# web062426

https://www.youtube.com/watch?v=sDj2UF316jw

[00:00] Uh, a lot of people reached out
[00:02] yesterday. Seems like a day can make
[00:05] people uh, a little nervous. So, I just
[00:07] wanted to give a uh, an intraweek quick
[00:10] video uh, really to remind you guys
[00:14] we're navigating through the AI midcycle
[00:16] slowdown and whatever you're hearing
[00:18] from people. And this is really it
[00:20] wasn't just the outreach from
[00:22] subscribers. It was also watching some
[00:25] videos yesterday and starting to hear
[00:29] people say, "Here we go. That was it."
[00:31] Um, the bubbles bursting. So, I just
[00:34] want to basically take you guys through
[00:37] uh a quick one and remind you where we
[00:40] are. So, first of all, here's the
[00:42] thematic portfolio. Made new highs last
[00:45] week, went down. I I want you to just
[00:49] look at what's really been happening
[00:51] since the end of May, which is the rate
[00:54] of change has slowed down and we've been
[00:57] doing some back and forth.
[01:01] I wrote this piece in early June saying
[01:04] the fireworks show was over
[01:06] and that now that people are aware of AI
[01:09] agents,
[01:11] we're past the point of repricing the
[01:14] market for AI agents. So that first
[01:16] quarter was a repricing guys and a
[01:18] repricing is seen through the earnings
[01:21] which I'll get through. So you just have
[01:23] to remember this was not speculation.
[01:26] This was not something that was driven
[01:28] purely by retail. This is all about
[01:32] earnings. This isn't about anything more
[01:35] than the agentic world has started to
[01:38] take over which means adoption has taken
[01:40] over. Most importantly, the thing
[01:43] driving the market, which you will hear
[01:45] forever, is token usage. Not token
[01:48] prices, token usage.
[01:52] I wrote this piece 10 days later. The AM
[01:55] midcycle slowdown is here. In macro,
[01:58] midcycle slowdown is the pause inside an
[02:00] expansion. Growth is still positive and
[02:02] the cycle is still intact, but the rate
[02:05] of change cools.
[02:08] Here is the rate of change of my
[02:10] thematic portfolio. The red line here is
[02:13] the 30-day. It peaked in miday at some
[02:17] insane level and started to deteriorate.
[02:20] So when markets have parabolic moves,
[02:23] they're they need within an expansion,
[02:26] they need to resolve themselves through
[02:28] some relationship of price and time. It
[02:31] can come from a violent fall of 20% in a
[02:34] day and then start to go higher again.
[02:37] Or it can happen by
[02:39] rising volatility, higher range,
[02:42] grinding higher. Every now and then we
[02:44] get these candlesticks that get make you
[02:47] think like it's over. But when you look
[02:49] back, all that happened was the rate of
[02:50] change reset. So if you go look here,
[02:53] the 30-day rate of change on my thematic
[02:55] portfolio is almost back to zero, which
[02:57] is where it was in March
[03:01] at the beginning after Iran. So we've
[03:03] done a lot of damage. Now, the 50-day is
[03:06] still up here, but it's going to be
[03:08] tracking lower. So, we're doing exactly
[03:11] the midcycle slowdown. Nothing should be
[03:14] shocking. I mentioned this. This is what
[03:17] got me after the subscribers. I watched
[03:20] some videos last night. People are like,
[03:21] "Yeah, anyone that says this is the
[03:23] second or third inning in AI is nuts.
[03:25] This is the end of it." Blah, blah,
[03:27] blah. Here's a candlestick, a big
[03:30] engulfing outside day
[03:33] in SKH Heinox. So SKH Highix
[03:37] suggests that that's it. We get kind of
[03:39] a bounce here. Would not surprise me if
[03:41] in SKH Highix we spend the time here and
[03:44] eventually get to the 50-day. Now the
[03:46] 50-day is rising. So the 50-day could
[03:48] get hit over here. This is what a
[03:50] midcycle slowdown and consolidation
[03:53] looks like. If the earnings start to
[03:55] drop off dramatically or if people
[03:57] believe the memory side has been
[03:58] completely resolved or all of the data
[04:01] center stop being built or all of the
[04:03] capex,
[04:04] none of that's going to happen, guys,
[04:05] because token usage is still going
[04:07] through the roof. But SKHEX, same thing.
[04:12] Here's the 30-day rate of change. Here's
[04:13] the 50. These are all healthy
[04:15] retracements that needed to happen. I
[04:19] give you guys a weekly sheet on
[04:21] technicals. I gave you something on
[04:23] saying if I were you, I'd save them in
[04:25] the same folder and then just go to
[04:26] Claude and ask them to do this every
[04:27] single time. Above the 20-day moving
[04:30] average, we peaked May 8th. Above the
[04:32] 50-day moving average, we peaked May
[04:35] 22nd. 200 day, we peaked May 29th. This
[04:39] is the health of the market. Are the
[04:41] 50-day still rising? Are the 200 days
[04:43] still rising? for these to turn down
[04:46] takes a lot more damage than just price
[04:48] coming back. So, this is what I care
[04:50] about. And look, in these cases,
[04:53] we're still near the peak on both of
[04:55] these. This is a healthy correction, and
[04:57] that's literally what Claude sees, too.
[04:59] The short term is choppy, but the
[05:01] intermediate and long-term breath have
[05:02] stabilized in a healthy range. Yesterday
[05:06] in this horrible down day and it felt
[05:07] bad for anyone who's got my thematic
[05:10] portfolio or any AI positions, the S&P
[05:13] was down one and a half percent. But
[05:15] more sectors were up. And these weren't
[05:17] like small moves. I mean, you got
[05:18] staples up 2%. You got healthcare up
[05:20] one. This was a rotation. We're at the
[05:22] end of a quarter. You've had a massive
[05:24] move. Pension rebalancings are
[05:27] occurring. Raia rebalances are
[05:29] occurring. Everyone's rebalancing. And
[05:32] it should be negative for stocks, good
[05:34] for bonds on one relationship. It should
[05:36] be bad for the US, good for global
[05:38] stocks on another relationship. And in a
[05:40] rebalancing, it should be good for the
[05:42] underperformers versus the over uh
[05:44] producers, which is what we've seen.
[05:47] Another sign that yesterday, and really
[05:49] this whole consolidation has been a
[05:51] consolidation, is that IWM continues to
[05:54] outperform SPY. So when you see a big
[05:57] down move in the S&P by 1.44% 44%. And
[06:00] as I'm going to show you, Beta was
[06:02] horrible yesterday. Normally, when Beta
[06:04] is horrible, IWM underperforms the S&P
[06:06] by a lot. Not only did it outperform
[06:08] yesterday, look at what it's been doing
[06:10] this month since the consolidation and
[06:12] the slowdown began. Small caps are
[06:15] outperforming. Good, healthy sign. Here
[06:17] was Beta. Beta has been atrocious during
[06:19] this period in terms of the volatility.
[06:21] So, overall since June 1st, Beta hasn't
[06:26] moved that much. much. I mean, we're
[06:27] down about 3%, but you can see how many
[06:29] moves we've seen violently, and that's
[06:31] the issue that shows up through here.
[06:33] This is the 50-day V of TMT MO. So,
[06:37] technology momentum factor is up at
[06:41] extreme levels. Basically, the only time
[06:43] in history is back in 2020 after COVID
[06:46] and then the dot bubble. This is where
[06:48] everyone just keeps jumping back to.
[06:50] They keep jumping back to the dot
[06:51] bubble. This is not the.com bubble.
[06:53] Demand is greater than supply, but I'll
[06:54] get into that again. But you do have
[06:56] incredibly high volatility. The reason
[06:58] this is important, high volatility makes
[07:01] it difficult for
[07:03] long short managers. It makes it
[07:05] difficult for people to stay in trades
[07:07] that are running market neutral. High
[07:09] factor volatility like this on a sector
[07:12] basis makes it impossible for people to
[07:14] just ride the trend. One of the ways
[07:16] bull markets continue is to climb a wall
[07:18] of worry, but sometimes they climb a
[07:20] wall of volatility as well.
[07:23] This is getting way too much press now.
[07:24] Um, this means nothing. Okay, I'm going
[07:27] to say it out loud. I've said it before
[07:28] that it would impact people's narratives
[07:31] and we've now gotten there. So, this has
[07:33] become a bubble in terms of the
[07:35] narrative. We are at a slowdown in the
[07:37] Western world which pays the most money.
[07:40] So, think of this as everyone's on
[07:42] vacation. Are more people apt to buy the
[07:45] cheaper stuff or the expensive stuff?
[07:47] the 200 subscriptions, all the token
[07:49] usage that is going on right now. We're
[07:51] just not getting usage from the
[07:53] expensive side. At the same time,
[07:55] businesses are trying to decide after
[07:58] price discovery what they should do. So,
[08:00] it makes sense for this to go down. My
[08:02] ultimate guess is since this is a
[08:04] commodity, meaning the token pricing
[08:06] side, is that we're going to continue to
[08:09] see what Brian Armstrong with other
[08:10] smart people have talked about, which is
[08:12] over time, the expensive models are
[08:14] going to become less of the pricing. And
[08:17] so, I expect this to not be trending
[08:21] higher like it was from here. This is
[08:23] not a sign of a bubble. This is not a
[08:25] sign of a peak. Stop looking at this as
[08:27] like some gauge on things. Now, I put
[08:30] this out. I did this on the webinar that
[08:32] I did for subscribers in the middle of
[08:34] June. Navigating the AI midcycle
[08:36] slowdown. So, this is the part where I
[08:39] want to remind you signal verse noise.
[08:42] Always remember when people tell you
[08:44] semis have had a huge run and they're
[08:45] going to have to pull back. It's
[08:47] possible, but their earnings per share
[08:49] are going to have to pull back as well.
[08:51] And the only way that's going to happen
[08:53] is on something that isn't going to
[08:54] happen anytime soon. So, the reason this
[08:57] went so much higher is because the
[08:59] earnings per share number skyrocket at a
[09:01] pace that it never has seen. Do I think
[09:04] it's going to be more like this? Maybe
[09:06] more like this. It isn't going to be
[09:07] like this the entire time because
[09:09] estimates have gone higher and the or
[09:12] the actual stocks have gone higher and I
[09:15] think we're going to be in an capex air
[09:17] pocket like I said in the last video. So
[09:20] expect this to start going like this
[09:22] which means the semi should stop moving
[09:23] in a parabolic nature and it should be
[09:25] more long short. Think about what Nvidia
[09:27] and Broadcon have done done over the
[09:29] course of the last year to 18 months
[09:32] despite the fact that their earnings are
[09:33] up dramatically. You can't get past this
[09:36] point either. So this is a great way to
[09:39] look at it. This is the last 25 years of
[09:42] year-over-year forward EPS. We have
[09:46] skyrocketed on SPX. And the reason I
[09:48] want you to look at this, these are the
[09:50] other times that happened. They were
[09:52] coming out of a recession.
[09:54] Here's the reality. When you have this
[09:56] kind of a move, this has historically
[09:58] been the beginning, not the end of
[10:01] something. So unless we get a dramatic
[10:04] turn in this, which I do not expect, I
[10:07] think you're going to have to be in a
[10:08] situation where the multiple compression
[10:10] story is actually happening. the S&P is
[10:12] not going up as fast as the earnings,
[10:14] which again is another sign when you see
[10:16] Nvidia trading sub 20p.
[10:20] You're not you're not in a bubble here.
[10:21] Um the surprise was how big it was, too.
[10:24] It's not just how fast it's going up.
[10:26] It's that this was the largest surprise
[10:29] uh in beats in years.
[10:32] Here's what normally happens seasonally,
[10:34] the fact that this went up. I'm just
[10:36] reminding you of the fact that this was
[10:38] all about earnings. And so if the
[10:41] market's going to correct, you're going
[10:42] to need earnings estimates to start to
[10:44] come down. And right now, revisions are
[10:46] still going higher. Now, if it was just
[10:48] the S&P 500, you could hear people
[10:50] argue, well, this is just two stocks.
[10:52] But it's not just two stocks, guys. This
[10:54] is a broad broad broad rally. I have a
[10:57] hundred names in my thematic portfolio,
[11:00] and not all of them are in the S&P. So,
[11:04] this is not an S&P thing. small cap
[11:06] numbers, whether it's their earnings,
[11:08] whether it's their profit margins,
[11:09] whether it's their revenues, here's
[11:12] revisions, all going higher, like we're
[11:15] coming out of a recession. Remember, in
[11:17] small caps, we've been anything that's
[11:20] positive is out of the norm. Except for
[11:22] coming out of the recession here and the
[11:24] tax benefits here, basically since the
[11:27] great financial crisis, small caps never
[11:30] do well. So, we're literally again
[11:32] seeing it in small caps as well. If you
[11:34] don't believe it on a small cap basis,
[11:36] then just go to MSCI and go to Aqui and
[11:39] just realize that it's not just here.
[11:41] The global earnings per share are going
[11:43] up and they're going up violently in
[11:45] here. The rest of the world is
[11:47] benefiting as well.
[11:50] I wrote this this week. I think this is
[11:52] an important thing to take. Originally,
[11:54] people wanted to ask if this was about
[11:56] Apple and whether it should be a
[11:58] positive or a negative. Uh I'm negative
[12:00] on Apple. I'm negative on uh all of the
[12:02] Mag 7 except for Tesla. Uh I'm negative
[12:06] and and Nvidia and I'm negative
[12:08] especially on the hyperscalers. So leave
[12:10] put Apple in with the hyperscalers and I
[12:12] think they fit into a different bucket.
[12:14] I'll talk about that later if their
[12:15] stock start stacks poorly um and it
[12:18] starts to look like people are price uh
[12:20] pricing in a multiple compression for
[12:22] them. But if this is another dot bubble,
[12:26] my answer is simple. Until Apple is
[12:28] overs supplied in Mac minis and Mac
[12:30] Studios, the bottlenecks are still here
[12:32] across the supply chain and demand is
[12:34] greater than supply. I spent a lot of
[12:37] time kind of thinking about this in my
[12:39] own experience. So, if you haven't read
[12:40] it, go read it. I think it's important
[12:42] just to have in the back of your mind. I
[12:44] do think Apple is a signal in the fact
[12:48] that they are a massive company. They
[12:51] can't get the memory and they obviously
[12:54] can get some, but they can't get Mac
[12:56] minis out. It just shows you how violent
[12:59] the demand was for the Agentic World
[13:01] when it took off. And as someone who
[13:02] right now would spend a lot of money
[13:04] buying more Mac Studios, I can't do it.
[13:08] This is the driving factor and this is
[13:09] the only thing that matters. Tokens
[13:12] being processed
[13:14] driving the need for more data centers
[13:16] for cloud revenue. It's driving more VA
[13:19] revenues for a uh anthropic and for AR
[13:22] as long as this is going higher for
[13:24] Google.
[13:26] This is a massive part of it. So you
[13:28] have to remember monthly tokens
[13:31] processed across their surfaces. And
[13:34] just think about it. This is not all the
[13:35] tokens in the world, but as long as
[13:37] Google stuff is growing like this, Azure
[13:40] stuff is growing like this, uh, for for
[13:42] Microsoft, if you're seeing the same
[13:44] thing for Amazon,
[13:47] you end up in a situation right now
[13:48] where you just have to remember that all
[13:50] of this stuff going on, tokens are food
[13:53] and money for AI agents. This is a
[13:56] representation of AI agents. We only got
[13:58] into the agentic world here and it's
[14:01] barely started at this point. I'm, you
[14:03] know, I mentioned in September I'll
[14:04] start bringing out a crypto side of
[14:06] this. I'm spending a lot of time on the
[14:08] crypto crypto ecosystem because we
[14:10] haven't seen the agentic side show up.
[14:12] Stripe has talked about the fact that
[14:13] the numbers are growing, but we're at
[14:14] such a small level until we hit that
[14:16] inflection point. I think the crypto
[14:18] world is going to continue to have this
[14:20] where the agents the agents are not
[14:21] there for consumption yet. They are
[14:23] definitely there for coding. That is the
[14:25] next stage. Uh way Lee put this out
[14:27] today again. Same thing. Don't be fooled
[14:30] by falling token per unit cost. So
[14:32] remember that chart. The relevant metric
[14:34] is per task cost and that's moving
[14:37] higher on the exploding agentic world.
[14:40] Final five, six slides here. Start
[14:43] understanding what loops are. So I want
[14:47] to make sure that we're always ahead of
[14:49] the game with my subscribers on what the
[14:52] next conversation is. Are loops the next
[14:55] hype cycle? The questioner asks, "Are
[14:57] they for real?" Boris Churnney from
[15:00] Anthropic answer was emphatic. Yes, they
[15:03] are for real.
[15:06] If you want to go learn more about it,
[15:07] this is where he said it. This was on uh
[15:09] at Meta Scale live fireside chat with
[15:13] Boris. You can go read it. This part in
[15:16] particular, this two and a half minutes,
[15:18] he spends times on loops as the next
[15:20] abstraction layer after agents. I'll let
[15:23] you read this, but this gets through the
[15:25] importance, but just think of them as
[15:28] working around the clock to do things
[15:30] where you don't need to be there
[15:31] monitoring them and watching and making
[15:33] sure you're hitting a button and going
[15:35] through it, but we're actually there.
[15:38] Agents make individual tasks faster, but
[15:40] loops make workflows faster. This is
[15:43] where the productivity move from
[15:45] incremental games to nonlinear. This is
[15:48] when you're going to start to see
[15:49] margins. is also when you're going to
[15:50] start to see the job impact come. You're
[15:53] also going to see token usage lately
[15:55] explode likely explodes with loops. One
[15:57] other thing on the loops before I read
[15:59] this. This is where power users. So, I
[16:02] keep putting these things out to teach
[16:04] you guys how to build the knowledge
[16:05] brain. Some of you are doing it. I'm
[16:07] getting phenomenal responses. Yesterday,
[16:08] I had a few people that reached out. As
[16:11] you do it and you get response, let me
[16:14] know. I really want to start keeping
[16:15] track of people that are doing it. the
[16:17] agency side of what I'm trying to bring
[16:19] is that you have to become a power user
[16:22] because if not your job will be replaced
[16:24] more likely by not being a power user
[16:27] because what these loops are doing is
[16:29] the power users are showing them how to
[16:31] use the computer with inside the context
[16:33] of an office. So if the person next to
[16:36] you is doing anything similar to what
[16:37] you're doing and they're using AI agents
[16:40] and teaching them how to do your job,
[16:45] then eventually the agent will be doing
[16:46] your job.
[16:48] This is why loops are so important in
[16:50] the AI infrastructure thesis. If every
[16:52] workflow becomes a loop, token demand no
[16:54] longer scales only with the number of
[16:56] users. It scales with the number of
[16:57] recurring processes those users
[17:00] automate. This is where agents start
[17:02] replacing the workforce. Remember,
[17:06] we're inside a consolidation. We're
[17:08] inside a rate of change. We're inside a
[17:09] midcycle slowdown to help you guys. That
[17:12] is why I created the pullback hunter.
[17:16] You can get that on the payw wall. And
[17:18] that's also where the AI consolidation
[17:21] playbook, what that sheet is supposed to
[17:23] do. All right, little update for you.
[17:25] Make sure you're not sitting there
[17:26] panicking as all this stuff is going on.
[17:29] Don't be surprised if a month from now
[17:31] we still have not moved out of this
[17:33] range. I really do believe the earning
[17:35] season bar is going to be much higher.
[17:37] The reaction might is probably going to
[17:39] be far more negative than it was in in
[17:42] Q1, but only on a rate of change basis.
[17:44] So navigate the AI made slowdown. I'll
[17:47] see you on the weekend.