09:58] scurve and we have this infrastructure
[10:02] layer scurve which we think is somewhat
[10:05] like 10% penetrated. And by the way we
[10:08] think it's still uh one of the best ways
[10:10] to play AI and we'll talk about how that
[10:12] feeds back through. Um but if you think
[10:16] about it, um even though you know 200 or
[10:20] I don't know how many 800 million people
[10:22] are using AI, they're just using AI 1.0
[10:25] which is like a a search engine on
[10:27] steroids. But now with these new
[10:29] primitives where you have claw on your
[10:31] computer linking it in, then you build
[10:33] skills. Companies are going to build
[10:35] people and companies are going to start
[10:36] building skills and then they're going
[10:38] to build true AI bots and then big
[10:41] corporations are going to build much
[10:43] larger but where are we in terms of the
[10:45] amount of people doing that? I mean
[10:47] Sunder said it's 10 bips of the uh
[10:51] knowledge workers the world. So
[10:53] Anthropic has something like 14 or 15
[10:56] million DAUs. Probably a small portion
[10:58] of those are truly doing AI the way you
[11:01] can do it. So that 10 bips, it's classic
[11:05] Scurve where these are the tinkerers and
[11:08] then it's going to go to the early
[11:09] adopters, then it's going to go to the
[11:10] early mainstream. But you're going to go
[11:12] from 10 bips to one to two or 3% to 5%
[11:17] to 15% in the next four years. And kind
[11:20] of a light switch this year went off in
[11:23] the enterprise where everybody realizes
[11:26] they need to do this now and do it fast.
[11:29] It's still
[11:30] >> like internet 1.0 I know when it's like
[11:32] you knew you needed a website in 1998
[11:36] but it's like hard to build that website
[11:39] but this is coming together fast and so
[11:42] you know we think the I don't the
[11:45] enterprise AI or enterprise application
[11:48] AI market is is like less than 1%
[11:51] penetrated and we've never seen you know
[11:54] we talk about S-curves we call this an L
[11:56] curve just straight up and then we'll
[12:00] take this to the infrastructure
[12:02] which is even we're at 10 basis points
[12:05] of people really using AI and we're
[12:08] already sold out of all the there's not
[12:10] enough compute in the world. So
[12:12] Anthropic has half of what they need
[12:13] right now and that's before this huge
[12:17] takeup. So Mark Andre said in the next
[12:20] four years one thing he's sure of is
[12:22] there's not going to be enough compute.
[12:24] >> Most software companies try to maximize
[12:26] your time on their app to juice
[12:27] engagement. Ramp does the exact
[12:29] opposite. RAMP understands that no one
[12:31] wants to spend hours filing expense
[12:33] reports, reviewing expense reports, and
[12:35] checking for policy violations. So, they
[12:37] built their tools to give that time back
[12:39] using AI to automate 85% of expense
[12:41] reviews with 99% accuracy. And since
[12:44] RAMP saves companies 5%, it's no wonder
[12:46] that Shopify runs on RAM, Stripe runs on
[12:49] RAM, and my business does too. To see
[12:51] what happens when you eliminate the busy
[12:52] work, check out ramp.com/invest.
[12:56] OpenAI, Cursor, Enthropic, Perplexity,
[12:58] and Verscell all have something in
[13:00] common. They all use work OS. And here's
[13:02] why. To achieve enterprise adoption at
[13:04] scale, you have to deliver on core
[13:06] capabilities like SSO, skim, arbback,
[13:09] and audit logs. That's where work OS
[13:11] comes in. Instead of spending months
[13:12] building these mission critical
[13:13] capabilities yourself, you can just use
[13:15] work OS APIs to gain all of them on day
[13:18] zero. That's why so many of the top AI
[13:20] teams you hear about already run on work
[13:22] OS. Work OS is the fastest way to become
[13:25] enterprise ready and stay focused on
[13:26] what matters most, your product. Visit
[13:28] works.com to get started.
[13:31] Every investor should know about Rogo
[13:33] because Rogo Aai's platform is not just
[13:35] another generic chatbot. Instead, it was
[13:38] designed to support how Wall Street
[13:39] bankers and investors actually work.
[13:41] From sourcing, diligence, and modeling
[13:43] to turning analysis into deliverables.
[13:45] For me, three key things differentiate
[13:46] Robo. First, it connects directly to
[13:49] your systems so it can work with your
[13:50] actual data. Second, it understands your
[13:52] workflows, how work really happens
[13:54] across a deal or an investment. And
[13:56] third, it runs end to end and produces
[13:58] real outputs the way the best people do.
[14:00] Auditable spreadsheets, investment
[14:01] memos, diligence materials, and slide
[14:03] decks that match your standards. This
[14:05] all comes from the fact that Rogo is
[14:06] built by finance professionals for
[14:08] finance professionals. And it's already
[14:10] being adopted by some of the most
[14:11] demanding institutions in the world. To
[14:13] learn more, visit rogo.ai/invest.
[14:16] I'm so curious when an investor like you
[14:18] who historically was a public markets
[14:20] investor, you could hit buy and buy
[14:22] whatever you want, is now operating in
[14:24] lots of the most important private
[14:26] market companies. We can talk about
[14:27] Stripe or Data Bricks or OpenAI or
[14:29] Anthropic. How do you get the positions
[14:32] at the size that you want coming from
[14:35] the legacy of been being able to just
[14:37] buy? How much of it is um creativity
[14:40] just directly with the company? If it is
[14:42] directly with the company, so they have
[14:43] it's a double opt-in. they have to
[14:45] decide to let you in too. How do you do
[14:47] that? Like what what have you learned
[14:48] about getting the allocation you want or
[14:51] the amount of equity you want in a
[14:53] private company given that you know that
[14:55] wasn't your original background?
[14:57] >> In that case, you know, we we got to
[14:59] know the company. One of our analysts
[15:01] knew people in in the finance group
[15:05] there and we actually we had a look at
[15:08] the at the $60 billion round and we we
[15:10] didn't do it. we didn't know the company
[15:12] as well and we um the gross margins were
[15:16] negative and and and frankly we hadn't
[15:19] seen coding explode the way it had and
[15:22] and one thing about public markets is
[15:24] you get to know companies
[15:26] over a long period of time and you can
[15:28] kind of invest on your own schedule. I
[15:30] got a chance to spend some time with
[15:32] Daario. I obviously listen to on
[15:35] podcasts and it I started to realize
[15:37] these guys their management team is
[15:39] excellent. the focus, the dedication,
[15:41] they had almost no turnover, the quality
[15:43] of code, and then the business plan was
[15:45] really starting to play out. And uh it's
[15:48] one thing to grow from, you know, 100 to
[15:51] a billion, but it's another to do nine.
[15:54] And then so we reached out to the
[15:56] company as much as we could. They took a
[15:58] meeting with us. We did a 90page
[16:01] PowerPoint deck where we used Claude
[16:03] Code to scour the internet for all the
[16:06] feedback we could about the coding
[16:08] market. and their and what their
[16:10] products were good at, where they might
[16:12] need to improve and we also did our
[16:14] whole overview of what the coding market
[16:16] would be. They welcomed us into this
[16:19] round and then we stayed close with the
[16:21] CFO and uh it's it's been great to build
[16:24] a relationship with them and I think we
[16:26] p punched above our weight in terms of
[16:29] the allocation. So that one was a total
[16:31] home run. In the rest of the the world,
[16:34] we are in this period where the unicorn
[16:36] market is bigger than most stock markets
[16:39] in Europe, maybe even combined. It's
[16:41] definitely bigger than Germany. It's
[16:43] definitely bigger than the UK. And we
[16:47] even before we invested in privates, the
[16:49] first one was 2020. We meet with these,
[16:51] we have to know these companies and you
[16:52] really have to know them now because
[16:55] sometimes they're the biggest companies
[16:57] in the space and and have huge impact.
[16:59] So we, you know, we do two to 3,000
[17:02] face-to-face meetings with management
[17:03] teams a year and about 10 or 15% of
[17:06] those are with privates and then we kind
[17:09] of focus in on the companies that we
[17:12] really want to learn about and find ways
[17:15] to meet with them and uh get involved in
[17:18] their rounds. And our first one was
[17:21] Stripe.
[17:23] And we had a large investment at the
[17:25] time. This is 20 2018 2017 1819 we and
[17:30] 2020 we own Audion which is a fantastic
[17:33] payments company and they're a next-gen
[17:35] cloud payments company taking from world
[17:37] pay and you know the cloud the cloud
[17:40] modern payments was 5% of total you know
[17:43] $80 trillion market or what have you but
[17:46] you can't invest in aud unless you know
[17:48] Stripe like the back of your hand so we
[17:50] did tremendous amounts of due diligence
[17:52] talked to 200 customers in Audon but
[17:54] when we asked about audio and we asked
[17:56] about Stripe and we realized this is
[17:58] Coke and Pepsi and um we said we got to
[18:01] find a way to invest and I finally got
[18:03] to meet the Coulson brothers in 2019 and
[18:06] so that was our first one. We weren't
[18:07] really known for privates. I've got a
[18:10] friend um who's who has a involved with
[18:14] a a venture firm that has tremendous
[18:16] amounts and I talked to him about it and
[18:18] I said let me know if you ever want to
[18:20] sell some and then I get a call from him
[18:23] during co in April
[18:26] of 2020. We knew a lot about Stripe. We
[18:30] didn't have the full financials, but we
[18:31] knew enough that at that valuation, I
[18:34] think it was 35 billion. We knew they
[18:37] had they disclosed we had over half a
[18:39] trillion of TPV. And we knew that
[18:42] Audience's take rate was 25 or 30 bips
[18:44] and we knew Stripes was 40 or 50. And we
[18:48] knew how many employees they had. So, we
[18:50] could kind of get at the profitability.
[18:52] It turned out the take rate was higher.
[18:54] It turned out they were being modest
[18:56] about their TPV. It was much higher than
[18:58] the 550. It was closer to the one 1
[19:00] trillion. And you know, we underwrote
[19:04] the thing under our assumptions and it
[19:05] was much better. And then we were able
[19:07] to upsize that from the seller to a $100
[19:10] million block. Sometime they like it
[19:13] that you know the VCs are going to own
[19:15] and then most of them are going to sell.
[19:17] They like it that we'll own and own in
[19:20] the public market which we did with new
[19:22] bank uh as well. all owned it for a long
[19:24] period of time in the public market as
[19:26] well.
[19:26] >> Maybe now's the right time to lay out
[19:28] everything you've ever learned about
[19:30] S-curves. Obviously, your firm is sort
[19:32] of predicated on this idea of technology
[19:35] adoption life cycles
[19:37] >> and investing in companies at the right
[19:39] time amidst a certain platform change or
[19:42] S-curve change.
[19:43] >> And I think everyone knows the basic
[19:45] idea of an S-curve and and the sort of
[19:47] uh the stages you mentioned, tinkerers
[19:49] and and early adopters and early
[19:51] majority. But I'd love you to go into
[19:53] the the super deep detail of what you've
[19:55] learned since this is the lens through
[19:57] which you've viewed markets and stocks
[19:59] for a long time. Bring us into like the
[20:01] nitty-gritty fine grading nuance detail
[20:03] of why S-curves can be so useful for
[20:05] investing. We have an investment
[20:07] framework. It's
[20:08] >> S-curve
[20:10] and we'll dive into each one.
[20:13] Competitive advantage and then
[20:15] underappreciated earnings power. And
[20:17] when you get the right part of the
[20:18] S-curve, you get exponential unit
[20:20] growth. If you have a very strong
[20:22] business model, which in tech there's so
[20:24] many of those for so many different
[20:26] types of modes, uh your earnings don't
[20:29] grow linearly, they grow exponentially.
[20:31] And that's the last piece. Invest when
[20:34] there's underappreciated long-term
[20:36] earnings power. And very often the
[20:38] earnings can grow from $1 to $10.50 to
[20:42] 20. And it happens way more than you
[20:44] think. And it allows you to buy some of
[20:47] the best companies in the world for
[20:49] extremely low pees. When we were buying
[20:52] Nvidia in 2023,
[20:55] we were paying four times earnings. When
[20:57] we bought Tesla in 2019 for the car
[21:00] scurb, we were paying five times
[21:03] earnings. When we were owning Apple, we
[21:05] were paying four times earnings. When we
[21:07] bought Amazon for AWS, we we were
[21:09] getting it for free. And you know, the
[21:11] world doesn't think exponentially.
[21:14] and they're so focused on the next year,
[21:16] the next quarter. Very few people
[21:19] believe you can accurately predict two,
[21:22] three, four years out. But if you follow
[21:24] and understand the S-curve and you you
[21:27] know the modes and you know how to
[21:28] model, you really can uh predict these
[21:32] these great things. So let's go to the
[21:34] scurve. So the S-curve is crucial
[21:36] because every technology follows this
[21:39] pattern where it comes out,
[21:43] you know, the I the smartphones were out
[21:46] 10 years before the iPhone. The internet
[21:48] was out 20 years before Netscape. AI has
[21:52] been out hidden inside of these
[21:54] companies, but it wasn't until Chachi PT
[21:57] took it public uh and ignited what it
[22:01] was. So electric vehicles, Tesla went
[22:04] public 15 years before 2019 when it went
[22:06] vertical
[22:08] it because there were so many barriers
[22:10] to adoption. The first smartphones, you
[22:12] know, they were clunky, they didn't have
[22:14] touchscreen, not Apple, there wasn't a
[22:17] wireless data system. And then uh and
[22:20] they were too expensive. They were $500
[22:21] or $600. Steve Jobs got the price to
[22:24] 200.
[22:26] There was AT&T had a 3G network. It was
[22:28] touchscreen. It was so easy your
[22:30] grandmother could do it. So Annie built
[22:33] an ecosystem and made it simple. So all
[22:35] the barriers to adoption were eliminated
[22:38] and then you rocket when those barriers
[22:40] are removed. That's the tornado of
[22:43] demand that everybody in the world knows
[22:46] they need this right away. And so that's
[22:48] the flip that happens. It happened with
[22:50] electric vehicles. The price was too
[22:53] high. Elon got the price to 40,000.
[22:55] Range anxiety was there. he got the the
[22:58] range to 300 miles. The supply chain was
[23:01] was finally in place so he could churn
[23:03] out millions of these things. So that
[23:06] triggers the inflection. Now the other
[23:08] nuance, it's not just, oh, it's taken
[23:10] off now. It's how tall, how big is this
[23:13] S-curve, how tall it is, so you know
[23:15] when to sell, how long to hold on, cuz
[23:19] we're underwriting out 2 or 3 years. We
[23:20] have to know what the growth looks like
[23:22] thereafter. And these S-curves can be
[23:25] dynamic. So when Amazon had AWS and it
[23:29] was a hidden line item inside of Amazon
[23:33] covered by retail internet analysts, not
[23:37] hardware chip, it was a new business
[23:40] model, what have you. But we realized
[23:42] the TAM for this, it was the largest TAM
[23:45] in enterprise IT ever because previously
[23:47] the TAM was routers, memory, storage,
[23:50] Dell, EMC, but they were doing it all.
[23:54] And so we figured out, you want to know
[23:57] how tall the S-curve is. So we figured
[23:58] out they were addressing 600 billion of
[24:02] IT systems directly addressing that. And
[24:06] then we said it's probably going to be
[24:08] 50% deflationary. Therefore, we're 1 or
[24:11] 2% penetrated. But then over time, we
[24:14] realized it it actually wasn't
[24:16] deflationary. If you talk to anybody
[24:18] now, they say if you build it yourself,
[24:19] it's about the same price. So that means
[24:21] the TAM was so much bigger. So there's
[24:25] mega S-curves and there's subass curves.
[24:28] You know, we've been lucky that we've
[24:29] had, you know, internet 1.0,
[24:32] uh, mobile, cloud, e-commerce,
[24:36] and now AI, which we can confidently say
[24:40] is the biggest and all these things
[24:41] build upon one another. So, you know,
[24:44] with with the electric vehicle S-curve,
[24:47] you you you have to pay attention too
[24:49] because, you know, at the time we we
[24:51] thought probably maybe 40 to 50% of the
[24:54] cars would go electric, but it did hit a
[24:55] big wall at 10 or 15%.
[24:58] Usually the S-curves go kind of all the
[25:02] way. Um, but in this case, for a variety
[25:04] of reasons, it didn't. So, you have to
[25:06] adjust and you have to stay on top of
[25:08] it. And generally you want to um when
[25:12] something gets to sort of 30 40%
[25:15] penetrated then you stop having
[25:17] exponential growth which means the sell
[25:20] side catches up and there's no longer
[25:22] big beats
[25:23] >> and is that when you sell typically
[25:25] >> generally we like we like the high
[25:27] growth and it was a mistake with Apple
[25:29] because in the first five or six years
[25:31] of Apple um it was awesome. I mean it
[25:34] was our largest position. and it would
[25:35] go up 50 70% a year um except for '08
[25:39] and then we sold in 20 2012 when it got
[25:42] to sort of 50% of the US had a
[25:45] smartphone and with Apple you know they
[25:48] maintained their leadership position it
[25:51] had a couple years of underperformance
[25:53] and then the multiple got low and they
[25:55] added several ancillary things and then
[25:59] they al also got to play in the uh the
[26:02] application because they get 30% % of
[26:04] the app. So they were able to compound
[26:07] very nicely, say 20%, but the the big
[26:09] years were in the 50, you know, the the
[26:12] 0 to 50% part of the curve.
[26:14] >> I'm so fascinated by this uh you know,
[26:17] sometimes decade plus long flatline at
[26:19] the beginning of one of these curves,
[26:21] which makes me wonder what you've
[26:23] learned about the right moment to buy or
[26:25] even start paying attention before you
[26:27] buy.
[26:28] >> How do you measure that? Is it always
[26:30] different? What are the pitfalls that
[26:32] you've fallen into? How do you know when
[26:34] when to we talked about when to sell,
[26:35] but how do you know kind of when to
[26:37] start thinking about buying in one of
[26:38] these things?
[26:39] >> Yeah. And you know, Andy Grove says sort
[26:42] of when you have strategic inflection
[26:43] points, you can't trust the data. And
[26:47] and strategic inflection points are
[26:49] about intuition, anecdotal evidence. I
[26:53] love this book called The Towel Jones
[26:55] Averages, a guide to whole investing,
[26:58] which is rightrain and leftrain. And the
[26:59] best investors have the right the
[27:01] creative side where they it's visual.
[27:03] It's connecting the dots. Um you know we
[27:06] invested in the mobile video game
[27:09] S-curve for so long. Mobile video games
[27:11] were just the screens were small on the
[27:14] phones and the processing power wasn't
[27:16] good. So you had all these uh casual
[27:18] games. But then I was in China and I saw
[27:20] this little 12-year-old boy with a huge
[27:23] phone and he was like playing a awesome
[27:25] video game. I'm like oh my god it's now
[27:27] coming to the phone. So, it's visual.
[27:30] Um, enterprise is hard cuz you can't see
[27:32] it. We go to the Gartner IT Symposium.
[27:35] 30,000
[27:37] American CIOS go there and like we saw
[27:40] this happen with Splunk where that used
[27:43] to be an amazing database company and
[27:46] like their their room where they were
[27:49] explaining was like standing room only
[27:51] or we saw that with VMware you know I'm
[27:54] talking like 30 years ago where they
[27:56] virtualized the server and like there
[27:57] was standing room only and you could
[27:59] just see the corporate demand just
[28:01] beginning and with AWS
[28:04] We went there and the grand ballroom was
[28:08] completely packed and that was at nine
[28:11] o'clock and at 10 o'clock the grand
[28:13] ballroom was completely packed 11:00. So
[28:16] you could you could actually see the
[28:18] demand exploding before it happened. So
[28:22] um we look for for all kinds of clues
[28:24] and there's a whole pattern recognition
[28:27] that happens. And by the way it's okay
[28:30] to be late. It's okay to miss the first
[28:32] one, two, three years in a lot of cases
[28:35] because if the top of the S-curve is
[28:38] half a trillion, um the growth can go on
[28:42] for a long time. So, you don't always
[28:43] have to be right there. It's okay to
[28:46] miss the first 100%. Peter Lynch, I
[28:49] started at Fidelity and he loved to
[28:51] mentor the young kids. So, I got some
[28:52] time with him. He said, "Wite out the
[28:55] chart.
[28:56] It's all about the future." Um, and so
[28:59] it's okay to miss, but but what helps
[29:01] about the S-curve is is sort of how long
[29:03] it goes for. Then there's sort of the
[29:05] the slope of the Scurve, which is
[29:07] important. And a lot of people think cuz
[29:10] we're in a modern world, everything's so
[29:12] fast, but there's a lot of factors that
[29:15] determine the pace of the adoption. And
[29:18] we we um commissioned this gentleman,
[29:22] Horus Du used to work with Clayton
[29:24] Christensen, to go look in history. And
[29:27] we have the big S-curves on our wall
[29:28] over the last 100 years. And the radio
[29:32] Scurve is one of the fastest ever. It
[29:35] took 7 years to reach like 100%
[29:38] penetration. But the dishwasher Scurve
[29:41] is like that because it needs to be
[29:43] plugged into the back end.
[29:46] >> What are some Yeah. What else did you
[29:47] learn? That's fascinating. What else did
[29:48] you learn?
[29:49] >> So like the B2B stuff can take a long
[29:52] time because it needs to be plugged into
[29:54] the existing systems. It's like it's got
[29:56] to be put
[29:56] >> the dishwasher
[29:57] >> inside the house and then um and
[30:00] consumers generally tend to go a lot
[30:03] faster. Um
[30:05] >> I love that the radio and the
[30:06] dishwasher, the two models for adoption.
[30:08] >> Yeah. And and I I covered internet of
[30:10] fidelity. I c, you know, my first stock
[30:13] was Amazon. That's a whole other story
[30:15] which is a lot of fun. But I also did
[30:17] B2B internet and you know there was a
[30:20] whole huge bullcase on that. But the
[30:23] basically the underlying infrastructure
[30:25] wasn't in place for B2B to happen.
[30:28] Ultimately happened 20 years later with
[30:30] SAS. And so that is a risk with AI in
[30:34] that you know these big companies are
[30:38] very security conscious. Uh they're can
[30:41] be slow to move. There's a lot of
[30:43] cultural issues with a with AI where you
[30:46] know you really need a few evangelists
[30:50] to push it through and the top
[30:53] management needs to push it through but
[30:54] then the IT's saying this is this is
[30:57] risky and that happened with cloud too
[30:59] that was one of the big things with
[31:00] cloud where it was too it was everybody
[31:03] was afraid it's unsecure to have your
[31:05] data in the cloud and then we saw the
[31:07] CIA do it and we saw Capital One and we
[31:10] talked to the Capital One CIO we that
[31:12] it's more secure in the cloud and then
[31:14] it really started to take off. But but
[31:18] those takeoffs
[31:20] maybe because SAS is like the dishwasher
[31:22] and because cloud is like the dish it's
[31:24] got to be plugged in
[31:27] it it meant that yeah it was growing but
[31:29] it was sort of a 30 to 40 maybe a 50%
[31:32] growth rate but what's amazing about AI
[31:34] is you just at least with consumers or
[31:37] even business you just open up
[31:40] >> the browser and it's there
[31:42] >> and so that's why we're getting this
[31:43] straight up
[31:44] >> and I think there's enough runway in the
[31:46] me in the near term going from 10 bits
[31:49] of people really using it to two to five
[31:52] or whatever which is going to cause it
[31:54] to keep on going straight up. So this we
[31:56] this we call it a backwards L curve. Um
[32:00] so it's really pretty exciting.
[32:02] >> What have you learned about uh when the
[32:04] group that ends up being the leaders
[32:06] separates itself from one of these
[32:08] competitive packs? So you're talking
[32:10] there mostly about overall growth of the
[32:13] S-curve and demand. There's always
[32:16] multiple players fighting for it. You
[32:18] know, you've invested, it seems like you
[32:19] kind of invest after someone has
[32:21] separated themselves from the pack, not
[32:22] try to pick the winners from the pack.
[32:24] Is that is that like roughly?
[32:26] >> Well,
[32:26] >> correct?
[32:27] >> Well, we're definitely So, you look for
[32:28] the S-curve, then we do an exhaustive
[32:32] study of everybody with exposure in that
[32:35] area and try and find the one with a
[32:38] very powerful competitive advantage. And
[32:41] a lot of people didn't like tech. Warren
[32:43] Buffett didn't like tech because he
[32:45] couldn't predict the future too fast.
[32:47] Yeah. And so the Scurve is our map for
[32:49] looking in the future. Now a lot of
[32:51] people were worried about tech because
[32:53] they thought there was so much
[32:54] disruption you could never trust a
[32:56] company to be a longived asset. And what
[33:00] we've found over the years is some of
[33:02] the competitive advantages
[33:04] within the digital world are more
[33:07] powerful, if not equally or more
[33:09] powerful than than in the offline world.
[33:13] You've got the network effect that was
[33:15] so powerful for LinkedIn, Facebook,
[33:17] Alibaba, you name it. Then you can
[33:21] become an industry standard. Oracle and
[33:24] Bloomberg are the industry standard.
[33:26] Oracle, you know, they charge a lot and,
[33:29] you know, there's free versions, there's
[33:31] open- source Oracle, but they had all
[33:34] the database administrators. They they
[33:36] had all the software that was tuned to
[33:38] work with them. So, they they basically
[33:40] had a chokeold on the relational
[33:42] database market forever.
[33:44] um you can get to scale very quickly
[33:47] because these scurves grow and all of a
[33:50] sudden Anthropic is doing 90 30 billion
[33:53] in sales or Amazon you know had so much
[33:56] scale and they got it quickly. So they
[33:58] got a Walmart size scale advantage in 5
[34:02] years versus 40 years for Walmart. So
[34:05] you can have network effects scale you
[34:07] can become industry standard. You can be
[34:11] a platform that people build on top of.
[34:13] You can have critical intellectual
[34:15] property, which was what Qualcomm had.
[34:18] You couldn't make a phone without paying
[34:19] them, or ASML has critical intellectual
[34:23] property. You can't make a chip without
[34:25] their lithography. And I think what's
[34:27] interesting is maybe these AI
[34:29] foundational companies, you know,
[34:31] they've got scale. Oh, you can also have
[34:32] brand. And brand's very important
[34:35] because Google, Amazon, they got to
[34:37] grow. They never had to advertise.
[34:39] Elon's never had to advertise for
[34:41] anything. and cost to acquire versus
[34:43] lifetime. It's the whole business model.
[34:46] And so almost all the companies I
[34:48] mentioned have Apple, they have all of
[34:51] these rolled into one. Um, so we can
[34:55] sometimes we can notice these things
[34:58] before the rest of the world. And one of
[35:01] our high points was we pitched Amazon
[35:03] for AWS at 2013 at the Robin Hood
[35:07] investors conference and we said the
[35:10] bulls have no idea what they're sitting
[35:11] on. Amazon's won the war but before it
[35:14] even started and at that time we said
[35:16] there's Coke and there's no Pepsi. Did
[35:18] turn out there was Pepsi but it was big
[35:20] enough to last. And we could see they
[35:22] had a seven-year lead. So first mover is
[35:24] important. Then they became a whole
[35:27] ecosystem and a platform. Then they got
[35:30] scale. So they were 10 times the size of
[35:32] everybody else. Nobody could invest in
[35:33] the R&D to to catch them. So um but
[35:38] you're right that if you don't have a
[35:40] competitive advantage, you can be in the
[35:42] best S-curve of all time
[35:44] >> and still lose out.
[35:44] >> But if your name was Rim, Palm, Nokia,
[35:47] HC, LG, Motorola, I can go on forever. 0
[35:51] negative negative negative negative. And
[35:53] that's what we saw at the foundational
[35:55] model layer where there's like 50
[35:57] companies trying to do that and they all
[36:00] have fallen away and two or three have
[36:04] emerged at the top and there's a lot of
[36:07] reasons to think they will continue to
[36:09] hold their position.
[36:10] >> So to take Google's a little trickier
[36:12] because they have this other huge
[36:14] massive complex business attached to the
[36:16] Gemini business. But if you take
[36:18] anthropic and open AI as pure plays and
[36:20] you dig through those and you reason
[36:22] through their competitive advantages,
[36:24] why aren't they susceptible to erosion
[36:25] of those things in the fullness of time?
[36:28] >> Yeah. Of all the S-curves we've we've
[36:30] done, AI is by far the most complex and
[36:34] the fastest changing. So it can be we
[36:38] have to keep in mind that there are
[36:40] risks
[36:42] but also the rewards are the highest cuz
[36:46] we're talking about a market in the
[36:48] trillions. You know we we just said
[36:50] cloud you know maybe cloud's 800
[36:52] billion. This might be you know we now
[36:55] think 3 to five but there's higher risk
[36:57] higher reward. But let's just say with
[37:00] anthropic now they have it looks like
[37:03] they have critical intellectual property
[37:06] generally they've been able to maintain
[37:08] their their high market share and code.
[37:10] Number two is uh they've built a a
[37:14] strong brand for enterprise to where go
[37:17] talk to any CIO and they'll just the
[37:19] first thing they'll say is claude.
[37:20] They're going to have escape velocity
[37:22] and scale. And what was scary for OpenAI
[37:25] and Anthropic fighting these big
[37:27] companies like Google was they had these
[37:29] huge cash cows. And to both of the
[37:32] management teams credited Open and
[37:34] Anthropic, they were able to
[37:37] work in these super capital intensive
[37:38] industries and find ways to raise
[37:40] capital. And certainly with Anthropic,
[37:43] with their 10x sales growth, it looks
[37:46] like and their fundraising ability, it
[37:47] looks like they've reached escape
[37:49] velocity. So now they have scale.
[37:52] And the other thing that Anthropic and
[37:55] OpenAI could have is Anthropic now that
[37:59] they're leading in code, they set that
[38:01] code back onto their model and it's this
[38:03] concept of the recursive improvement.
[38:06] And if you look at the pace of their
[38:08] innovation, it's accelerating.
[38:11] >> Um, and so maybe they can have this
[38:13] liftoff stage. you know, Open AI has,
[38:16] you know, they they were focused on so
[38:20] many different other sectors, but
[38:23] they're starting to do better in
[38:25] enterprise and their coding tools good
[38:27] and they're starting to see accelerating
[38:30] growth on that side. And then look, the
[38:33] consumer franchise,
[38:35] it it looks like enterprise right now is
[38:38] much better because you're, you know,
[38:39] you and I, we're willing to pay a lot
[38:41] because it's replacing human beings. you
[38:43] know, consumer, maybe you can get
[38:46] advertising, but maybe they would pay
[38:48] for a a clawbot type assistant if you
[38:51] could make that perfectly well for them.
[38:54] Um, but they have gazillion eyeballs
[38:56] there. But you're right, things do
[38:59] shift, but it usually on the we have
[39:01] this
[39:03] charts that we almost do for all of our
[39:05] pitches. On the internet, the leader
[39:07] goes bigger, faster, and wins. And it's
[39:09] it's it's happened you know most of the
[39:12] time the leader gets it. Shopify becomes
[39:13] the leader. It just keeps on going.
[39:15] Amazon the leader keeps on going. SAS
[39:17] company XYZ just you get the lead. It
[39:20] compounds on internet company compounds
[39:22] on itself. And and another thing is you
[39:24] need to be big. Another is scale. You
[39:26] need the compute and you got to pay for
[39:28] the compute because so there's only so
[39:30] many people that can do that. So that
[39:32] those are some of the modes that we
[39:33] think are now showing up. Now there are
[39:36] some exceptions to that rule. usually
[39:38] with the paradigm shifts AOL and then
[39:41] dialup went to broadband and and they
[39:44] didn't make make the change. You know,
[39:46] Netscape came out early and it wasn't as
[39:49] strong of a business model. But I think
[39:51] if you talk to anyone in the valley or
[39:54] any startups, you know, they'll tell you
[39:56] that they're building on top of these
[39:58] three and the world's a huge place and
[40:01] the economy is a huge place that that
[40:02] they'll be able to differentiate within
[40:04] those. I'm so curious then what you
[40:06] think all of this means for software. Um
[40:08] when I look through your portfolio, I
[40:09] don't see a ton of uh big software
[40:12] companies, enterprise software
[40:13] companies. I I don't know if you once
[40:15] had them and sold them or or how you
[40:17] thought about it, but it's hard to have
[40:19] the experience of building really
[40:21] useful, cool little tools, even if
[40:23] they're still toys, and not have the
[40:25] thought of, wow, you know, like if I
[40:28] spend enough time on this, even if I'm
[40:29] not technical, maybe I could build a,
[40:32] you know, an ERP equivalent replacement
[40:35] or something for my company. There
[40:37] doesn't seem to be a fundamental reason
[40:38] why that's not possible and and then
[40:40] those companies could be in lots of
[40:42] trouble. Seems like everyone has a a
[40:44] strong view on this one way or the
[40:45] other. I'm curious how you've approached
[40:47] those sorts of companies given that you
[40:49] don't seem to own a ton of them.
[40:50] >> We were at certain points maybe 5 years
[40:53] ago, we might have had 40 or 50% of our
[40:55] portfolio in software. And early on in
[40:58] in our April 2023
[41:01] seminar, we said definitely invest in
[41:03] chips first and and we said but at the
[41:06] application layer initially we thought
[41:10] these companies are huge. They have huge
[41:12] sales forces. They can take these AI
[41:14] APIs and and build products and they
[41:17] have the data. This is going to be
[41:18] amazing for software.
[41:21] And pretty quickly we realized their AI
[41:25] products were not very good. They
[41:29] weren't moving the needle. Nobody could
[41:31] charge for them. We basically sold
[41:33] almost all of our software, almost all
[41:36] of our application software. We still
[41:38] have one or two small ones, but entering
[41:42] this year, we were actually net net
[41:44] short. And uh it really helped us in the
[41:48] first quarter. There's so many layers.
[41:50] The old way of software is like using a
[41:53] pen and paper or it's like a horse and
[41:56] buggy. The new way of software is like a
[41:59] jet engine or frankly like the
[42:01] transporter from Star Trek. It's so
[42:06] revolutionary changing that it feels
[42:09] like it has to be disruptive
[42:12] now.
[42:14] even if it's not disruptive now or right
[42:17] away. Uh so the software companies have
[42:20] another problem which is
[42:23] their list on the to-do list or priority
[42:26] list of any CIO has fallen a lot. So
[42:29] even if AI is not going to be
[42:32] disruptive, they're spending it on
[42:35] anthropic tokens because there's faster
[42:38] ROI there. Um, second,
[42:42] um, if they're spending all that money
[42:44] over there, it pushes on the budget, so
[42:47] that hurts them. Third, a lot of
[42:50] software companies were able to raise
[42:52] price every year. Um, and now they're
[42:56] probably nervous about doing that. Then
[43:00] fourth, we'll see what happens with jobs
[43:02] cuz I don't, you know, there's smart
[43:04] people on both sides of that, but we are
[43:07] seeing some companies really gut their
[43:10] jobs or whatever,
[43:12] >> freeze hiring and so that hurts on
[43:13] seats. Maybe in terms of them building
[43:17] their own apps, maybe it just um
[43:21] >> you know, if you want to be optimistic,
[43:22] it's it's taken them a while to do that.
[43:24] We talked about how early the primitives
[43:26] of AR are. So maybe they have just taken
[43:29] a while to get to something they can
[43:31] commercialize, but
[43:34] you know, they might not have the right
[43:35] people. They might not know it's a
[43:37] different selling motion from selling a
[43:40] fixed system versus, you know, if you're
[43:42] installing something that does human
[43:44] work, you got to be right at the side to
[43:46] make sure it's really getting done. So
[43:48] you need the FDE
[43:50] for deployed engineers and they might
[43:53] not have the right people internally to
[43:56] do that. Then of course there's the the
[43:58] risk of you can build it yourself. The
[44:00] bulls will say, well, they're never
[44:02] going to build their own ERP system. And
[44:05] that's probably right. And it is true
[44:06] that technology, old tech is very
[44:09] sticky. Like mobile video games didn't
[44:12] hurt console games and uh the tablet
[44:16] didn't hurt the PC and the smartphone
[44:18] didn't hurt the PC and uh there's a lot
[44:21] of integrations and work that goes into
[44:23] these software. So that's all true and
[44:27] companies do like to buy from they don't
[44:30] like to build themselves that much. So
[44:34] that's all true but you can't imagine a
[44:37] world where in 1 2 3 4 5 years um you
[44:42] could have a brand new AI native company
[44:45] going after each one of these very
[44:48] strong incumbents and it might their
[44:49] data advantage could get obiated. it
[44:51] might be easy to take it out and put the
[44:54] new one in with AI and such. So, what's
[44:57] good if you like so is the valuations
[45:00] are very high and everybody knows
[45:02] they're under pressure. Some people are
[45:04] tempted to buy these, but the AI um
[45:07] coding tools are just getting better and
[45:09] better. So, we'll we'll have to wait and
[45:11] see. And we're we're watching these
[45:12] software companies very closely to see
[45:15] if they're getting any revenue that can
[45:17] change that trajectory. But it's hard
[45:20] because if you're a company like
[45:22] Salesforce, you've got 40 billion in
[45:24] sales
[45:26] and now you you might have 500 of ARR
[45:29] 700 of AR of AI. So you've got this huge
[45:32] base. Now maybe this starts to work but
[45:35] it takes a while. And in software
[45:38] there's the rule of 40 which is your
[45:41] growth rate plus your operating margin.
[45:44] And if you've got a 20% growth rate and
[45:46] 20% that's good. For AI, we have a new
[45:50] kind of rule of 40. We call it well,
[45:52] it's really for chip investing. But if
[45:55] what percent of your sales are AI, say
[45:59] 30%, and what's your market share in
[46:01] that category? Say 30%. You'd be 60.
[46:04] That's a great place to look because
[46:06] you've got exposure and you've got a
[46:08] strong market position. Problem with
[46:10] software is their AI is 1 or 2% at this
[46:13] stage and it's a long way to go. Um, one
[46:17] thing we are picking up though now
[46:19] lately and this is halfbaked, but AI
[46:22] could make some of these software
[46:24] platforms more important because what's
[46:26] the first thing you do with claude? You
[46:27] plug it into Slack. If that can become a
[46:31] key repository, that will make Slack a
[46:34] permanent fixture within the
[46:36] organization. And so maybe these agents,
[46:38] maybe the next wave of AI will be these
[46:40] agents that use tools and they might
[46:43] operate inside of the existing incumbent
[46:46] software tools to use them like a human
[46:48] being would.
[46:49] >> Just to pull in that thread, uh it seems
[46:51] like the commonality of the tools they
[46:52] might use that are the most sticky would
[46:54] be network-based tools. Uh so Slack is a
[46:56] great obviously a great example of the
[46:58] software in Slack itself is I don't know
[47:00] leaves something to be desired. It's not
[47:02] the software is not the special part.
[47:03] It's that everyone is there,
[47:05] >> right? But I'm curious yeah what kinds
[47:07] of things you would want. Is it just
[47:10] network you know the presence of a
[47:11] network effect? Is that the only thing
[47:13] that really matters?
[47:14] >> It's still early in our thinking here
[47:16] but I don't know even even even maybe
[47:19] you know workday or the HR systems or um
[47:25] the big systems of record
[47:27] you know the agents may be running on
[47:30] top of on top of them. CRM is going
[47:33] headless or they're making a headless
[47:35] version and that's sort of the bare case
[47:37] too that you get relegated to just being
[47:39] a database but you know there's a human
[47:43] interface to it then they need to make
[47:45] the AI interface which is no interface
[47:47] it's just them going right into the data
[47:50] and so you know you lose that customer
[47:54] interaction but if if the
[47:57] agents are going right to right to CRM
[47:59] and doing the work inside of CRM M that
[48:03] that will solidify CRM so you won't have
[48:06] to think it's going away.
[48:07] >> Can we talk about chips? You've
[48:08] referenced them a few times.
[48:10] >> Inf infrastructure chips, you know,
[48:12] everything around the data center maybe.
[48:13] I don't know how you conceive of it.
[48:15] >> Why is this so interesting to you? I
[48:17] love the the modified rule of 40 for
[48:19] percentage that's AI and percentage
[48:21] market share in the category. That's an
[48:22] interesting stat.
[48:23] >> What companies shine on that today? What
[48:25] are what are lagards, you know, that are
[48:27] surprising? For the past 40 years,
[48:30] nothing has changed in the data center.
[48:33] Even with cloud, we're basically Intel
[48:36] x86.
[48:38] It became the data center chip sometime
[48:40] in the '9s. And um and compute grew in
[48:44] the cloud era and it grew compute
[48:47] workloads grow you know 25 to 40% every
[48:53] year but Moore's law is improving at
[48:55] that rate.
[48:57] So it didn't require tremendous
[48:59] innovation
[49:01] and there really was almost no growth in
[49:03] hardware for years and years and years
[49:07] and the whole industry basically
[49:10] commoditized every part every chip every
[49:13] part of the server the printed circuit
[49:15] board to the memory to the enclosures to
[49:19] the networking
[49:21] you know there was no innovation
[49:24] you would go from one gig to 10 gig
[49:27] That would take 7 years. And when you do
[49:30] switch in the first year, it does take
[49:32] some innovation to get to 10 gig and
[49:34] would create a little cycle, but then it
[49:35] would commoditize.
[49:38] And now you go to AI and
[49:43] the workloads are growing 10x every year
[49:47] and they're pushing every single aspect
[49:51] of this hardware to the physical limits
[49:53] of what it can do. And so, not only are
[49:57] you creating tremendous unit growth, but
[50:02] the industry, we call it the
[50:03] decommoditization of the hardware
[50:05] industry. And I I met with Shawn Maguire
[50:08] like 3 years ago, and he said, "I wish I
[50:10] could come back and be a a hardware
[50:12] hedge fund because all the companies are
[50:14] public and they all have powerful IP."
[50:17] And Sequoia made some of their best
[50:18] investments back in the hardware day
[50:20] with Apple and Cisco and others. And
[50:23] we're in this renaissance of chips. So
[50:26] not only do you have tremendous unit
[50:30] growth,
[50:31] but you it's requiring tremendous
[50:34] innovation and what that means, you
[50:37] know, at every aspect of the server. And
[50:40] so you know memory which used to be a
[50:43] pure commodity, this high bandwidth
[50:46] memory is stacked 10 chips on top. you
[50:50] know the input outputs are 10x what they
[50:52] were before like took Samsung for years
[50:55] to do it and it's a critical critical
[50:58] piece and then that is constantly
[51:00] upgrading so they're on the same you
[51:03] know they've got to be working with
[51:05] Nvidia for three or four generations in
[51:07] advance we we had this with Celestica
[51:11] Celestica
[51:13] was a contract manufacturer and this has
[51:16] been a disaster industry since 1999. It
[51:21] went all offshore to China. It was
[51:23] commodity, but they hung on and they
[51:26] they kind of kept
[51:28] Celestica's heritage was IBM
[51:30] supercomputing and they kept all that
[51:33] talent and skill. And then we noticed
[51:36] they were the sole supplier of the
[51:37] Google TPU server. We're like, "Oh my
[51:40] god, this was like three years ago. The
[51:42] stock was trading at eight times
[51:44] earnings." And they had this whole and
[51:46] then they also had this whole business
[51:48] of selling Ethernet white box which is
[51:51] code word for commodity white box
[51:54] Ethernet switches into the clouds.
[51:58] It it turns out that these are excellent
[52:02] businesses. Not only do they have
[52:03] tremendous growth, but to do an AI uh
[52:08] server computer, it's it's liquid
[52:11] cooled. It's running so much hotter and
[52:14] you know it's two or $300,000
[52:17] piece of machinery whereas an old server
[52:20] was $5,000. If it breaks you just throw
[52:22] it away. If this thing breaks the whole
[52:24] thing goes down. So you become like
[52:27] critical infrastructure like selling a
[52:29] critical part on a plane. You'll never
[52:32] get swapped out. And then they they it
[52:34] turned out they were quite good at
[52:36] liquid cooling and you know a lot of
[52:39] other people tried to do it and failed
[52:41] and so they've retained that position.
[52:42] Then it also turned out that the
[52:44] Ethernet market was because you were in
[52:48] the old days you would go from 100 gig
[52:51] to 400 to 800. It would be a 7-year
[52:56] cycle to upgrade. Now they're upgrading
[52:58] every year and that's really hard to do.
[53:02] Then there's a whole software layer, the
[53:04] open source sonic layer. The the guys at
[53:06] at Celestica invented were some of the
[53:08] people that wrote that open- source
[53:10] software. They work very closely with
[53:12] Broadcom. So what we thought was just a
[53:15] great growth driver turned out to be
[53:17] great competitive advantages and they
[53:19] have like 50 60% share of the cloud
[53:22] Ethernet switch market which is a
[53:24] crucial market for um AI because AI is
[53:28] incredibly network intensive. And then
[53:30] even something like the printed circuit
[53:32] board. I mean a regular server you need
[53:34] 10 layers. These AI servers you need a
[53:36] 40 layer and there's very few PCB
[53:40] suppliers that can make this. And um
[53:43] there's all kinds of complexities in
[53:45] there. And we also own Elite Materials
[53:47] which makes the leading ingredient which
[53:49] is copper clad laminate which goes into
[53:52] these boards. And so the PCB
[53:56] uh units are growing, the layer counts
[53:59] are rising. So you've got like a
[54:02] 50 to 60% keer just in the units and
[54:06] then the ASPs are rising and then the
[54:09] gross profits are rising and your
[54:12] visibility which used to be hey we'll
[54:14] call you next week if we need you to
[54:16] like hey we need you for the next four
[54:18] years to be like designing this road map
[54:20] with us. So you've you've gone from a 5%
[54:24] grow or low margin to you know a 35% 40
[54:29] 50 topline kager for the next four years
[54:32] with rising margins
[54:34] and then on top of that there's
[54:36] shortages of everything. So even if it
[54:38] is a commodity it's going to be a great
[54:41] cycle. So we see that up and down the
[54:44] supply chain. You find these companies
[54:46] like Corning like they make the fiber.
[54:49] Um they've got some ridiculously high
[54:52] share of the fiber. I was reading this
[54:55] uh Microsoft
[54:57] data center they just built. There's
[54:59] enough fiber to circle the world four
[55:01] and a half times in that one thing.
[55:04] And their fiber is thinner and more
[55:07] bendable and can be specially
[55:10] manufactured to the exact specs. and
[55:12] it's higher margin and it's the fastest
[55:14] growing part of their business. And then
[55:17] they're doing, you know, in networking
[55:19] there's scale out which is kind of
[55:21] connecting all the server racks
[55:24] together. Then there's scale across
[55:26] which is connecting the data centers
[55:28] together. And when you want to build one
[55:30] of these huge clusters and you can't get
[55:34] all the power in one place for training,
[55:36] you want to wire them together. But the
[55:39] wires you need like 10x the wire has to
[55:41] be so much thicker. So that's creating
[55:43] huge growth. And where the real kicker
[55:46] comes in is when you do scale up. That's
[55:49] connecting every GPU in the rack to the
[55:52] other ones. That's done over copper.
[55:54] Eventually that'll be done over fiber.
[55:57] when that happens that two to three X's
[56:00] Corning's opportunity. So you just have
[56:04] at every layer of of the rack,
[56:07] >> everyone's overwhelmed.
[56:09] >> Everyone's overwhelmed. But the story
[56:10] like in the power supplies, every Nvidia
[56:13] chip or rack uses
[56:17] n 50 to 125% more power. And like
[56:20] literally that drives the ASPs of Delta
[56:24] and Advanced Energy. I just I think it's
[56:27] it's I can't believe these stories when
[56:29] I hear I'm like wait so your ASPs are
[56:32] going to like go up 40% for the next
[56:36] four years in a row and it's higher
[56:39] margin. The broader picture is like
[56:41] we're going to be the AI demand if we're
[56:44] right with this L curve. We're already
[56:47] short, you know, the DRAM market, the
[56:50] NAN market, the PCB. We're already like
[56:54] 30, we're 30% short all these things as
[56:58] we are now.
[56:59] >> As your business scales up, everything
[57:01] gets more complex, especially your
[57:02] compliance and security needs. With so
[57:04] many tools offering band-aids and
[57:06] patches, it's unfortunately far too easy
[57:07] for something to slip through the
[57:09] cracks. Fortunately, Vanta is a powerful
[57:11] tool designed to simplify and automate
[57:13] your security work and deliver a single
[57:15] source of truth for compliance and risk.
[57:17] There's a reason that Ramp, Cursor, and
[57:19] Snowflake all use Vanta. It frees them
[57:21] to focus on building amazing
[57:22] differentiated products, knowing that
[57:24] compliance and security are under
[57:26] control. Invest like the best listeners
[57:27] get a special offer of $1,000 off Vanta
[57:30] when you go to vanta.com/invest.
[57:33] I know firsthand how complex the tech
[57:35] stack is for asset management firms. And
[57:37] seemingly every new tool and data source
[57:39] makes the problem even worse, adding
[57:41] more complexity, more headcount, and
[57:42] more risk. Ridgeline offers a better way
[57:44] forward. One unified platform that
[57:46] automates away that complexity across
[57:48] portfolio accounting, reconciliation,
[57:50] reporting, trading, compliance, and
[57:52] more, all at scale. Ridgeline is
[57:54] revolutionizing investment management,
[57:56] helping ambitious firms scale faster,
[57:58] operate smarter, and stay ahead of the
[57:59] curve. See what Ridgeline can unlock for
[58:01] your firm. Schedule a demo at
[58:03] ridgeline.ai.
[58:04] >> The measure of percent AI, percent
[58:06] market share. Do you care more about the
[58:08] absolute or the rate of change of those
[58:10] metrics?
[58:11] >> It's good because I took we did this
[58:13] presentation in two 2024 where we
[58:16] actually listed everybody's market share
[58:19] and everybody's and then I I asked
[58:21] Claude to plot plot it to a thing and it
[58:24] actually didn't get it right because
[58:26] what it didn't get is the rate of
[58:27] change. So the rate of change is
[58:29] important and that's incredible too
[58:31] because you go from 10% to 30% and your
[58:35] growth rate accelerates and your margins
[58:36] accelerate. So rate of change is very
[58:39] important.
[58:39] >> Why don't more people get this right in
[58:41] public markets? Like if your whole
[58:42] framework is S-curve, competitive
[58:44] advantage, underappreciated earnings
[58:46] power. It feels like the movie's been
[58:48] played out a lot over the last 25, 30
[58:51] years.
[58:52] >> My mom said, "Why do you tell everyone
[58:54] your secret? It's like it's why does the
[58:57] casino teach people how to play
[58:59] blackjack? It it's harder. It's really
[59:01] hard to do. It's it's you have to have a
[59:03] a deep you have to be comfortable
[59:05] investing. You know, we've been doing
[59:07] I've been doing tech for 20 years at
[59:09] Whale Rock. We've got a team that's been
[59:12] doing this, covered many cycles. We know
[59:14] the different. So, very few people, no
[59:16] one's paid attention to hardware and
[59:19] chips at all. So, you've got all these
[59:21] newbies coming into it.
[59:23] >> You and Gavin, that's it. and Gavin's
[59:24] done a great job. people weren't
[59:26] comfortable with it and it's it's harder
[59:28] to do than it seems and the chart you
[59:30] know a lot of these companies their
[59:31] charts are up so it's scary can I buy
[59:33] and then you also have to have the
[59:35] holistic view because if you don't have
[59:37] conviction so every you know every time
[59:40] with Nvidia over the last four years
[59:42] it's like oh they had a great year oh my
[59:45] god it's got to be a bubble and then
[59:48] they had another great year and it's
[59:50] like 6 months of marking time it's got
[59:52] to be a bubble this is like getting out
[59:54] of hand this is pretty scary like and
[59:57] the the bare cases are not like totally
[01:00:02] without merit but if you can see the
[01:00:04] whole picture and understand how these
[01:00:06] things are unfolding and gain conviction
[01:00:08] in that frankly if you're just a semi-
[01:00:10] analyst so many semi analysts missed it
[01:00:13] because they didn't see what was really
[01:00:15] happening at the foundational model
[01:00:18] layer so it helps to have have the big
[01:00:20] picture it helps to have you know
[01:00:22] decades and scores of scurves that
[01:00:25] you're looking at and and where it plays
[01:00:27] in different things.
[01:00:28] >> What what in this whole picture, you
[01:00:30] know, I would describe your your stance
[01:00:31] so far in the first hour discussion as
[01:00:33] like very bullish on on the impact that
[01:00:36] AI is going to have and the returns
[01:00:37] available as a result. What makes you
[01:00:39] the most concerned or uncertain or is it
[01:00:42] just the rate at which all this stuff
[01:00:44] changes and like what keeps you worried
[01:00:47] amidst what seems like pretty extreme
[01:00:49] bullishness? I mean, one thing that
[01:00:51] bothers me is there's a lot of
[01:00:53] negativity in the general population
[01:00:56] about AI and there's a lot of negativity
[01:00:59] in some aspects of the government. You
[01:01:01] know, I think Maine just banned data
[01:01:03] centers and
[01:01:05] 80 only 20% of the people are optimistic
[01:01:08] about AI and potential for negative
[01:01:11] regulation. But I do think kind of the
[01:01:13] genie is out of the bottle. Another risk
[01:01:16] is that if AI sort of slows down in its
[01:01:21] improvements, I think there's a whole
[01:01:23] lot of AI adoption to happen even if the
[01:01:26] models didn't improve. But Jensen said
[01:01:29] this, you know, years ago when he was
[01:01:31] talking about his GP crap, just the
[01:01:33] graphics chips. If good enough is good
[01:01:36] enough, I won't have a business. Now
[01:01:40] every year he made the graphics a little
[01:01:42] bit better and people always wanted the
[01:01:45] best in AI. If anthropics sort of hits a
[01:01:49] wall and stops improving or open AI then
[01:01:52] the open source models will catch up and
[01:01:56] um and then it might be a race to the
[01:01:58] bottom and it might be you know it won't
[01:02:01] be good for the stocks probably. It
[01:02:03] could be good for the chip companies.
[01:02:05] chip companies don't care
[01:02:07] >> who's winning tokens, right?
[01:02:08] >> Who wins.
[01:02:09] >> So, that's another positive and they'll
[01:02:11] benefit if if open source, you know,
[01:02:14] Jensen really wants open source to like
[01:02:16] take off. It's all he kept on mentioning
[01:02:18] at at his last GTC. So, that could be a
[01:02:21] risk. Another thing is if one or two of
[01:02:24] the players falters and loses its
[01:02:26] position and can't compete, that could
[01:02:29] be like a lot of compute that they don't
[01:02:32] need in the future. Now, if AI is so
[01:02:34] big, somebody else will suck that up.
[01:02:36] And we saw that with, you know, Oracle
[01:02:38] cancelled a big deal and then Meta went
[01:02:40] right in. But let's just say Meta
[01:02:43] decided not to be involved
[01:02:46] with AI. Hey, we can't keep up. It's
[01:02:48] just going to be a waste of our
[01:02:49] resources. So, we we watch that very
[01:02:51] carefully and um in general, we see
[01:02:55] more, you know, more more companies
[01:02:58] truly going after this and even
[01:03:00] Microsoft going trying to build their
[01:03:02] own. So I think those are those are some
[01:03:04] of the key risks.
[01:03:05] >> Seems like you really have done very
[01:03:07] little in the application layer of AI.
[01:03:10] Historically the apps ended up being
[01:03:11] most of the market cap you know not not
[01:03:13] the infrastructure and there wasn't
[01:03:15] really a model layer in the past. I
[01:03:16] guess you could say it was the clouds or
[01:03:17] something.
[01:03:17] >> Yeah.
[01:03:18] >> Why focus so much on the bottom layers
[01:03:21] of Jensen's five layer cake versus
[01:03:23] things in the application layer that are
[01:03:25] actually getting used by consumers?
[01:03:27] Well, we do, you know, part of OpenAI is
[01:03:29] they have chatbt which which is an
[01:03:31] application, but we think the
[01:03:33] application layer well a it always comes
[01:03:36] later. So, you know, the first three or
[01:03:38] four years of the iPhone and then the
[01:03:41] applications really took time. So, maybe
[01:03:44] it's just starting. Um but to date um we
[01:03:48] found that area to be pretty risky
[01:03:50] because where does the where does the
[01:03:53] foundational model end and where does
[01:03:55] the application begin and can can the
[01:03:58] applications build enough of a moat um
[01:04:02] where they can fend off and um and build
[01:04:06] and build businesses in that. and um and
[01:04:10] we we thought we would see it in some of
[01:04:12] the incumbents like a a CRM and they're
[01:04:15] starting and maybe just a matter of time
[01:04:17] but we really haven't seen it in the
[01:04:19] enterprise world and there there are
[01:04:21] some you know very good
[01:04:25] startup application companies out there
[01:04:27] but the ecosystem is not clear you know
[01:04:30] like when we started the ecosystem and
[01:04:32] chips was clear when we started the
[01:04:34] foundational model ecosystem wasn't
[01:04:36] clear now it's clearer to us and at the
[01:04:39] application layer it's still kind of
[01:04:41] unclear and a little bit dangerous
[01:04:44] because um but there will be great
[01:04:46] application companies built you know we
[01:04:49] really were watching Brett Taylor at
[01:04:50] Sierra Brett was CEO of CRM he wrote
[01:04:54] Google Maps he was CIO of Facebook and
[01:04:58] uh he he's building this fantastic
[01:05:00] company called Sierra we're not involved
[01:05:02] but that's where the rubber hits the
[01:05:04] road will he be able to turn this into a
[01:05:06] huge company and he's doing quite well.
[01:05:09] We'll see. It's a matter of timing when
[01:05:12] these things really start to to to come
[01:05:15] in into their own and prove they're
[01:05:17] sustainable. It usually doesn't start in
[01:05:19] the first 3 or 4 years. It comes a
[01:05:21] little bit later.
[01:05:22] >> At your office, you have this this giant
[01:05:24] uh award wall for the research. I can't
[01:05:26] remember what it's exactly. It's for the
[01:05:28] best research job or project of the year
[01:05:30] given to an analyst. And I think you won
[01:05:32] it. you gave it self awarded in their
[01:05:34] own when you're by yourself, but you've
[01:05:37] got this now long 20 year history of a
[01:05:39] year one or one or more people, you
[01:05:41] know, put their name on this wall for
[01:05:43] having done the best job on a research
[01:05:45] project that year. I'm so curious about
[01:05:47] the nature of that research and how it's
[01:05:49] changing as a result of all of this.
[01:05:51] Say, you know, the person that's going
[01:05:53] to win the award this year and the sort
[01:05:55] of work that that requires a human to do
[01:05:58] when so much of the work that probably
[01:05:59] would have won you the award in, I don't
[01:06:01] know, 2009 or something could probably
[01:06:03] be fully automated or done in an hour
[01:06:05] with cloud code or something today. How
[01:06:07] is the nature of research and what gets
[01:06:09] you on that whale rock award wall
[01:06:11] changing in real time? I would like to
[01:06:14] say that we're so advanced in our AI
[01:06:16] systems that it's a huge change so far.
[01:06:20] I mean, it's it's helping us get up to
[01:06:21] speed and we have a handful of of great
[01:06:23] apps, but it's not yet it's not
[01:06:27] supplanting the job of the analysts. And
[01:06:29] so much of what we're doing is we're
[01:06:31] meeting with as many companies as
[01:06:32] humanly possible. We're developing
[01:06:35] relationships with with the management
[01:06:37] teams that we cover. We're talking to
[01:06:39] the competitors. The system we use is
[01:06:42] right out of common stocks and uncommon
[01:06:44] profits which was written by Philip Fish
[01:06:47] Fischer in the 1950s. And it's the
[01:06:49] scuttlebutt approach. It's growth
[01:06:51] investing. It's it's get out there and
[01:06:54] talk to suppliers, uh, customers,
[01:06:57] competitors, looking for the key
[01:06:59] characteristics of these leading
[01:07:00] companies and really developing
[01:07:02] conviction in them. Now, if it's a new
[01:07:04] complicated area like ABF substrates or
[01:07:08] PCBs, we're able to get up to speed on
[01:07:11] those things quickly, but it can't pick
[01:07:14] stocks for you in any kind of a way. I
[01:07:17] will say that, you know, if you're an
[01:07:19] analyst who's good at the blocking and
[01:07:22] tackling and there's a role for that,
[01:07:24] but that role is you need to have
[01:07:28] obviously the insight on top. So, we're
[01:07:30] now like using AI to write notes uh or
[01:07:34] you know review the quarter or and those
[01:07:37] notes are much better but there better
[01:07:39] be a really good paragraph on top which
[01:07:43] is the wisdom. What does this mean? How
[01:07:45] does this deal with our thesis? Um what
[01:07:48] changed? You know, don't just be a
[01:07:50] reporter. Um so the AI can be a great
[01:07:52] reporter. It can't it can't quite pick
[01:07:55] into the future. And like the job that
[01:07:57] the guys did on app 111 two years ago. I
[01:08:00] mean I think we got two of the best adte
[01:08:02] guys around and they you know they
[01:08:05] convinced me to buy I knew adte I
[01:08:08] started actually nearby here in New York
[01:08:10] at at that internet advertising startup
[01:08:13] and after I did banking.
[01:08:16] So I knew internet advertising and ad
[01:08:18] tech which is historically a terrible
[01:08:21] industry. Um, but Michael and Sam really
[01:08:26] figured out the Apploven story like
[01:08:28] before anybody and they followed it when
[01:08:30] it was private. They know all the
[01:08:31] competitors. They know all the
[01:08:33] intricacies of, you know, there's all
[01:08:35] this terminology and um they, you know,
[01:08:39] Sam went to the Las Vegas app
[01:08:42] advertising conference and we went to
[01:08:44] con and, you know, we talked to scores
[01:08:46] and scores of of people. So um and they
[01:08:50] did the work on the model and developed
[01:08:52] a great relationship with Adam Ferogi.
[01:08:54] He's one of the best managers out there.
[01:08:56] And um I don't see AI doing that.
[01:08:59] >> What role does talking to other
[01:09:01] investors outside of your firm play in
[01:09:03] your life? Like
[01:09:05] >> I one of the great things is just the
[01:09:07] friendships I've built with so many
[01:09:10] smart investors
[01:09:12] and and frankly Philip Fischer said part
[01:09:15] of his process was like get to know a
[01:09:17] good 10 or 15 like-minded people around
[01:09:20] the country and share ideas and um
[01:09:25] and a you know they're great great
[01:09:27] friends to make a lot of them have been
[01:09:29] on your podcasts and uh and and you
[01:09:32] develop good friendships and and you you
[01:09:35] share ideas, talk ideas. It's important
[01:09:37] that it's a two-way street. Um, I call
[01:09:40] it the tripod. When I like something
[01:09:44] and then my analyst likes it and then
[01:09:47] somebody who I really respect also likes
[01:09:49] it. That's three legs of the stool can
[01:09:52] really help the conviction.
[01:09:54] >> What have you learned about shaping the
[01:09:56] products that you offer your investors
[01:09:59] across the history of the firm? It's not
[01:10:01] just one monolithic structure anymore.
[01:10:04] >> There's there's several things that if
[01:10:05] I'm an investor and I want to give you
[01:10:06] money, I can there's a couple ways I can
[01:10:08] do that.
[01:10:09] >> How did you arrive at those things? And
[01:10:10] and how do you how could you turn that
[01:10:12] experience into um advice for other
[01:10:15] investors that are trying to provide
[01:10:17] their LPs with the right set of options?
[01:10:20] >> For the first 15 years, it was a long
[01:10:22] short fund and we you know, you want to
[01:10:24] be focused and if you defocus that can
[01:10:26] be hard. So we we grew that and we got
[01:10:29] that to the scale that we wanted to.
[01:10:32] We're 20 years old, maybe 10 years in,
[01:10:34] people started to ask for a long only
[01:10:36] product. And so in 2020, we we launched
[01:10:42] uh the long only fund. So we're 6 years
[01:10:45] on that. And um that's now larger than
[01:10:50] the long short. The bulk of the assets
[01:10:52] are in these two. In maybe 2015, we we
[01:10:56] formalized that we might be doing
[01:10:57] privates. And so we gave investors the
[01:11:01] option to opt in or opt out and you
[01:11:03] could do 15% or 25%. So, but we didn't
[01:11:07] break the seal on the privates until
[01:11:08] 2020. In 2021,
[01:11:11] we offered um a hybrid fund that could
[01:11:14] be 80%
[01:11:17] uh into privates. sort of similar
[01:11:19] approach but if you wanted more exposure
[01:11:21] to privates. Um and then very recently
[01:11:25] we launched the whale rock meggaap tech
[01:11:28] fund and we just think there's a huge
[01:11:32] structural underweight of the largest
[01:11:35] tech companies in the world because a we
[01:11:38] also realize that a a lot of our
[01:11:40] performance over the years was from some
[01:11:42] of the largest companies whether it be
[01:11:45] Apple or Amazon or Tesla and and people
[01:11:50] just it's hard to overweight these to
[01:11:52] the to the amount. And so a lot of our
[01:11:54] largest pools of capital endowments or
[01:11:57] what have you, they realize they they
[01:11:59] have been massively underweight, the
[01:12:01] largest tech companies in the world for
[01:12:03] the last because they only have, you
[01:12:06] know, they have a lot of privates. They
[01:12:09] don't have a ton of public and then
[01:12:12] maybe half the public is international.
[01:12:14] And then of their public bucket, they
[01:12:18] don't want to they there's a belief that
[01:12:20] there's no alpha in large cap. So they
[01:12:22] underweight large cap and they have a
[01:12:23] lot of small and mid managers that are
[01:12:25] stock pickers because it's intuitive
[01:12:27] that large cap can't have alpha. Um and
[01:12:31] then in their hedge fund portfolio, even
[01:12:33] if it's long bias, they're not going to
[01:12:34] have 15% and Nvidia and all these other
[01:12:37] things. And we realize that there's a
[01:12:41] huge that this people are worried that
[01:12:43] there's these big companies. This is
[01:12:44] just a product of the digital economy in
[01:12:46] that, you know, in tech, the leader
[01:12:48] usually grows bigger and wins and
[01:12:50] develops very high market share quickly
[01:12:53] and and there's great competitive
[01:12:55] advantages and and they're also selling
[01:12:57] around the globe. So, this is going to
[01:12:58] lead to massive profit pools and massive
[01:13:01] market caps and it's just going to
[01:13:03] happen into the future. And so, most
[01:13:06] endowments are betting against this.
[01:13:08] They're they're because they're
[01:13:10] completely underweight this. And finally
[01:13:13] somebody came to us and said you know
[01:13:15] what should we do which index we got to
[01:13:17] and I'm on the board of Hamilton College
[01:13:19] and they were trying on their investment
[01:13:22] committee they were trying to figure out
[01:13:23] this and so we kept on hearing it and
[01:13:25] finally uh one of our clients was like
[01:13:28] we said we'll we'll do this for you
[01:13:31] because there's a lot of alpha to be had
[01:13:32] and the mag 7 or the fang or whatever
[01:13:35] it's going to be different and you know
[01:13:37] in 2022 they all rallied but like last
[01:13:40] year they were very divergent and this
[01:13:43] year they're down and so we created the
[01:13:46] the Whale Rock Mega Cap Tech Fund which
[01:13:48] is the top 30 the universe is the top 30
[01:13:51] market caps globally and then we pick
[01:13:54] you know the 12 or 13 that are the best
[01:13:57] and I think there's tremendous alpha in
[01:14:00] the largest cap because if you think
[01:14:02] about it a small cap it just takes one
[01:14:05] person to to figure out it's good and
[01:14:07] move it up but it takes a hundred
[01:14:10] people, 100 diversified PMs to realize
[01:14:14] Google's not a loser, it's a winner. And
[01:14:17] can we figure that out before
[01:14:21] 95% of those generalist PMs
[01:14:25] do it? And you know, we've been able to
[01:14:27] do it.
[01:14:27] >> We like your odds in that.
[01:14:28] >> Yeah, we like your odds in that. And so
[01:14:30] there is alpha to be had there. And then
[01:14:32] as an asset category, it's great because
[01:14:35] these companies by definition have
[01:14:36] wonderful modes and maybe they're not
[01:14:39] the super S-curve, but sometimes they
[01:14:42] are. I mean, Nvidia sure is, and TSM is
[01:14:46] really levered to it, and Heinix is
[01:14:49] extremely levered to it, and ASML is
[01:14:51] levered to it. So, it's a great um so
[01:14:54] that's a new we're four months into that
[01:14:57] one. And so the right way maybe to think
[01:14:58] about it, it it it sort of sounds like
[01:15:00] really what you've built is a research
[01:15:02] machine to understand the world through
[01:15:05] the lens of companies and that the thing
[01:15:08] you're constantly trying to improve is
[01:15:10] that research machine and then the way
[01:15:11] that you would then express that through
[01:15:13] products is multiplied. But if I was to
[01:15:15] try to understand Whale Rock, it would
[01:15:16] be to investigate the research machine
[01:15:18] first and foremost. We call it the whale
[01:15:20] rock learning machine and it's a group
[01:15:22] of 10 highly experienced individuals
[01:15:25] that you know Warren Buffett reads books
[01:15:28] and we read books and we read blogs and
[01:15:30] we but we're also in tech you got to go
[01:15:33] out and talk to people. So we do 2500
[01:15:36] 3,000 facetoface meetings with
[01:15:37] management teams and you know Mer and
[01:15:40] Buffett talk about compounding
[01:15:41] knowledge. We've been compounding that
[01:15:43] knowledge for 20 years. you know,
[01:15:46] there's changes to the team, but broadly
[01:15:49] there's a lot of um consistency to it.
[01:15:53] Um Andrew and Michael have been with me
[01:15:55] for 19 and 18 years and the average
[01:16:00] experience level on the team is 10 or so
[01:16:02] years and that includes some of the new
[01:16:04] newer people. And uh yeah, that research
[01:16:07] engine can support all these all these
[01:16:10] products and it's the same people that
[01:16:12] do publiclix and the private. So, we're
[01:16:14] not going to scour the world and turn
[01:16:16] over every A B. But when we see
[01:16:19] something that fits into our system,
[01:16:21] we're able to act on it.
[01:16:23] >> It's so much fun to do this with you.
[01:16:24] When I do this, I ask the same
[01:16:25] traditional closing question of
[01:16:26] everybody. What is the kindest thing
[01:16:28] that anyone's ever done for you?
[01:16:30] >> Well, I got to say it's definitely my
[01:16:32] father who, you know, I was super lucky.
[01:16:36] My father um graduated Cornell a double
[01:16:40] e electrical engineering, pivoted to
[01:16:43] Wall Street and um uh had a great career
[01:16:46] at Goldman Sachs and he was um he ran
[01:16:51] corporate finance in the 80s and then
[01:16:53] ran private equity as chairman in the
[01:16:56] 90s and uh he was just whips smart but
[01:17:00] he he had such humility and was such a
[01:17:03] great gentleman and uh when I started
[01:17:07] Whale Rock, you know, friends and
[01:17:08] family, he was the first call, but he
[01:17:10] said, you know, I've been at Goldman for
[01:17:12] for 41 years. How about I come and join
[01:17:16] you? I'll be the gray hair. I'll be the
[01:17:18] oversight. I'll be the chairman. You do
[01:17:19] what you do. You build the firm in uh
[01:17:22] Boston. Build the team, run the money,
[01:17:25] I'll help raise some money. And we got
[01:17:27] to work together for 6 years until he
[01:17:30] passed away in 2011. But I just feel so
[01:17:33] lucky to have worked with him. You know,
[01:17:35] it's not easy running a fund. We never
[01:17:37] raised our voice. And he was just an
[01:17:40] amazing mentor to so many people. And
[01:17:43] when he passed away,
[01:17:46] um I got so many letters from people who
[01:17:48] said, "Your father was just such an
[01:17:52] influence on me. He was such a
[01:17:53] gentleman. He was such a great mentor to
[01:17:55] me." And so I just feel so lucky uh to
[01:17:58] have worked with him. And if I could be
[01:18:00] half the person that he is, I'd be
[01:18:03] completely winning. And
[01:18:04] >> how did he do that? How did he What was
[01:18:06] his method? Why did so many people say
[01:18:08] that?
[01:18:09] >> Um
[01:18:11] I don't know. He He just He was He was
[01:18:13] modest. He was whipsmart. He was wise.
[01:18:17] He was also known as a um a great
[01:18:22] investor, which isn't the most common
[01:18:23] thing at a lot of investment banks. He
[01:18:25] also was on their commitments committee
[01:18:27] and kept him out of a lot of tougher
[01:18:29] situations and yeah he was very warm and
[01:18:32] he people would could go into his office
[01:18:35] with with with problems and he handled
[01:18:38] it handled it with grace and um whether
[01:18:41] it's a personal problem or a work issue
[01:18:43] or what have you and he just had this
[01:18:45] soft way and he also had a great sense
[01:18:47] of humor.
[01:18:48] >> Lucky.
[01:18:48] >> Yeah. I'm so lucky. So Alex, thanks so
[01:18:51] much for your time.
[01:18:52] >> Thanks so much.
[01:18:58] Your finance team isn't losing money on
[01:19:00] big mistakes. It's leaking through a
[01:19:01] thousand tiny decisions nobody's
[01:19:03] watching. Ramp puts guardrails on
[01:19:05] spending before it happens. Real-time
[01:19:07] limits, automatic rules, zero
[01:19:08] firefighting. Try it at ramp.com/invest.
[01:19:13] As your business grows, Vant scales with
[01:19:15] you, automating compliance and giving
[01:19:16] you a single source of truth for
[01:19:18] security and risk. Learn more at
[01:19:20] vanta.com/invest.
[01:19:23] Ridgeline is redefining asset management
[01:19:25] technology as a true partner, not just a
[01:19:27] software vendor. They've helped firms 5x
[01:19:29] and scale, enabling faster growth,
[01:19:31] smarter operations, and a competitive
[01:19:33] edge. Visit ridgelineapps.com to see
[01:19:36] what they can unlock for your firm.
[01:19:38] Every investment firm is unique, and
[01:19:40] generic AI doesn't understand your
[01:19:41] process. Rogo does. It's an AI platform
[01:19:44] built specifically for Wall Street,
[01:19:45] connected to your data, understanding
[01:19:47] your process, and producing real
[01:19:48] outputs. Check them out at
[01:19:50] rogo.ai/invest.
[01:19:53] The best AI and software companies from
[01:19:55] OpenAI to cursor to Perplexity use work
[01:19:57] OS to become enterprise ready overnight,
[01:19:59] not in months. Visit works.com to skip
[01:20:02] the unglamorous infrastructure work and
[01:20:04] focus on your product.