Hyperscalers Are Collapsing. Is the AI “Bubble” Crashing or Consolidating?

[0:00] I'm in Maine. Um,
[0:03] finished up a week.
[0:05] June has been a uh a wild month and I'm
[0:09] going to take you guys through it. Uh,
[0:11] the rotation continues. Uh, just a
[0:14] brutal month for the hyperscalers, the
[0:15] Mag 7 included. Uh, we're going through
[0:19] what I believe uh people have to adapt
[0:22] to, which is a midcycle slowdown. Uh, I
[0:24] don't think it's over and I think the
[0:26] earnings period is going to be very
[0:27] different than the first quarter
[0:29] earnings period. Uh, and Mike Ron was a
[0:31] preview of that with the earnings and
[0:33] the commentary explosive. Let's go see
[0:36] what the stock did for the week. Uh, I'm
[0:39] going to cover loops and tags. As the AI
[0:42] progression continues in terms of token
[0:44] usage, I wrote a paper this week on the
[0:46] AI mindset. I'm going to go through
[0:48] that. Uh, healthcare and the biotech
[0:51] renaissance. I've talked about Eli Liy.
[0:53] I've talked about biotech. Um, more news
[0:55] on that stuff this week. Again, I'm
[0:57] going to keep emphasizing the midcycle
[1:00] slowdown means the bottom three layers
[1:03] of the five layer cake to me are built
[1:07] in. And I don't mean the growth rate's
[1:09] going to stop, but I do believe that the
[1:12] uh the ability to just sit in things and
[1:15] think you're going to get a five and 10
[1:16] bagger is over. Uh, so it gets harder in
[1:19] the application stage is where I think
[1:20] you're going to start to see uh better
[1:22] sharp ratios. Uh, Stripe Sessions,
[1:26] I I referenced it last year. Um, it
[1:28] happened about a month ago. I finally
[1:30] got the chance to spend the time as it
[1:32] was put up on YouTube going through it
[1:34] and I'm going to connect it to the
[1:36] Agentic Commerce and you guys are going
[1:37] to hear me keep talking more about this
[1:39] in the middle of as more people reach
[1:42] out and say the crypto winter is here.
[1:44] for all of you crypto people. I continue
[1:47] to focus my attention on crypto uh for
[1:50] next year with a bottom happening
[1:53] anytime but we have no signs of it yet.
[1:56] So rather than sit there and try to pick
[1:57] the bottom we are in a bare market but
[2:00] let's go through the fundamentals behind
[2:03] it. So for the week S&P was down 2% but
[2:08] this is not a uh a fall.
[2:11] Healthcare up 8% real estate up four
[2:14] utilities I don't have to go read them
[2:15] all you can see what's on the bottom
[2:18] that is basically the definition of the
[2:20] mag 7 consumer discretionary info tech
[2:23] commercial services
[2:25] uh rotation rotation uh Q's down almost
[2:31] 5% and again I said this back in this
[2:34] week now we've had the third or second
[2:37] worst week beating the one from the
[2:39] beginning of the month So
[2:42] Q's are starting to feel a little bit at
[2:45] the same time that that's going on
[2:46] surprisingly as Beta is getting crushed.
[2:50] IWM continues to outperform up one and a
[2:53] half%. So yes, in the same way that we
[2:55] had a big up big up week here on the
[2:58] bounceback. This week we've had our
[3:00] third week in a row. But if you take it
[3:02] back further and you go from early
[3:04] April, which again you would have
[3:07] thought that oil would have had a bigger
[3:08] impact on small caps, but everything is
[3:11] good in small cap land. Inflation
[3:14] expectations seen through the swap rates
[3:17] um continue to fall as
[3:20] getting banks saying three rate hikes.
[3:23] And I I'm just going to keep
[3:25] emphasizing, guys, the more you focus on
[3:26] the macro and stupid things like what
[3:29] the Fed chair is going to do from an
[3:30] investment standpoint related to AI,
[3:33] every time that you migrate into this
[3:35] area of this, if we were going to have
[3:37] Fed rate hikes, would IWM be
[3:39] outperforming right now? Um, again, it's
[3:42] not that it won't matter to some degree
[3:45] for a period of a week or two. It
[3:47] obviously has had an impact on on
[3:49] capitulation in the debasement trade. Uh
[3:52] the dollar has continued to rally, but
[3:54] here's what's going on in inflation
[3:56] expectations. And again, we've taken out
[3:59] the lows of last year on the one-year. I
[4:02] mean, again, I said it before. I thought
[4:04] we'd be having as this went higher on
[4:06] the back that people were
[4:07] underestimating the risk from the
[4:09] straight of four moves. I listened to a
[4:11] lot of the people that went through.
[4:13] Didn't think it would ever get to the
[4:14] point that they did, but definitely
[4:15] thought we'd be seeing in the second
[4:16] half of the year. But here are the
[4:18] expectations coming down. So whether
[4:20] it's part of the the basement trade or
[4:22] not doesn't really matter. And it's not
[4:24] just swap rates. Here's tenure rates. I
[4:26] mean again, we have a raging economy. We
[4:28] have employment that's stronger than
[4:30] expected. We still have uh CPI above 4%.
[4:33] We have the PCE core coming out. And yet
[4:36] somehow or another, we are lower than we
[4:38] were before the evil Kevin Worsh spoke
[4:41] and was super hawkish. Two-year rates
[4:43] have given up the entire move. So every
[4:45] one of those macro podcasts you listen
[4:47] to saying here it is. It was a hawkish.
[4:49] It was nothing. Absolutely nothing.
[4:53] Here's the important stuff. Hyperscalers
[4:54] down 14% monthto date. This is across
[4:58] the board, guys. I'm not going to go
[5:00] through a lot of news this week because
[5:02] I'm going to try to keep this relatively
[5:03] around 45 minutes. But you got a lot of
[5:06] stuff going on. These are big moves
[5:08] monthto date. I mean, these are big. And
[5:10] I've told you I want to be on the short
[5:12] side of hyperscalers relative to the
[5:15] thematic AI trade. Um, here's what's the
[5:18] reality. Profit margins uh for the MAG
[5:22] 8. This includes Netflix, does not
[5:24] include uh Broadcom, but I mean this is
[5:28] not happening for fundamental reasons.
[5:30] This is a rotation. These are the
[5:32] spenders. You have to lump Nvidia in.
[5:35] Apple obviously had bad news this week.
[5:37] I wrote a paper on Apple. Tesla hasn't
[5:40] been able to go out. People reach out to
[5:41] me on Tesla. I still love Tesla, but the
[5:43] problem is right now any company with
[5:47] high pees is under pressure. This is not
[5:50] a market to find high pees. And
[5:52] especially when Tesla, like SpaceX, like
[5:55] Bitcoin, like all of this, these things
[5:58] need retail traders. They need retail
[6:00] energy. And if retail energy is finding
[6:03] money from the math in the market, which
[6:05] is through the earnings coming through
[6:06] on the AI trades, and you're getting
[6:09] four to 10 baggers over a one-year basis
[6:11] for I mean, it's shocking. We've we've
[6:14] had more than a doubling in 10% of the
[6:16] S&P over the last uh year and and four
[6:19] months. So, they haven't needed
[6:21] anything. When the mega caps are going
[6:23] down to me, this is a a representation
[6:26] of what I've shown many many weeks,
[6:29] which is I call it the benchmark
[6:31] arbitrage. It's a reweing away from
[6:33] these high multiple names. Nvidia is not
[6:36] a high multiple name, but the rest of
[6:37] these for the most part started the year
[6:39] that way. So, as their earnings and
[6:41] their profit margins continue to grow,
[6:43] as they lay off people,
[6:46] their stocks are getting pounded because
[6:48] they're spending lots of money. I wanted
[6:50] to just highlight this because if you go
[6:52] back to the end of the great financial
[6:53] crisis and take March and you go through
[6:56] and look at a total realized annualized
[6:58] return for the S&P, you're talking about
[7:02] 14.7%.
[7:06] Look where we are through the halfway
[7:07] mark this year at the exact same pace.
[7:10] Yet this is called a bubble. So of the
[7:13] last 15 years, we are going exactly at
[7:16] the pace that we have over that time
[7:18] period. Now here's the Nikkei up 38%.
[7:22] Here's the Cosby up 100%. Here's the TX
[7:25] up 53. Even Europe is equal to the US.
[7:32] I I bring this up because if you're not
[7:34] involved in these from an AI
[7:35] perspective, because that's what's
[7:36] driving these, they just have a higher
[7:38] benchmark waiting towards AI than we do.
[7:42] Our spenders are fueling this the same
[7:45] way that if you broke down the S&P and
[7:47] went to the AI names or my thematic
[7:49] portfolio, my thematic portfolio is up
[7:51] around this mount for the TX. So, what's
[7:55] happening is the receivers of the money
[7:57] are driving everything,
[8:00] navigating the AI midcycle slowdown. I
[8:02] did a webinar this week uh for the
[8:05] subscribers just to make sure that
[8:07] people focus on getting used to this. I
[8:10] went through a bunch of things. I'm
[8:11] going to go through some more now. Uh
[8:13] but just remember
[8:15] this is what I wrote. This is this was
[8:17] for the the video update. So, this is
[8:19] what I did this week. Again, for the
[8:20] subscribers, if you didn't see it, it is
[8:22] on the website. But again, I go through
[8:24] a lot of different things regarding um
[8:27] intra week to make sure that people
[8:29] don't panic and don't just sit there and
[8:31] start capitulating all over things. And
[8:33] then I remind you that okay, I wrote
[8:35] that the fireworks show was over on June
[8:37] 4th. The AI midcycle slowdown, building
[8:40] intelligent there, but in here in macro,
[8:42] a midcycle slowdown is the pause inside
[8:44] an expansion growth is still positive.
[8:48] It's just that the rate of change comes
[8:51] down.
[8:54] You've got all of this stuff. So, the
[8:55] second that AI starts to slow down, and
[8:57] I mean the second, I have no idea how to
[9:00] describe to you how many people that I
[9:02] communicate with
[9:04] from from my world that send me things
[9:08] because they know I'm focused on it and
[9:09] they're looking to see if I agree with
[9:11] them. And it's all of this bubble stuff.
[9:13] Things are going to crack. Things are
[9:14] going to go down. We are in the middle
[9:16] of a very important month, guys. June is
[9:18] a massive expiration month. Um, you've
[9:21] had that. You've also got the pension
[9:23] rebalancings at a time where stocks have
[9:25] massively outperformed
[9:27] bonds. And at the same time, you've had
[9:29] the AI names completely outperformed.
[9:32] So, it's not shocking that there's
[9:34] cracks. But to say it seems more
[9:37] serious, again, the midcycle slowdown
[9:39] should bring this out. Valuations of AI
[9:40] related stocks are now high relative to
[9:42] our optimistic views. Okay, so people
[9:44] are in it. part of the AI. That's what
[9:46] goes on with the stock market when you
[9:49] have a serious growth phase. 60% of
[9:52] companies now watching AI budgets are
[9:53] moving to cheaper models. Talk about
[9:55] this every week. I continue to believe
[9:57] it's there. And every smart person that
[9:58] is inside AI keeps saying the same
[10:00] thing, which is eventually most
[10:02] companies will be using open source or
[10:04] the cheaper models for the majority of
[10:06] their compute. In some numbers, I've
[10:08] heard 99%. If you haven't thought about
[10:10] that, you don't need the most expensive
[10:13] models to do the hardest work. We are in
[10:15] the very early innings, the very early
[10:18] innings of figuring out what's going on
[10:19] with um how to use AI for a cheap way.
[10:22] So, I do believe that you're going to
[10:24] see more and more of the token index
[10:27] weakness type thing. And like I said on
[10:29] to the subscribers this week, I really
[10:31] do believe that token index has already
[10:32] peaked. So, you're going to be looking
[10:34] at a chart. Best case scenario looks
[10:35] like 10-year yields where it just kind
[10:37] of goes sideways. Worst case scenario is
[10:39] it starts going through the deflationary
[10:40] pressure that makes sense from AI. That
[10:43] doesn't mean that Jevans paradox isn't
[10:45] alive and well and the compute numbers
[10:47] don't go through the roof. I'll go
[10:48] through some of that because we have
[10:50] more agentic activity going on.
[10:53] Bernstein put this out and again I'm
[10:55] only bringing this up because this is
[10:57] your second derivative. Now they don't
[11:00] have this thing peaking until sometime
[11:02] in 2027. When I go through the micron
[11:04] earnings you can see it. But the main
[11:05] point is look at the explosion in terms
[11:09] of the earnings and then obviously the
[11:12] forecast.
[11:13] When this slows down, it should mean
[11:16] that the stocks start stop going up at
[11:18] the same price. Now, for confirmation,
[11:21] here's my thematic portfolio.
[11:24] This is the 30-day rate of change. So,
[11:26] look, it peaked in May. Here's the
[11:28] 50-day rate of change. It peaked early
[11:30] in June. They're collapsing. All that
[11:33] means is the growth rate is slowing down
[11:34] in the thematic portfolio. We're
[11:36] consolidating
[11:38] and as I go through that should be
[11:40] expected with everything with Micron
[11:42] massive blowout of earnings. You got
[11:44] them doing a deal with Enthropic all
[11:47] kinds of stuff and the stock finished
[11:50] down for the week. Lots of activity wide
[11:53] range blah blah blah outside the prior
[11:55] week's range. I mean, we're talking
[11:56] about massive stuff. Like, when you go
[11:58] from,300
[12:00] to 1,300 points and basically we were
[12:02] stuck in a a 40point range last year
[12:05] doing nothing. I'm just telling you that
[12:08] again, Micron is a symbol of that
[12:09] slowdown in the rate of change. Here's
[12:11] another thing. Uh, for those of you,
[12:13] again, I highly recommend doing this
[12:15] type of stuff. I will do a video very
[12:17] shortly on how to use these Excel files
[12:20] using Claude Co-work. Um, and I'll give
[12:23] you guys the the prompts. I know I did
[12:26] this for
[12:27] uh
[12:29] uh something in terms of going through
[12:31] it, but here you go. The amount of names
[12:33] above the 20-day with inside my thematic
[12:35] portfolio of the 100. We're now down to
[12:37] 44. We peaked May 8th. We were at 85%
[12:40] above the 50-day. We're now at 59 at the
[12:42] end of the week. The 200 day still rock
[12:46] solid. The up 50-day uh uh up slope. We
[12:51] had another one turn another two turn
[12:53] positive this week. were actually at
[12:54] highs and this is there. These are the
[12:56] things that matter for a bull market. So
[12:57] again, we're in a bull market. We're in
[12:59] a consolidation. We need to have them.
[13:02] What is a consolidation? For those of
[13:03] you who
[13:05] have never really thought about a
[13:07] consolidation with inside a bull market,
[13:08] a consolidation is the market's way of
[13:10] making a worker trend harder to own. It
[13:13] resets successes through some
[13:14] combination of price time and
[13:15] volatility, shaking out weak hands and
[13:17] preserving the larger trends. Now, I put
[13:18] up three separate ones here. I don't
[13:21] think we're going to go through this
[13:22] one. So a sideways one usually depends
[13:25] on something that is moving at a very
[13:27] slow pace. This we could end up in
[13:30] something like this. Um I think this
[13:33] level here would be around the uh 50-day
[13:35] for the index. We're still well above
[13:37] it. Uh but it could be more the 100 day.
[13:39] Whatever the case is, you could get this
[13:41] kind of slowdown. This is the one I
[13:43] think we're in and this is the one that
[13:45] has gone on. So that would mean that
[13:47] volatility in time. So you have to
[13:49] remember that consolidations are
[13:50] allowing the moving averages to catch up
[13:53] to where the prices are. That should
[13:55] happen when the growth rate or the
[13:57] second derivative in terms of earnings
[13:58] and all of that stuff starts to slow
[14:00] down. For confirmation of the fact that
[14:02] we're in the third one, here is the
[14:04] volatility again for tech momentum. This
[14:06] is the 60-day V. The Morgan Stanley tech
[14:09] momentum. It just continues to move
[14:11] higher. So at an 80 V, we're talking
[14:14] about 5% a day is the normal move.
[14:18] I mean, you're talking about big moves.
[14:20] Now, everyone can sit here and say,
[14:22] "Well, this is like the.com bubble."
[14:24] This is not like the.com bubble.
[14:28] Here's industrial momentum, which is the
[14:31] other big part of the AI trade in my
[14:33] thematic portfolio. So, the industrial
[14:35] momentum 60-day vault is also getting up
[14:37] to levels only associated with COVID. We
[14:39] don't have this going back. Morgan
[14:40] certainly only has that from there. You
[14:42] can see the upward trend in this that
[14:44] has basically been going on since
[14:46] Caterpillar and Vertive and all of these
[14:48] different components that are more
[14:50] related to the industrial side started
[14:51] to trend higher. So here's the chart on
[14:55] the thematic portfolio and again since
[14:57] mid May when the 30-day rate of change,
[15:00] you can see that we went from a period
[15:01] of just trending higher. This is a
[15:03] consolidation guys. That's what this is.
[15:05] This isn't some big top forming. We keep
[15:08] making higher highs and higher lows.
[15:10] We'll see. Here's the 50-day right here.
[15:12] It's down from where we were. That's
[15:14] another what 3%. Uh we'll see what
[15:18] happens, but we've done some work. Um
[15:22] again, 50-day here, you could look at it
[15:24] on this. The reason uh I wanted to look
[15:27] at these at different stages is you
[15:29] could pick your poison in terms of the
[15:30] correction. If you want to focus on just
[15:33] this year, this is when Opus 4.5 came
[15:35] out. uh 50% retracement would take us
[15:38] down to here, which would also be the
[15:39] lows in May. I could see that as well.
[15:42] If you want to look at it more from
[15:43] where it was from uh uh the bottom of
[15:46] last year after the tariff fears, you
[15:49] would think a bigger correction, but I
[15:51] don't think people jumped in. Based on
[15:53] my conversations, I don't think people
[15:55] accepted DRAM as a bull market until
[15:58] September, October. So, I'm using Opus
[16:00] 4.5 as the the true point to look for to
[16:04] use for retracements. Um, if you
[16:06] followed my advice and you're long my
[16:10] thematic portfolio relative to the
[16:12] hyperscalers,
[16:15] this isn't a consolidation. This
[16:16] continues to be a bull market. So, for
[16:18] the week it was up again and again, the
[16:20] hyperscalers are the spenders.
[16:22] Here's the equal weight S&P. So, with
[16:25] everyone freaking out, me getting stuff
[16:27] saying I'm now net short, I believe the
[16:29] market's going to collapse, blah blah
[16:31] blah, here's the equal weight S&P. This
[16:33] is not about anything more than the
[16:35] hyperscalers going down. And yes, there
[16:38] are retail that is heavily involved in
[16:41] Taiwan and Korea and in the US in the AI
[16:44] trade just like silver, gold, and
[16:46] palenteer a year ago and crypto. Those
[16:49] things continue to be in a quote unquote
[16:52] bare market. The debasement trade is
[16:54] unwinding. We just keep getting these
[16:57] rolling rolling rotations with inside an
[17:00] earnings driven bull market. This is not
[17:02] a speculation driven bull market. So how
[17:04] do you deal with retail? Retail is very
[17:07] sophisticated. You could burn them out
[17:08] by having this volatile chop and they
[17:11] start looking for the next good sharp
[17:13] ratio. Um equal weight is working.
[17:17] New York Stock Exchange breath same
[17:18] thing right at all-time highs. I showed
[17:22] this this week. I just want to make sure
[17:23] it does not pay to get mega bearish as I
[17:27] heard people reach out this week. I mean
[17:29] I I just here is the forward EPS. So you
[17:32] need earnings to come down plain and
[17:34] simple because the stock price is not
[17:38] going up relative to how much earnings
[17:40] are going up. So earnings have gone
[17:42] parabolic. The S&P is up 7.4% year to
[17:44] date. That's because of the
[17:47] hyperscalers. So this is not some like
[17:50] raging bull market. This is just
[17:52] earnings are going higher. So you're
[17:54] going to have to forecast when earnings
[17:55] go down. Two things on this chart. It's
[17:57] very hard for earnings to go down.
[17:58] Second thing is you don't do really well
[18:00] when you're shorting the stock market at
[18:03] these points when we've been here. These
[18:05] have been coming out of recession
[18:06] periods. I don't see how this is going
[18:08] to turn down. I do believe the growth
[18:10] rate is going to slow at some point
[18:12] because a lot of this has to do with
[18:14] Micron and SanDisk and Western Digital
[18:17] and all the semis and maybe this won't
[18:19] be as good but I don't know when that is
[18:21] going to
[18:23] turn down.
[18:25] Remember the earnings surprise blew off.
[18:27] I mean, again, these are all facts.
[18:28] These are not speculation. If you don't
[18:30] believe it and you think this is all the
[18:33] two names in the S&P or whatever garbage
[18:35] people send out to scare you, here is it
[18:38] for the S&P 600. Again, we had a bare
[18:41] market, unchanged earnings, margins,
[18:45] everything for small caps for for from
[18:48] the time Chat GPT came out until the end
[18:51] of last year. And now it's all going
[18:53] higher. If you don't believe in it,
[18:56] you're missing it. And here it is for
[18:58] the globe. This is global EPS for 2027.
[19:04] Here it is for the information
[19:05] technology side. Longterm expected
[19:08] long-term earnings growth soared to 38%.
[19:11] Again, you can fade this. You can go
[19:13] back and say, "Yep, look what happened
[19:14] here. This was it, too." So, fine. This
[19:17] needs to come down. We need revisions to
[19:19] come down. We don't have any of this
[19:21] yet. here is how much semis have gone up
[19:23] relative to their earnings looks
[19:25] justified to me. It's all about this. So
[19:29] if you guys think that tokens are going
[19:31] to come down or the price of tokens is
[19:33] magically going to all of a sudden
[19:35] decrease dramatically and that
[19:36] everyone's going to move to Chinese open
[19:38] source models which oh by the way is not
[19:40] possible um unless you're using the
[19:44] garbage models. Um, as someone who would
[19:46] love to, I wrote a paper about this,
[19:48] which I'll get into, as kind of the
[19:49] canary in the coal mine for when you
[19:51] should start thinking that token prices
[19:52] are going to start collapsing at a
[19:54] really fast pace. We don't have enough
[19:57] memory for that right now. So, the token
[19:59] needs, the demand just continue to grow
[20:01] exponentially. Quadrillion Q trillion
[20:05] back here. These numbers were not
[20:07] expected. I wrote a paper I talk about
[20:09] this back here about the inference
[20:10] needs. We are accelerating so fast that
[20:12] this number is just going to continue to
[20:13] go. That is what's driving the earnings
[20:16] is that this is the food that is
[20:18] necessary and if you don't like the
[20:20] accounting the capback spenders don't
[20:23] need they can move their spending over
[20:25] the next five years to give them time
[20:27] for the revenues to come in. That is a
[20:29] different bet. I'm not sure they're ever
[20:31] going to get the revenues on the
[20:32] timeline that they need which is why the
[20:34] multiple compression is real. So
[20:36] multiple compression is an uncertainty
[20:38] over their ability to get their revenues
[20:40] in the door while their profit margins
[20:41] are still growing. Think about it. It's
[20:43] being built into the market. All those
[20:45] risks that you think are happening,
[20:46] they're being built in through the
[20:48] hyperscalers. If you don't believe in
[20:51] what is coming, forget what has
[20:53] happened. Everyone else has to build
[20:55] their own AI data centers. They need
[20:58] chips. Japan just announced a $2.3
[21:01] trillion investment plan. Japan is about
[21:02] 3% of global GDP. So, multiply that by
[21:06] three, you're up to 70 trillion.
[21:09] I I you have to think about this from
[21:14] this. We get confirmation of this every
[21:17] day. Why would you bother fading this?
[21:20] You don't have to invest in all them
[21:21] because at some point I do believe that
[21:23] the buildout will not be profitable for
[21:26] the same companies, the applications and
[21:29] all this stuff. There's going to be
[21:30] tremendous competition. We'll build this
[21:32] into the pees at some point ahead way
[21:34] ahead of when it is. If you looked at
[21:36] the micron chart, the scary thing about
[21:37] it is once it starts to go down the
[21:39] other direction, you're talking about a
[21:41] big fall-off. The market will not have
[21:42] the right PE at that point. I feel very
[21:44] confident about that. So, I wrote this.
[21:46] This is my way of saying to you guys as
[21:49] a user who would love to use the Chinese
[21:53] models. I would love to use GLM 5.2. I
[21:56] would I would love to use them, play
[21:58] around with them. For anyone who does
[21:59] coding, they would love to use them. The
[22:01] problem is there are no Mac minis or Mac
[22:04] studios guys. There aren't any. Go read
[22:06] this report about why Apple is such an
[22:09] important thing. Why their architecture
[22:11] is so important because to for me to use
[22:14] GLM 5.2 which as I show is very close to
[22:17] the best models that's a lot of money. A
[22:21] lot again Mac Studios again sold out on
[22:25] Apple store except for the smallest
[22:27] models which won't be able to work. The
[22:29] 512
[22:31] configuration was pulled earlier this
[22:33] year. Again, you could have had as many
[22:35] as you wanted in January. This all
[22:37] changed in March or yes, I mean, it
[22:40] started to change in February. I bought
[22:41] my last load in February. That was when
[22:43] OpenClaw came out and I wanted to go run
[22:46] it. I did not expect the model changes
[22:49] to happen this fast. So, the model
[22:51] capabilities have improved dramatically
[22:53] before the hardware can. Why Micron's
[22:55] earning blowout are a headache for
[22:56] Apple. you started to see the price
[22:58] rises again. At some point, this will
[22:59] change. But if Apple, one of the most
[23:02] and until Nvidia, the largest and most
[23:05] important supply chain groups that
[23:07] should have been able to secure memory
[23:09] because of their orders for everything,
[23:11] they're in the queue.
[23:14] This is the deal that Micron announced
[23:16] with Anthropic earlier in the week.
[23:17] These are all just confirmations that we
[23:19] still don't have anywhere enough enough
[23:22] memory and we don't have enough memory
[23:24] makers and there's no way to just as I
[23:27] like to say if you have four billion
[23:29] extra people show up on the planet
[23:30] tomorrow you can try to grow as much
[23:32] food as you want you just don't have
[23:34] enough of the supply chain to do it so
[23:35] they're earnings blow out with respect
[23:38] to supply our customers are recognizing
[23:40] that supply shortages in memory and
[23:42] storage will take considerable time to
[23:43] improve even as we expect industry
[23:45] supply to improve gradually in 2028,
[23:48] 2028. We're in 2026. We currently do not
[23:51] have line of sight as to when memory
[23:52] supply will be able to catch up with
[23:54] increasing demand. There's no way to
[23:56] know this. Demand is moving too fast. I
[23:59] continue to say that whatever timeline
[24:01] people had for memory is best reflected
[24:03] by Andre Carpathy on Dwar Patel's
[24:07] interview in October. Go listen to it
[24:09] today. Go listen to the Dwar Patel and
[24:11] listen to what Andre Carpathy said about
[24:14] AI agents. He was so wrong. So wrong in
[24:18] terms of how soon they would get here
[24:20] and start to impact things. They are
[24:22] here and he writes about it
[24:23] consistently. He was wrong by years. If
[24:27] he was wrong as a power user who uses it
[24:29] all today, how the hell would they have
[24:31] the supply and the demand just keeps
[24:33] increasing because now we have RSI
[24:36] happening. So we have recursive
[24:38] self-improvement which means the models
[24:39] are improving their own through the
[24:41] memory through the efficiency side which
[24:43] means more people want it because the
[24:45] models are getting better but we don't
[24:47] have the memory to run the hardware to
[24:49] be able to do it. If you're again
[24:52] doubting here are the earnings. Now the
[24:56] majority of this is obviously price.
[24:58] When you have a shortage and you can't
[24:59] make enough you're just raising your
[25:01] prices on what available you have. And
[25:03] that's why eventually when supply does
[25:05] catch up demand, you're going to see a
[25:06] violent move lower in this as you saw in
[25:08] the Bernstein. But here's the reality.
[25:10] You're talking about the last three
[25:11] quarters. I mean, these numbers are 10
[25:13] years worth of earnings.
[25:17] These are important. You can go through
[25:18] I'm not going to read everything on your
[25:20] own, but the reason I wanted you guys to
[25:21] see this is because we are in the early
[25:25] innings of the significant innovation,
[25:27] productivity that can be unleashed by
[25:28] everything of the global economy over
[25:30] time. This is what Jensen Yuang says all
[25:31] the time. It's what I do all the time.
[25:33] It is mind-boggling to me that people
[25:35] don't realize when they think of this,
[25:37] they only look at the spending on the
[25:39] data centers and the ROC and the
[25:41] hyperscalers. Shorting the hyperscalers
[25:43] has worked. No need to sit there and go
[25:46] crazy except only on a relative basis.
[25:48] They're not getting pounded. Their
[25:49] profit margins are still growing.
[25:51] They're still laying people off or not
[25:52] hiring people. They're spending more
[25:54] money because the prices are going
[25:55] higher. And we have no idea when they're
[25:56] going to get revenues in, but they can
[25:58] raise endless amounts of capital. This
[26:00] isn't a bubble. These companies are not
[26:02] going out of business. No matter how bad
[26:04] this stuff is, at least in the short
[26:05] run, they have plenty of money to spend.
[26:08] We have behind this the smartphones,
[26:11] the high-end PCs, consumer devices,
[26:13] autos, industrial applications,
[26:15] robotics, robotics, and humanoids. These
[26:18] are going to use so much memory. So much
[26:21] memory.
[26:23] Even as we expect industry supply to
[26:25] improve gradually in 28, we currently do
[26:26] have line of sight as to when will be
[26:28] there. It gets back to the statement
[26:29] that he said,
[26:31] "All system performance is
[26:33] architecturally dependent on memory,
[26:34] subsystem performance, and capacity."
[26:36] Until this changes, you have to
[26:38] understand this is literally like with
[26:40] all of the food in the world, if every
[26:43] single person said, "You know what? You
[26:46] won't die if you eat beef." If everyone
[26:49] took all of their spending for food and
[26:52] it went to beef as opposed to everything
[26:55] else, we'd have no beef. So the question
[26:58] is how do we get this up? this is
[27:00] something new and that's what the
[27:02] argument is and you've reached a point
[27:04] where it's a strategic asset which means
[27:08] the government is going to be there to
[27:09] support building this stuff and you've
[27:11] heard Trump measure uh mention now
[27:13] Micron uh I think multiple times in the
[27:17] past month and the importance
[27:20] business model transformation if you're
[27:22] looking again to say this is not micron
[27:24] of the past this is not a se so whenever
[27:26] people go I've seen this before in
[27:27] memory eventually just cyclical and blah
[27:29] blah blah. We are pleased to announce
[27:31] that we've completed 16 structurally
[27:36] strategic customer agreements with
[27:37] customers across the data center,
[27:39] consumer, and auto. This is the lineup
[27:41] of long-term deals. They represent 20%
[27:44] of our volume. These are long-term deals
[27:47] that are being done.
[27:53] When completed, we expect approximately
[27:54] half or more of our company revenue to
[27:56] be under these sea SCAS.
[28:00] The SCAs are structured as take or pay
[28:02] agreements with binding commitments to
[28:04] purchase specific volumes over this
[28:06] multi-year term.
[28:08] I mean, isn't this like a subscription
[28:11] agreement? I like I just
[28:14] again if you're going to be bearish on
[28:16] these things these guys are not just
[28:18] telling you the value that's there but
[28:19] they're also telling you that they need
[28:21] more and more of this and there's no
[28:22] solution coming anytime soon because
[28:24] hardware takes big sol takes time um in
[28:29] terms of the automotive and robotics
[28:31] which again are not here yet but this
[28:33] will start and it's going to go fast and
[28:35] I've talked about this the mix of L2 and
[28:36] above vehicles is more than doubling
[28:38] this year to over 20% expected to exceed
[28:40] 40 that is what matters matters is that
[28:42] even though car purchases aren't going
[28:44] higher, the mix towards higher memory
[28:46] names is because you want an AI model.
[28:49] Humanoid robots carry 10 times the
[28:51] amount of memory as an average car.
[28:56] Here are the earnings. Again, you're at
[28:59] a point now for 2027 at seven and a half
[29:02] times earnings. If they only make what
[29:04] is forecast,
[29:07] I I can make the argument for people
[29:09] that the stock is fairly close to where
[29:12] it should be, but the likelihood of them
[29:14] not beating each time seems insane given
[29:17] the the trend. So that's why we're
[29:19] sitting there unchanged for the week. Um
[29:21] my whole point of getting out of Micron,
[29:23] which again for the pain point was
[29:25] getting out of here in the midst of this
[29:27] thing. um is that I thought I'd have
[29:30] better opportunity, which as of now has
[29:32] worked for names that I thought could
[29:34] more than double this, but if I would
[29:37] have
[29:38] been in it now, I doubt I'd be getting
[29:40] it rid of uh any more of it. So, that
[29:43] was a mistake. Um I just want to put
[29:46] this up again because I want you to go
[29:47] read the a the uh Mac Mini thing on
[29:51] Apple and just use it as a canary in the
[29:52] coal mine. And just remember with all
[29:54] the humanoids and all the things that I
[29:56] just mentioned for memory, you also have
[29:58] every country in the world that now
[30:00] realizes that they can't depend on the
[30:03] best models in the US. Sovereign AI is a
[30:06] necessity. Everyone in the globe needs
[30:08] to build out their own data centers and
[30:10] house their own AI.
[30:13] That's why this matters. Now, I continue
[30:16] to believe that we're moving into the
[30:18] application. So if we have a 7 and a
[30:20] half PE on micron and we no longer can
[30:22] expect a 20 bagger and we can't expect a
[30:24] 10bagger maybe not even a threebagger.
[30:28] Okay well then this is priced much
[30:30] better than it was back in January. And
[30:32] I think that's obvious to everyone that
[30:33] is the midcycle slowdown. I think the
[30:35] applications now when they start
[30:37] utilizing AI and we start seeing the
[30:38] benefits these will have a better sharp
[30:41] ratio. The problem with these is they
[30:43] are cyclical to some degree. So they're
[30:45] always going to have this commodity
[30:46] element in terms of the multiple for the
[30:48] application side. If you can find names
[30:50] that are trading with PEG ratios below
[30:52] one, say Eli Liy, uh, and you can go
[30:56] find these, you're going to be able to
[30:58] have a better time. The memory industry
[31:00] has been structurally transformed by the
[31:01] proliferation of AI. We were only in the
[31:03] early innings of significant innovation.
[31:05] I heard someone uh giving some lame
[31:08] thing this week uh on a video about how
[31:10] anyone saying we're still in the early
[31:12] innings is just wrong. We're in the
[31:15] early innings of AI. There's no doubt.
[31:17] In memory, I again I think we're way
[31:19] past the early innings, but I think we
[31:22] still are in the later innings, but the
[31:24] later innings are going to be more
[31:25] stressful as you get closer to the end.
[31:27] Data civer growth will increasingly
[31:29] complemented by AI enable features and
[31:31] all of this stuff. and then humanoids in
[31:34] there. Again, just reminder that these
[31:37] are all the things coming and that is
[31:38] why we are in a consolidation at this
[31:40] point because we've done all this. All
[31:43] right, Loopy. Um,
[31:47] highly recommend spending time on
[31:50] watching this video.
[31:52] Uh, sorry, this is a a news article on
[31:55] it, but loops are the next are are they
[31:57] the next hype cycle? Uh, this is Boris
[32:00] from Claude, the head of code. His
[32:02] answer was an emphatic yes, they are for
[32:04] real. Agent loops are now starting to
[32:07] dominate. This is the interview um I for
[32:09] the subscribers. I will send out the uh
[32:13] link and all the details. It was at a
[32:15] meta event. He describes loops as the
[32:18] next abstraction layer after agents.
[32:20] Agents repeatedly running tasks such as
[32:22] code review, reading feedback, opening
[32:24] PRs every few minutes. The jump from the
[32:27] fable, he says the jump from Opus 4.8 to
[32:30] fable feels as large as or larger than
[32:32] the leap to 4.5 which was a major leap.
[32:36] Um,
[32:38] think of loops as basically the ability
[32:41] to just keep running, repairing,
[32:44] debugging. You're taking the human out
[32:46] of
[32:48] the project. So basically on a workflow
[32:52] this becomes really important. And the
[32:54] reason this is important is this is
[32:55] where with loops now and the fact that
[32:57] we're at this point, you really can't
[32:59] replace many people in the workforce
[33:03] until you reach the loop part. So you
[33:04] need AI agents to start. And this is the
[33:06] thing I just want to make sure people
[33:07] go. In October of last year, there were
[33:10] no AI agents happening. So the first
[33:12] phase of the job situation for profit
[33:14] margins was actually not replacing the
[33:16] people. It was by not hiring new people.
[33:18] We've been doing that. That's why
[33:20] there's been no job creation especially
[33:22] for the places that are workflow places.
[33:25] So if you take out nurses and take out
[33:27] the health care side we've created no
[33:28] jobs over the course of the last 16
[33:30] months. I think people know that. But on
[33:32] the AI related stuff business services
[33:35] information they're negative. Now, I
[33:37] don't expect any kind of doom gloom
[33:40] scenario in jobs, but I do believe that
[33:42] when you get to the loop side, you have
[33:44] to realize if you have a 100 people in a
[33:46] marketing area and two of them are power
[33:48] users and they're using AI every day and
[33:49] the other 98 don't even care about it
[33:51] and they're just kind of pressing a
[33:52] button and doing whatever they're using
[33:54] Microsoft Copilot. What loops allow the
[33:56] two power users to do is completely
[33:58] teach an agent to basically go through
[34:02] and do their workflow. And then once
[34:04] they do it for them, if the other 98
[34:06] have similar workflows, then the agents
[34:09] are going to be doing the other 98. And
[34:10] guess who's not going to be around? It's
[34:12] going to be the other 98. This is where
[34:14] we are. This is where we start to get
[34:16] into the next phase of the profit
[34:18] margins. And I bring this up because if
[34:20] you guys are bearish on AI, these are
[34:22] all the things you're bearish on. Token
[34:25] usage likely explodes with Loops. So
[34:28] once Loops come, these are running all
[34:30] around the clock. The benefits start to
[34:32] come out.
[34:33] This is why loops are so important for
[34:36] the AI infrastructure thesis. If every
[34:38] workflow becomes a loop, token demand no
[34:40] longer scales only with the number of
[34:41] users. It scales with the number of
[34:43] recurring processes that you users
[34:45] automate. This is why again when you
[34:48] look at the token usage and you look at
[34:50] the ARR for anthropic, they're getting
[34:52] signed up everywhere because CIOS are
[34:54] chomping at the bit to have AI bring
[34:56] them ROIC.
[34:58] If you guys want to go read something,
[35:00] just go type this in. I'm going to read
[35:01] this. The term was not invented out of
[35:03] nowhere. In a single week of June of
[35:05] 2026, that's this month, several groups
[35:07] ran into the same thing at almost the
[35:09] same time. What set it off was a post by
[35:12] Peter Steinberger, famous of OpenClaw,
[35:15] which passed 8 million views. One should
[35:17] no longer be prompting coding agents,
[35:19] but designing the loops that prompt
[35:21] them. At nearly the same moment, Boris
[35:24] was saying the same thing. You have
[35:27] Andre Carpathy talking about loops. If
[35:30] you want to think about loop from a
[35:32] hedge fund perspective, if you're
[35:34] watching this, just think about all the
[35:36] quant strategies you have and what goes
[35:38] on. You can go read this loop
[35:39] engineering from a quant strategy. But
[35:41] quant trading is already a loop. You
[35:43] pull data, generate signals, you back
[35:45] test it, execute, monitor risk, repeat.
[35:48] The only difference is they need
[35:49] hundreds of humans sitting inside the
[35:51] loop. You do not. Okay? So, the
[35:54] competition for quant strategies just
[35:56] continues to intensify where you've
[35:58] needed the human more and more. You're
[35:59] not going to need the human more more
[36:02] Andre Carpathy put this out on one other
[36:04] thing which I'm not going to spend a lot
[36:06] of time on. This is introducing claude
[36:08] tag. Another disruptive thing to the
[36:11] workforce. Another heavy heavy thing on
[36:14] using agents. The agentic world is
[36:17] accelerating in front of our eyes. It
[36:19] has huge implications for profit margins
[36:22] and for token usage and we still do not
[36:24] have enough memory for them to run. Club
[36:26] basically joins the team in a serious
[36:28] way. You can talk to it as you would
[36:30] talk to a person. It can help. The best
[36:32] way to think about it without reading
[36:33] through all this. It's kind of like a
[36:35] loop. You're on Slack. You're talking
[36:37] with your team. An idea comes out of it
[36:39] and it just runs off and goes and builds
[36:41] it. Something I I view it as a data my
[36:44] data scientist taking my ideas and by
[36:46] the time I get in the next morning, it's
[36:47] already built and I didn't have to ask
[36:48] for it. It's just reading through the
[36:50] conversations. Again, token usage to
[36:53] explode. do not get bearish AI with
[36:55] token usage going up unless we're going
[36:58] to see rates massively go higher which
[37:00] is not happening. We're going to see
[37:01] inflation go higher because of all the
[37:02] memory needs which is not going on. The
[37:05] earnings are going to continue to grow
[37:07] as long as this is happening. Eventually
[37:09] we will hit a point that they won't be
[37:10] growing like this. I don't see it coming
[37:13] up anytime soon. For those of you who
[37:15] continue to use what I'm doing from an
[37:18] agency perspective, I wrote this this
[37:20] week. I put it on Substack. I got so
[37:22] many people that responded to it. I
[37:23] wanted to send it out to the subscriber
[37:25] side as well. Go read it. I go through
[37:28] the AI mindset. Give it to your kids.
[37:30] Um, literally, you're supporting what I
[37:33] do when you do this, but send it to
[37:35] everyone. People need to be learning AI.
[37:37] And if the only thing you get out of
[37:39] this is some sort of being on top of the
[37:41] news uh for AI, so you don't have to sit
[37:44] there and doubt it. So, it inspires you
[37:46] to try. At the same time, from an
[37:48] investment and trading perspective, you
[37:49] get the tools to help you make you
[37:51] better. I give you a cheat sheet of the
[37:52] list of the names that are directly
[37:54] impacted by AI and you get the agency
[37:56] side to change your brain to have an AI
[37:59] mindset and I go through it in this. It
[38:02] is a really cool thing but you need to
[38:04] change your brain. So the AI mindset
[38:07] becomes a loop. What I wanted to do now
[38:09] was take the mindset portion again. Make
[38:13] sure you read this because the paper
[38:15] goes into Andy Duke. It goes into my
[38:17] basian uh beliefs and it goes into
[38:19] Sherlock Holmes which you can think is
[38:21] quantico training and the loop is
[38:23] basically going to be what I do to
[38:25] create thematic portfolio. So how did I
[38:26] come up with the five separate themes?
[38:29] Many of you have seen this. This is the
[38:31] architecture that I put together where I
[38:33] would find a podcast with some dot or
[38:35] some marble as it used to be for green
[38:37] marbles. Some piece of information that
[38:39] I think you can make money on. I run a
[38:41] skill on it. Then it goes into deep
[38:43] research. Then it goes through a
[38:44] consolidator which is a mixture of
[38:45] experts with a judge approach. Then I go
[38:48] through the screening. The human element
[38:49] is in there. I've shown this before. But
[38:52] here it is now. And now I'm working with
[38:54] Claude on putting it in a loop. And what
[38:56] would happen? So now think about it. It
[38:59] goes through this process, but it just
[39:01] continues to run on its own. New
[39:03] information comes in on the same theme.
[39:05] What's going on in power? What should
[39:07] happen? Imagine if your entire process
[39:09] was being done for you. That's what I'm
[39:10] working on now. And this is a kind of
[39:13] visual on the way it would look so you
[39:14] guys can think about it. In the
[39:17] meantime, uh I put this up because I
[39:20] just realized you guys may want to use
[39:22] the system on your own. So for the next
[39:24] four weeks, uh on the subscriber line,
[39:27] I'm going to put it up. What this is
[39:28] going to give you is all of the prompts
[39:30] for each of these. So you guys can go do
[39:32] this on your own. But more importantly,
[39:33] if you want to go create your own loop,
[39:34] I'm giving you the information that you
[39:36] can go do it on your own. The reason I'm
[39:38] doing this is because of the outcry that
[39:40] I got from this. Outcry is the wrong
[39:42] word. The right word is you guys are
[39:45] amazing. Um, every day I get another
[39:48] thing on the success of people building
[39:50] a knowledge brain. For those of you who
[39:51] weren't able to do it and didn't reach
[39:53] out for you were different than other
[39:55] people. I sent many, many people and
[39:57] just said, "You're doing this wrong."
[39:59] Meaning, don't reach out to me. Take
[40:02] what I gave you here. Create the brain.
[40:04] I've had gamblers call me up and say
[40:06] they're using this for the people that
[40:08] they like for things. They're using it
[40:10] for um gambling on football games in
[40:13] terms of a coach and trying to stay on
[40:15] top of the change in what the coach is
[40:18] saying with regards to injuries. If you
[40:21] take and you really think about this,
[40:23] and I'm thinking about other ways to do
[40:25] it. People have created a jordy brain by
[40:27] taking all of my content off this
[40:29] website and putting it in. I'm working
[40:31] on doing that for people. So again, if
[40:33] you haven't done this yet, if you
[40:35] haven't got your kid at college a
[40:37] subscription to at least keep them up on
[40:39] all of this, you're doing people a
[40:41] disfavor. Th this is really, if you can
[40:44] do this, this is the one that I put out
[40:46] that has been a game changer in terms of
[40:48] seeing the empowerment that people have.
[40:51] Go build your own knowledge brain. I
[40:54] very seldom put this up here and I
[40:56] usually put it in the description, but
[40:57] so you guys can go. This is where you
[40:59] can go visit the website and see it. If
[41:00] you guys have any questions, Mark
[41:02] Whailing is the man. That's his email
[41:04] address. You guys can find him there.
[41:06] Now, in terms of nobody dying,
[41:10] Brian Johnson put out a tweet this week.
[41:12] Um, I didn't include it in this uh
[41:15] mainly because I forgot. I'm just
[41:16] reminded of as I bring this up. But
[41:17] Brian Johnson is someone that I I I've
[41:20] read. We all know he's a little wacky
[41:22] out there, but at the same point, he is
[41:24] doing some cool things for people who
[41:26] believe in both data, but also believe
[41:27] in longevity. and he's talking about
[41:29] people living forever the same way that
[41:31] moonshots is, the same way that I am.
[41:34] Darian Modi in a recent interview, I
[41:36] think biotech is about to have a
[41:37] renaissance driven by AI. My instinct is
[41:38] we're about to cure a lot of diseases.
[41:40] AI healthcare is going to be a massive
[41:42] theme. Now, there have been a lot of
[41:45] focus on people leaving Google,
[41:48] especially from deep mind focused on
[41:50] this point and moving over. Guys, we are
[41:54] very close when you get to RSI. We are
[41:56] very close at this. I'm telling you
[41:58] there is nothing that can be bigger than
[42:01] people not dying of cancer the people
[42:04] living for another 50. It changes macro.
[42:07] It changes everything. You have to think
[42:09] about this and we're seeing it in the
[42:11] market. The rotation from tech to
[42:12] healthcare signals a shift in market
[42:14] sentiment towards defensive sectors amid
[42:16] tech volatility boosting bioarmmer
[42:19] valuations. Now here's the chart we just
[42:23] consolidated for this year in biotech
[42:25] after a bull market. remember the cons
[42:27] consolidation side and again upward
[42:29] consolidation and now we're starting to
[42:31] move higher again and look at all these
[42:34] wicks people are still this is the way
[42:36] Eli Liy looked when it had broken out
[42:38] too oh to speak of Eli Liy which I've
[42:41] publicly said I will believe will be the
[42:43] largest company in the world within the
[42:46] next 5 years the reason is because I
[42:49] believe the GLP ones and now GLP 3es
[42:52] people are underestimating as I travel
[42:56] the country. This country needs GLP3s in
[42:59] a massive amount. And whether it's
[43:01] because of obesity or whether it's
[43:02] because of addiction, doesn't really
[43:05] matter. You're going to start to see
[43:07] more and more stuff. And these guys have
[43:09] spending money left and right to
[43:12] basically buy up companies, to buy up
[43:14] IP, to buy up AI. They already have
[43:16] their own data center. Eli Liy I'm a big
[43:19] fan of. This week, the president touts
[43:22] Eli Liy's factory in there. There's also
[43:25] speculation that Trump was taking uh
[43:28] their new GLP3 before it was released.
[43:30] He they've refused it. Um this is really
[43:33] cool. Deploys a weight loss cache on app
[43:35] store for scientists.
[43:38] Monero maker is collaborating with small
[43:39] biotech on AI as a tool for drug
[43:41] discovery. Again, I view this as the
[43:43] collaboration side within is going to
[43:45] speed up the process and the
[43:47] partnerships. Um for those of you who
[43:49] know my love of the Aura Ring, which I
[43:51] don't have on my finger. Oh, yeah, I do.
[43:53] I have it right here. See, there it is.
[43:55] This is not an advertisement for Aura
[43:57] Ring, but for those of you who do know,
[43:59] this is a big deal to me. I'm not going
[44:01] to go take you through it. It's a big
[44:03] deal because every week I started a
[44:05] second Substack at the beginning of the
[44:07] year with a focus on HRV because I had
[44:10] spent the last five years trying to get
[44:12] my HRV higher and was successful in
[44:14] getting it way away from my biological
[44:16] age. and the main or way away from my
[44:19] actual age to uh get it up at a point
[44:22] where
[44:24] I'm younger than I am. And so I changed
[44:26] it based on HRV. I do a weekly on this.
[44:29] You guys can go read it, but this is all
[44:31] through the aura ring. And so again, the
[44:33] application side to me is the big story
[44:36] here. Um just a shout out to Michael
[44:39] Seymble. I think for people looking to
[44:41] again go through the risks involved with
[44:44] the hyperscalers at this point. I view
[44:46] Frontier Lab projections of what they
[44:48] will be c when they will be cash flow
[44:49] positive as speculative, uncertain, and
[44:52] subject to revision. I don't disagree
[44:54] with any of this. So on all of you
[44:57] thinking that I'm permeable on AI, that
[45:00] is not the case. I do not know if
[45:02] anthropic is going to start to go
[45:04] sideways. I do not know if Open AI is
[45:06] just going to continue to go at this
[45:07] point and all the spending they're doing
[45:09] is going to end up with problems for
[45:10] them. I do believe that they are
[45:12] government necessities and the
[45:13] government will be there to support
[45:14] them. So it's not going to lead to any
[45:15] of the doom gloom stuff that people say
[45:18] if they reach this point. It is not
[45:20] because the AI is not getting better.
[45:21] It's not because their products are not
[45:23] needed. It's not because their products
[45:24] are not going to change the world
[45:25] through everything. So they will be
[45:28] fine. But when they get their money
[45:30] could be a valuation for their equity.
[45:32] And that's why I want to be on the
[45:33] spender side. Here's the hyper cloud
[45:35] revenue which is accelerating. But
[45:37] again, when is it going to get to the
[45:38] point where it's big enough? So we don't
[45:41] know.
[45:44] Um,
[45:46] I wanted to make sure that I'm putting
[45:47] up the things that are going to remain
[45:49] headwinds in this. So, you've got more
[45:51] people talking about the fact that
[45:53] cheaper ones will be used. You've got
[45:56] the fact that there's GLM 5.2, which is
[45:59] not that far off, but look at the cost
[46:00] differential between Fable 5. You're
[46:03] talking about a 90% cheaper model, but
[46:06] again, you need to go buy the hardware,
[46:08] and the hardware is very expensive. uh
[46:10] if you want to keep it on a closed
[46:12] computer as opposed to giving the
[46:14] Chinese companies all your data. Um the
[46:17] importance of this story with the
[46:19] revenues is big because again they are
[46:20] dominating the economy. I don't care
[46:23] what any economist says to you that this
[46:25] doesn't matter. Obviously, when
[46:27] twothirds of the economy is consumption,
[46:29] you can always say that AI isn't an
[46:31] impact. But based on the fact that
[46:34] there's no job creation, based on the
[46:37] fact that wages are coming down, why are
[46:40] we seeing
[46:41] Johnson red book at the highest level in
[46:43] the last 25 years except for after the
[46:46] helicopter printing? Why is that going
[46:48] on, guys? We had another huge print
[46:49] again. It's because AI is driving the
[46:52] stock market higher, creating net
[46:53] wealth. net wealth is going up and so
[46:55] you don't need as much income and
[46:57] savings particularly when most of the
[46:58] spending is happening at the higher end.
[47:01] Um, and this just shows how crowded
[47:03] everything is. I don't want to refute
[47:05] any of that. That's why the midcycle
[47:06] slowdown is the story. I don't think it
[47:08] changes. And just so we get back to how
[47:11] this leaves people angry and the voting
[47:14] that we're seeing with mom Donnie having
[47:16] more success. Um, this is probably the
[47:19] best chart as Warren Pies puts out. Tech
[47:22] expenditures
[47:23] versus residential fixed income
[47:26] investment. So, housing sucks.
[47:30] Tech AI is in parabolic boom. This is
[47:33] the reason why the public
[47:36] will not see a tightening.
[47:40] Yes, three right hikes, guys. Ignore
[47:42] them. Um, stripe sessions. So, I
[47:46] mentioned this real quickly. Go listen
[47:48] to this keynote. Again, I'm sending out
[47:50] the subscriber details on it. This is
[47:52] important for where we're going, and
[47:55] I'll make it clear when we get through a
[47:57] couple of these. Um,
[48:00] we're at the start of an AIdriven
[48:01] singularity moment showing that new
[48:03] business creation on Stripe has gone
[48:05] parabolic. My business is on Stripe. You
[48:09] have to go read what's happening in
[48:10] their businesses and especially this
[48:12] soloreneur thing of people like me that
[48:15] are solo business owners,
[48:18] how they're running their business and
[48:19] doing this without any other people. You
[48:21] have to go look at the agentic commerce
[48:23] suite. None of this is possible guys
[48:26] with what I talked about. As of Opus
[48:28] 4.5, there were no AI agents. Andre
[48:31] Carpathy was saying would take 10 years.
[48:33] We're now eight months since Andre Carpy
[48:35] said that and we're getting loops and
[48:37] tags and everything is accelerating to
[48:39] recursive self-improvement. Everything
[48:41] is happening. We are at the commerce
[48:43] point and the economy changes. So you
[48:46] hear this again and again from me the
[48:49] way you know the S&P 500. I publicly
[48:51] said I don't know what companies will be
[48:53] here in a decade. I don't know if any
[48:55] public companies will be around in a
[48:56] decade. Once you hit the point of things
[48:59] accelerating that fast and the
[49:00] competition goes excel, remember the S&P
[49:03] is time and obso is time to
[49:05] obsolescence. This is the reason why
[49:07] Bitcoin fits on the other side for me.
[49:08] So in this thing they talk a lot for
[49:11] crypto people. Stripe sessions is your
[49:14] crypto playbook and this is what is
[49:15] happening now and this is what you need
[49:17] to think with new CLI users. This is the
[49:19] beginning of the agentic side.
[49:23] They went through this in detail. AI
[49:25] native businesses are moving from Vibe
[49:27] coding, which was a year ago, to Vibe
[49:29] deploying. So, it's one thing to build
[49:31] an app, it's another thing to deploy it
[49:33] and start getting money in the door.
[49:35] They show what their Stripe projects is,
[49:37] which allows you to have the CFO, the
[49:39] COO, and all the infrastructure needs be
[49:41] done via an agentic tool.
[49:45] This is the signal of solar developer
[49:47] leverage. So you can grow your business,
[49:50] create it and then have it all done and
[49:52] deployed
[49:54] very very quickly by them. It can be
[49:57] global. My business is global.
[50:00] Early evidence of autonomous economic
[50:03] activity.
[50:04] It connected the CLI growth to the a
[50:06] that agents will become autonomous
[50:08] economic actors. First they will operate
[50:10] infrastructure then they deploy apps. AI
[50:12] agents are no longer just helping people
[50:14] write code. They are starting to operate
[50:15] the infrastructure layer of the
[50:17] internet. That my friends is what the
[50:20] crypto part is for. So let's go through
[50:22] the sequence because they're basically
[50:24] telling you first agents write code.
[50:27] This is the language part. Agents speak
[50:30] code. If human beings didn't have
[50:33] language,
[50:35] there'd be no economy. They're now
[50:37] talking to each other. Then they operate
[50:40] the infrastructure. They deploy and
[50:42] package real products. a transact, which
[50:44] is the next step. So, we need to do all
[50:46] of this before we can get to the
[50:48] transaction part. For that to happen, we
[50:51] need the guardrails. This is where
[50:52] crypto shows up. This is where
[50:54] everything changes. I'm bringing this up
[50:57] because it only just started. So, I want
[50:59] you to look at the summary. I just
[51:00] pasted that last page on what is
[51:02] happening from the Stripe sessions,
[51:03] highlighting this, and I want you to use
[51:08] the Jensen knowledge brain. So if you've
[51:10] created this, take that photo that I
[51:12] just had, copy paste it, put it in, type
[51:14] this prompt in there, and say, how does
[51:16] this show up in the five layer cake? The
[51:18] entire stripe sequence. Okay, he's
[51:20] taking the five layer cake. So this is
[51:22] the response from my brain, the Jensen
[51:24] Yuang brain. Aentic commerce, payments,
[51:26] checkout, financial guardrails live in
[51:29] layer five, the application layer. So in
[51:31] the same way that I say that Eli Liy's
[51:33] in the application layer, I continue to
[51:35] say that the financial guardrails crypto
[51:38] are 100%
[51:40] the agentic side. You do not have AI
[51:44] agents continue to move forward and we
[51:46] couldn't be there without the first ones
[51:48] happening. So it's going to start right
[51:50] now, guys. Every month that you come
[51:52] back, there will be more AI agents
[51:54] transacting with each other and they're
[51:56] the ones building the layer of the
[51:58] businesses to talk to each other. So,
[52:00] this is only going to increase from
[52:01] here. So, for those of you not paying
[52:02] attention to crypto,
[52:05] good luck a year from now playing
[52:06] catch-up. I will be launching my crypto
[52:09] version of this YouTube in either
[52:11] September or October. I'm working on the
[52:12] data now. I will not do it until I have
[52:14] all the data done. I'm working with
[52:15] OpenB. Shout out to Diddier uh from
[52:18] OpenB, the founder. Uh I highly
[52:21] recommend using OpenB for a lot of this
[52:23] stuff, too. But he and I are working
[52:24] together on this. So I just want to make
[52:26] sure you read through this stuff in
[52:28] terms of what happens with all of this
[52:30] stuff. It's important. So we went from
[52:33] 20 23 to 25 IQ reasoning. We had to get
[52:36] the model smart enough so that they
[52:37] could do coding on their own. This
[52:39] started late 25. It's accelerated with
[52:40] the Opus 4.5 4.6 4.7 4.8 and now fable
[52:44] into loops into co into uh tags. All of
[52:48] that shows that we are making that
[52:49] progress. We now are building the
[52:52] agentic infrastructure. This is what
[52:54] Stripe talks about. And then we move
[52:56] into the agentic commerce and eventually
[52:59] into the physical autonomy side. Each
[53:02] layer unlocks the next. So even as AI
[53:04] gets cheaper per unit of intelligence,
[53:06] total command. If you invest in all of
[53:08] these, this was Nvidia. This is
[53:11] everything that's worked in my thematic
[53:12] portfolio. Okay. Now we're getting into
[53:15] this part which is crypto.
[53:20] where AI agents and crypto collide.
[53:23] Again, crypto is the native rail.
[53:24] Tokenization starts and you're now
[53:27] moving at a speed that is impossible to
[53:29] measure. Volumes are going higher,
[53:30] transactions are going higher, velocity
[53:32] of money is going higher. The dormant
[53:34] assets were releasing 2/3 of the 700
[53:37] trillion on the planet, which is not
[53:38] part of the global economy. It will be
[53:40] coming. It will be the largest monetary
[53:42] increase in the history of mankind
[53:44] happening on a regular basis due to
[53:46] tokenization. The collateral for that
[53:48] world is Bitcoin.
[53:50] Which is why when the debasement trade
[53:52] is happening, you're getting a
[53:53] capitulation on anything that doesn't
[53:56] have a valuation. Gold doesn't have a
[53:58] valuation, silver doesn't have
[53:59] valuation, Bitcoin doesn't have
[54:01] valuation, Palanteer doesn't have
[54:03] valuation, everything. SpaceX doesn't
[54:05] have one, Tesla doesn't have one. You're
[54:06] getting all of these things to unwind
[54:07] because the easy part of investing in
[54:09] the infrastructure thing is you're
[54:10] getting to buy things with PEG ratios
[54:12] below one. But the midcycle slowdown is
[54:14] where it gets harder. So here's the
[54:16] crypto chart. This is Bitcoin. Looks
[54:17] horrible. I'm going to keep saying what
[54:19] I've been saying now for weeks since it
[54:20] failed again here, which is no other
[54:23] way. This is a bare market. 200 days
[54:24] moving down. We keep going lower and
[54:26] lower. We made new lows in this. You do
[54:28] not need to step in and be a hero. If
[54:30] what I'm saying is right, I'm not doing
[54:32] that anymore. I do have plenty of
[54:34] Bitcoin in my possession. I even bought
[54:36] a little bit in here. I stopped myself
[54:38] out. I will keep playing this game in
[54:40] here, moving through these moving
[54:41] averages, getting stopped out just to
[54:44] make sure that when this starts, I do
[54:46] have some size. This is what I did with
[54:47] Micron. Micron I just kept buying
[54:49] Bitcoin. I already have plenty for
[54:51] anything. If this ends up happening and
[54:53] it does work out the way I think. If
[54:55] this is a big move, there's no need to
[54:56] be involved early, guys. Continue to
[54:58] trade your names that have positive
[55:00] momentum and don't worry about it. Uh
[55:02] the good news is I get a bunch of people
[55:03] saying Michael Sailor is is going to
[55:05] collapse. This whole thing is going to
[55:06] go through. Blah blah blah. Every time
[55:08] that I seem to get here where I get a
[55:10] bunch of email on strategy, that's
[55:12] usually near the bottom since I mean the
[55:14] guy owns a ton of Bitcoin. Uh, I don't
[55:17] see what it is. It's all Bitcoin. If
[55:19] Bitcoin goes down, strategy is going to
[55:21] keep going down. Um, final two slides.
[55:24] This I showed this before. So, this is
[55:26] what I'm working on or been working on.
[55:29] 40 names, tokenized index, some public
[55:33] companies, six.
[55:35] The rest of them are tokens. They
[55:37] represent eight different verticals or
[55:39] sectors to show the economy. And the
[55:41] reason I want to show this again and
[55:43] again and again, this is Bitcoin making
[55:45] new lows. This is the blue line which is
[55:47] those 40 names, okay, equal weight that
[55:50] represent the ecosystem for the AI
[55:52] agents. This is what I'm putting
[55:54] together. The interesting thing, it
[55:55] didn't make new lows, so it did break
[55:57] away. I believe the ecosystem needs to
[55:59] lead the way out and then Bitcoin will
[56:01] join, but this is the way it should be
[56:03] in. Have a great week, guys. Um, I will
[56:06] see you from Maine for the rest of the
[56:07] summer. Uh, stay cool. It's uh 59
[56:11] degrees right now.