The Fireworks Show Is Over: Why This Is An AI Rotation, Not a Bubble Unwind

[0:00] Let's start out with a shout out uh to
[0:03] the Nick fans, of which I am a long time
[0:06] suffering Nick fan. Very excited about
[0:08] what's going on. Um stocks fall. Is this
[0:13] a bare market or a rotation? Uh I will
[0:16] go through a bunch of papers that I I
[0:17] wrote this week which um talked about a
[0:21] lot of the things I think that happened
[0:22] but also about the way I'm thinking
[0:23] about the things going forward. And then
[0:25] I will release a paper next week for
[0:27] subscribers on the new economic cycle.
[0:30] I'm also working on a paper on the uh
[0:33] 800 volt DC. Uh that'll be a new theme.
[0:37] And I'm also working on the application
[0:40] part of the five layer cake which fits
[0:42] in with more on Eli Liy which you're
[0:45] going to see more and more about. And I
[0:46] did release a very long extensive piece
[0:49] on Eli Liy uh for subscribers on Friday.
[0:54] Uh I do believe this is a very big deal.
[0:58] Uh S&P down 2.6% for the week. Basically
[1:02] the worst week since liberation day
[1:04] week. Um Q's down 4 and a.5% easily the
[1:09] worst since then. And again just like in
[1:11] the S&P the S&P was up nine weeks in a
[1:13] row. Q's were up eight of nine weeks.
[1:15] Then you get into small caps. They were
[1:17] down but again a very strong rally. So
[1:19] none of this should be a surprise.
[1:22] Um, yes, the charts, you know, you're
[1:24] going to have people come out and say
[1:25] that was it. Looks like a bare market is
[1:28] starting
[1:30] doesn't at all to me. Um, looks like
[1:32] this is the beginning of of what I
[1:35] believe is a necessary rotation for
[1:38] those of for those people looking. Um, I
[1:40] do believe that like memory stocks, it's
[1:43] very possible
[1:45] uh that those have made a top and the
[1:48] fact that I keep getting push back
[1:49] whenever I'm negative on memory at this
[1:51] point. I'll go through the reasons why
[1:54] uh and at least give you the bare case,
[1:56] not bearish for memory, but I do not
[1:59] think it is the place to have your
[2:01] capital. If you get a move down in
[2:02] memory 40%, I think your riskreward has
[2:05] changed. So you've already had kind of a
[2:07] move, but remember I bailed out of my
[2:08] micron in the 600 and 700 area. We
[2:11] closed in the high 800. So the reason I
[2:14] was getting out there is I already
[2:15] thought the riskreward was I couldn't
[2:17] get doubles and triples anymore in
[2:18] memory. I still think you can get
[2:21] doubles and triples in Marll and in Eli
[2:23] Liy and stuff like that and I will go
[2:25] through. So in the cues you broke the
[2:27] 20-day. The reason I have this arrow
[2:29] here is to show you that after
[2:30] liberation day, we went through this
[2:32] long period um much longer than what we
[2:35] just did. But you can also see it wasn't
[2:36] as sharp. And in this case, we finally
[2:39] closed below the 20-day. We just finally
[2:42] closed below the 20-day. We hadn't even
[2:44] really ticked it. At least back here, we
[2:46] kind of hit it and never closed below
[2:47] it. But then once we did break below it,
[2:50] that ended the euphoric period. And it
[2:52] happened in the summertime. This is just
[2:54] before the summertime. But then you did
[2:56] go through some digestion and that's
[2:58] kind of what I called the fireworks
[3:00] piece is that the all you can eat uh
[3:03] memory and semiconductor and
[3:04] infrastructure buffet where you could
[3:06] just buy any of them and they would gap
[3:08] higher 10 to 30% because the market was
[3:11] not on the agentic trade and they were
[3:13] not yet positioned for it. The catchup
[3:15] is over and now that everyone is
[3:17] basically on the same thing which is
[3:18] what the fireworks paper was about. Uh,
[3:22] I just think it's now a two-sided market
[3:24] and you're going to have to pick your
[3:25] poison. But I would not be getting
[3:27] bearish. And I'll go through the the
[3:29] foundational reasons why. The S&P, same
[3:32] exact thing. Finally broke below the
[3:34] 20-day. In the case of the S&P, it
[3:37] continued to move higher. I think the
[3:39] same thing will probably play out this
[3:41] way. I'd be much more worried about the
[3:43] Q's than I am the S&P. And I will go
[3:46] through the reasons why. Um, another
[3:48] reason to not be too bearish is how
[3:51] strong the market was. I do believe that
[3:54] sustainable tops or bad places for the
[3:57] market happen with divergences and RSIs.
[4:00] And in the case of S&P, we made the
[4:04] highest weekly RSI going back to 24. So,
[4:07] I think we'll make new highs and make
[4:09] divergences like we did here. And I
[4:13] think until that shows up, I'm just not
[4:15] going to be that worried about it.
[4:16] Here's the weekly chart for Q's. Same
[4:18] exact thing. Highest RSI in over a
[4:21] couple years. Now, the big thing that
[4:24] showed up was that beta, this is the
[4:26] Morgan Stanley uh volatility um factor,
[4:31] long short uh was down 10 and a half% on
[4:35] uh on Friday. If you take this back all
[4:37] the way to 2000, I think you have to go
[4:39] all the way back to 2000 to find
[4:41] anything that was even similar. So,
[4:43] again, it was a big move on Friday in
[4:45] beta. Um, this is pure. So, this is
[4:48] sector neutralized. The Morgan Stanley
[4:50] one still has sector exposure. This is
[4:52] the pure beta or volatility verse
[4:56] profitability or quality. And it was the
[4:58] worst um day since going back to
[5:02] liberation day last year. So, we did see
[5:05] a cleanse. Um, if I was going to pick a
[5:08] place where we already have divergences.
[5:11] So, RSI, this is a lower peak than what
[5:13] happened here. The volume here in
[5:16] memories and micron in particular was
[5:18] very high and this looks a little bit
[5:20] more like a reversal pattern. But again,
[5:22] I've bailed out in here.
[5:25] I can see where people can go. Maybe we
[5:28] get down here and we bounce and we go
[5:29] through it. But I think we need time to
[5:31] go through this. And I will go through
[5:32] the reasons why I am worried about
[5:34] memory when all of a sudden every single
[5:38] person around the planet is positive on
[5:40] memory. I spoke at the New York Stock
[5:41] Exchange this week. There were a lot of
[5:44] brokers from Asia and in particular from
[5:46] Korea. Um
[5:48] they all confirmed what we've read. Uh
[5:51] retail has basically put in an all-in
[5:54] bet on SKH Samsung and the Korean
[5:56] market. So since we were already seeing
[5:58] divergences there that I'd been talking
[6:00] about with sectors like construction and
[6:02] machinery breaking down and they are
[6:04] trading at kind of one-mon lows still um
[6:07] and the breath was horrible in Korea. I
[6:10] do think that this story that I put out,
[6:12] the fireworks show is over. Agentic AI
[6:14] moves from discovery to digestion. So I
[6:16] published this on uh on Thursday
[6:20] in the morning and basically the
[6:22] firework show is over and the last
[6:24] couple of weeks have been the climax of
[6:25] an incredible display, but the agentic
[6:27] AI buildout has just begun. And I kind
[6:30] of go through this as finishing up as my
[6:32] father taught me the odds on the tote
[6:34] board are now fair. So let's sit out the
[6:36] race until we see better odds. So for
[6:38] everyone who's looking, the thematic
[6:39] portfolio that I put together is based
[6:41] on the agentic world. It has much higher
[6:43] highs to go. When I talk about memory,
[6:45] that is one specific channel that has
[6:46] had a parabolic move. You could have the
[6:48] entire index go higher and memory just
[6:50] kind of bounce from a lower level and it
[6:52] won't matter. Um Eli Liy is in the
[6:55] basket and it was up on Friday. Uh it
[6:58] was up last week. Marll was up big last
[7:01] week and it was up after market hours as
[7:04] I'll show why. Um, so when I talk about
[7:06] the fireworks show is over, that just
[7:08] means the easy let me just put my money
[7:10] on any of these I think is over. And I
[7:12] think now you have to do your homework.
[7:13] It's a long short side, but I believe
[7:15] the agentic side will continue to
[7:17] outperform, but maybe for the next three
[7:19] to six months.
[7:21] It's it's a much choppier um
[7:23] perspective. Uh I will take you through
[7:25] what I started to buy this week, but get
[7:27] a chance read the paper. I spent a lot
[7:30] of time thinking about it. And the
[7:32] reason for all of this was really broken
[7:34] down by three events that I wrote papers
[7:37] after each one. The January CES that was
[7:40] when Jensen uh introduced the Vera
[7:42] Rubin. It was a huge surprise to the
[7:45] industry in the AI world and for
[7:48] investors it had only been about six
[7:51] weeks before where they had been saying
[7:54] it was a bubble. So at the end of
[7:56] October the consensus was that AI was a
[7:58] bubble that the capex wouldn't work. Jim
[8:00] Chainos was doing interview after
[8:01] interview or at least uh posting and
[8:05] then he did do one podcast which I
[8:06] referenced which was we were questioning
[8:09] everything and that's when Oracle
[8:10] started to peak and everything started
[8:12] to go down. Open AAI was questioned and
[8:13] there were fears that it would be under
[8:15] uh Sam Alman did his interview with Brad
[8:18] Gersonner. All of these things happened
[8:20] just before CES. So when CES came out
[8:23] and he basically talked about tokens per
[8:24] watt, the need of memory, opticals,
[8:27] inference, and he basically said this is
[8:29] it. That's when the five layer cake was
[8:31] born. He mentioned it, but he really
[8:34] explained it in detail during the Larry
[8:36] Frink interview in Davos, which was a
[8:38] couple weeks later. So at this point,
[8:40] traveling all over, talking to mutual
[8:42] funds, talking to hedge funds, people
[8:44] didn't buy into the AI situation. By the
[8:46] time we went through the March Morgan
[8:49] Stanley Tech Conference, that's when I
[8:51] started to hear a shift from people.
[8:52] That's when they were reaching out
[8:53] saying, "What do I buy? What's what's
[8:55] not too late?" And in there, it was
[8:58] because companies like AMD and Intel and
[9:00] Nvidia and everyone, Dell was talking
[9:03] about how big their backlog was
[9:05] becoming. And by the time we got to
[9:07] Computex last week, it was the same
[9:09] story. It happened. I saw nothing in
[9:12] there that was new. Nothing. Um, and in
[9:16] the interim, Broadcom beats numbers,
[9:18] great results, stock trades down. Sarah
[9:21] Bros IPO comes out, trades down 50% from
[9:24] there. Google announces an $85 billion
[9:27] financing, as I'll show. Meta is now
[9:29] talking about it. You've got all of
[9:30] these IPOs coming out. And all of this
[9:33] is related to the exact same thing,
[9:35] guys. It's to fund this buildout. So,
[9:38] this was the surprise.
[9:40] This kind of stuff looks like more of a
[9:42] top meaning they're coming out to raise
[9:45] money. It's not just IPOs like we have a
[9:47] great business go buy our stuff. It's we
[9:49] need the money not for monetization for
[9:52] the people that have been invested. We
[9:53] need the money to go buy more chips and
[9:55] more everything. But I think people are
[9:57] getting crazy and believing that the
[9:59] bottlenecks and the shortages and the
[10:00] stuff won't prevent them. Everyone
[10:02] believes we need lots of money. As I go
[10:04] through this, as someone who's been on
[10:05] this, I really do believe people are
[10:08] missing a part of this market that
[10:10] they're going to have to think about.
[10:11] And I will say it here, there will be a
[10:14] lot of the capex spend which is not
[10:17] needed.
[10:20] I've been writing about this since I
[10:22] wanted to just highlight. So before CES
[10:25] or right at the same week, I wrote this
[10:27] thing on the inflection point when AI
[10:29] crossed the threshold in December.
[10:32] I wrote why 2026 is the revenue
[10:34] inflection year. AI agents are here. So
[10:37] remember again as you're looking at me,
[10:38] whatever you think about what I've
[10:40] created, what's gone on. I'm not just
[10:42] trying to be contrarian. The odds have
[10:44] shifted. My father taught me the top
[10:47] board has shifted. Everyone is in.
[10:49] They're now worried more about are did
[10:51] they buy something at the peak? And it's
[10:54] because the certainty is there. And I
[10:56] don't believe anyone should ever have
[10:58] certainty over a three-year lookout
[11:01] ever. And I will go through why. You had
[11:04] speculation. There's no way to get
[11:05] around it. In the semiconductors, the
[11:07] gross option premium,
[11:10] the Cosby over the last last six
[11:12] sessions, the Cosby's up 12% yet breath
[11:15] was negative each day. And not by a
[11:16] little bit. I mean, come on, guys. Like,
[11:19] at some point, it just gets completely
[11:21] ridiculous. the skew the single stock
[11:25] put one month put call has now collapsed
[11:27] to the slowest level in Goldman Sachs
[11:29] entire database so it was easy to make
[11:31] money retail was involved mutual funds
[11:33] were involved hedge funds were involved
[11:35] everyone was playing catch-up and
[11:36] remember because of the Iran situation
[11:38] and how many strategists came in I
[11:40] really don't think people jumped in at
[11:42] least with one foot until after the
[11:43] Morgan Stanley TMT uh side which was
[11:46] after the beginning of Iran and when
[11:48] people started to see a bounce in Iran
[11:50] so that's why I think it was locked up.
[11:52] Um I' I've never mentioned John Husman
[11:57] on here. Um and it's because almost
[11:59] everything I've ever read by the man has
[12:01] been bearish, but I think his work is
[12:04] quite good actually. Um and I like
[12:06] anything that looks like this where I
[12:08] don't have to get people's opinions on
[12:10] things. And the only reason I bring this
[12:12] up is because for people who want to be
[12:15] super bearish, you can look at this. And
[12:17] I think this is the kind of stuff that I
[12:19] would build. I use the word cluster at
[12:22] the same time. I'm looking for warning
[12:24] flags. This is a an element of what I
[12:28] like to do, which is I don't want to
[12:29] pick and choose. I just want to take as
[12:31] much data as possible. And again, he did
[12:34] have a lot of things that went up. So,
[12:35] if you want to be bearish, if you want
[12:36] to read this and ignore what I'm saying,
[12:38] all good. You got a lot of signals in
[12:40] here that occurred during bull markets.
[12:43] So, I wouldn't sit there and go crazy on
[12:44] it, but I thought it was interesting.
[12:47] So, here's where we get into the reality
[12:49] of stuff. These are all facts. You're
[12:51] going to have to expect this to change.
[12:54] S&P earnings.
[12:56] This is what has happened this year,
[12:58] guys. And again, normally what goes on
[13:00] is this.
[13:02] Normally, every year
[13:05] we go into the year and then we start
[13:06] revising the numbers lower. It is like
[13:08] clockwork. Well, this year up. So,
[13:11] normally everything goes down by about
[13:13] 2% and this time it just rocketed
[13:15] higher. This is overlaid with nominal
[13:18] GDP
[13:20] which highlights the profit margins. It
[13:22] highlights the productivity gains. I
[13:24] think you're making a huge mistake
[13:26] getting in in the train of AI. Uh
[13:29] forward earnings nominal verse real and
[13:32] what happens in a recession. So I'm only
[13:34] showing this because I'm going to start
[13:36] showing you guys this more and more.
[13:38] Before a recession occurs, before some
[13:40] kind of big problem occurs, you start
[13:43] getting leaning lower. This one is
[13:45] really important and I'll I'll let you
[13:46] know why. This is the.com bubble. So,
[13:49] I've shown before that things didn't get
[13:51] bad in the.com bubble until late in
[13:54] 2000. The NASDAQ went down and then the
[13:57] rest of the market went down. So, I'm
[13:59] looking for a rotation, not a bare
[14:01] market. If the rotation fails, it's
[14:03] going to be because the economic data
[14:05] fails. And I'll show you what that'll
[14:07] look like. Here are semiconductor sales.
[14:10] These are not pees. These are sales.
[14:14] profit margins. All these yellow lines
[14:18] before we get recessions, guys, even in
[14:20] the.com bubble, here's the.com bubble.
[14:23] Profit margins go down.
[14:25] Profit margins make all-time highs. And
[14:28] not only that, there is no mean
[14:30] reversion. For 50 years, your you
[14:33] corporate profit margins always
[14:34] reverted. You have to think about 22,
[14:38] 21, 22, 23, 24, 25, 26,
[14:43] guys. This is when chat GPT was launched
[14:46] in late 22. Sorry.
[14:51] Forward earnings change versus is the
[14:54] market expensive. Here's the S&P on a
[14:56] forward PE ratio right now.
[15:00] This is not a bubble, guys. I could say
[15:02] it again. And if the S&P were to fall
[15:04] 20% and earnings don't go down, here's
[15:07] what we'll be. So again, the reality is
[15:09] when you have these corrections and what
[15:11] goes on, I just want you to think about
[15:13] them. COVID, that was a decision by the
[15:16] government. Rate hikes, that was a
[15:18] decision by the Fed.
[15:21] Tariffs, that was a decision by the
[15:22] government. Unless you believe that
[15:24] Donald Trump is going to do something.
[15:26] So maybe we get a correction. And the
[15:28] reason is because oil prices go to 300.
[15:31] I got no problem with that. that it's on
[15:32] the distribution, but if it does, some
[15:35] other things are going to show up in the
[15:36] market first. Credit spreads never lie.
[15:40] They always lead. Very seldom have you
[15:43] not seen in history. Credit start to go
[15:45] because that's the way the capital
[15:46] system was worked on. It's all change.
[15:49] And this is what has happened while
[15:51] profit margins have been exploding as
[15:52] credit spreads have just continued to go
[15:55] down.
[15:57] Jobless claims, same exact thing. Look
[16:01] how cyclical they are. Every time before
[16:03] we get a recession, we get jobless
[16:05] claims going higher. Even in the.com
[16:07] bubble, we haven't had a budge yet.
[16:10] Nobody's getting fired and and getting
[16:12] uh claims.
[16:15] PMIs, another thing, and this is
[16:17] overlaid with capital goods just showing
[16:19] the massive expenditures going on in AI.
[16:21] Do you really believe that the
[16:23] expenditures are going to stop in AI?
[16:24] You're nuts.
[16:26] So, credit spreads are near all-time
[16:28] tights. Job lossless claims show no
[16:30] growth in insurance claims. Earnings are
[16:31] growing rapidly and revisions are
[16:33] growing. PMI manufacturing just came
[16:34] out. Blah blah blah. Are these
[16:36] conditions normal for a sustained bare
[16:37] market? No. What you're describing is
[16:40] close to the photographic negative of a
[16:42] sustained bare market. Guys, don't
[16:45] listen to people in their bearish
[16:47] things. Don't pick an economist who is
[16:49] perma bear and read what they're
[16:51] writing. Throw it out. If you want, copy
[16:54] what they do. take the snapshot and put
[16:57] it into all of the LLMs and ask them all
[16:59] the same questions. I wrote something
[17:01] last week that I use LLMs to do this
[17:03] because they don't have a bias and
[17:04] someone said they do have a bias. They
[17:06] don't have a bias when it comes to
[17:07] calling recessions, guys. Just put them
[17:09] into all five then and take the average
[17:11] of all five for the answers. It's an
[17:13] easy way to deal with it. The honest
[17:15] caveat, the one variable missing from
[17:16] your list is valuation, which I didn't
[17:18] include in there for a reason. And
[17:20] that's because 2000 was one. But even in
[17:22] 2000 only became a sustained bare market
[17:25] once the macro broke. Claims rose. ISM
[17:28] fell, earnings decline, profit margins
[17:31] came down, the valuation crack started
[17:33] after that. Handicapping it. This is a
[17:35] low base rate environment for a
[17:37] sustained bare market. You'd have to
[17:38] have a lot of stuff go on. If those
[17:40] things are going to happen like they did
[17:42] in 2007, the market will still make new
[17:44] all-time highs and you'll be able to
[17:45] change the risk going forward. The
[17:47] probability of an 87 style move at this
[17:50] point. You can make the argument there.
[17:52] That is not where we are. Bearish
[17:54] sentiment is everywhere. We do have
[17:56] retail heavily involved, but they've
[17:58] been making money for a long time now.
[17:59] And if they're using call options, guess
[18:01] what? Call options have a defined loss.
[18:05] Stocks, owning stocks,
[18:08] your defined loss is a lot bigger than
[18:10] putting money into a call option. So, if
[18:11] they're using call options and having
[18:13] replaced stock, but they also changed
[18:15] their minds and they have no trouble
[18:17] getting bearish. So, if I was going to
[18:18] pick where the S&P is likely to go, I
[18:20] showed it before. Here's the 50-day
[18:21] moving average. Here's the prior
[18:23] consolidation on what I would argue was
[18:26] a wave four. And then we've got
[18:31] the 62% retracement. So, that's where I
[18:34] think it could go. If it went below
[18:36] that, I think it would be quick. And
[18:38] just so you get a sense on Friday when
[18:41] it felt like the world was ending and
[18:42] the S&P was down close to 3%
[18:45] five sectors were up. This was a
[18:47] rotation guys and not surprisingly the
[18:50] ones that were down were the ones
[18:51] directly related to AI. Okay.
[18:55] Jeff Degraphth
[18:57] very very good at this particular stuff
[19:00] looking for areas to invest in. He's
[19:01] been talking a lot about healthcare
[19:03] lately. Adam Parker also been talking
[19:05] about healthcare from another
[19:06] perspective. The capital that exited the
[19:08] chip complex didn't leave the market. It
[19:10] rotated into healthcare and financials.
[19:12] The day's two strongest sectors. I like
[19:15] me some healthcare. Mark Newton. Another
[19:19] one. I see good likelihood of rotation,
[19:21] rotation, rotation as healthc care fins
[19:23] come back to life. Okay, these are two
[19:25] guys don't work together, compete with
[19:27] each other. And that rotation out of
[19:30] here, some into there in my opinion, and
[19:34] some into here, some into there. It's
[19:36] this that I'm worried about. You have to
[19:38] be more selective in the chip front. The
[19:40] application side, that is what the Eli
[19:43] Liy paper is about. I think you should
[19:45] read it if for no other reason just to
[19:47] understand what the application side is
[19:48] and why we are entering one of the most
[19:50] important times in our lifetime with
[19:53] regards to pharma
[19:56] and the ability for people to deal
[20:00] with longevity.
[20:02] This is a p the p t title of the paper
[20:04] as of now that's going to go out Monday
[20:06] or Tuesday. From labor verse capital to
[20:08] compute verse energy.
[20:11] I'm just going to read this part.
[20:12] Yuang's other formation is even simpler.
[20:14] The input is electrons. The output is
[20:16] tokens. That is the new production
[20:17] function. The factory consumes energy
[20:19] and produces intelligent. The irony is
[20:22] that the great software winners were
[20:24] already showed us that the labor versus
[20:25] capital relationship was breaking. This
[20:27] again is what AI is about. So in the old
[20:29] world it was labor verse capital. You
[20:31] had credit when profit margins went
[20:34] down, you started to see asset prices
[20:36] come down, you had problems with debt,
[20:38] and it would lead to people being fired.
[20:40] Well, now it's just compute verse energy
[20:42] and human beings are not a part of the
[20:44] equation. Credit's not a part of the
[20:45] equation. And I know that's funny to
[20:46] see, but the reason I'm bringing this
[20:48] down, Microsoft, Alphabet, Meta, Amazon,
[20:50] and other platform companies scaled
[20:52] revenue margins and market value without
[20:53] scaling labor or brickandmortar
[20:55] locations. They didn't have any debt to
[20:57] win the battle. It was all based on
[20:59] revenue per employee. They didn't have
[21:00] to hire as many people. Their profit
[21:02] margins are above the S&P 500 and they
[21:04] are the ones now funding the next leg in
[21:07] a extreme competition to avoid
[21:10] obsolescence and most of them in my
[21:12] opinion will lose. The business cycle
[21:14] has changed. So this is the old one.
[21:17] Credit expands. You borrow and hire to
[21:20] build capacity. You hire people. Demand
[21:23] slows. I mean everyone has done this
[21:25] who's managed people. Morgan Stanley
[21:27] open an office in Brazil. You hire more
[21:28] people. They gave me more headcount. I
[21:30] refused it because I didn't believe in
[21:32] the future of Brazil at the time coming
[21:35] out of the emerging market crisis.
[21:37] Demand slows, margins compress, layoffs
[21:39] transmit. That's not going to happen
[21:41] ever again because the hieres
[21:45] don't consume. They don't have children.
[21:48] They eat tokens. We are building compute
[21:52] and memory. The cycle of AI is about
[21:55] bottlenecks and shortages. It's about
[21:57] the power grid, about chips, about
[21:59] networking. It's the physical world. The
[22:01] AI cycle breaks when digital demand
[22:03] outruns physical supply. Now, instead of
[22:06] breaks when you're going through it,
[22:07] this is when you have a AI quote unquote
[22:10] recession or when the AI trades aren't
[22:12] working. I think we're in that mode for
[22:14] some of the things now. Tokens are the
[22:17] food source of these digital workers. A
[22:18] human employee consumes wages, benefits,
[22:20] office space, management, time, and
[22:21] trading. Then they have families, they
[22:23] buy cars, have children, and leverage up
[22:25] through a home. A digital employee
[22:26] consumes compute, memory, electricity,
[22:28] and data. That changes the bottleneck,
[22:30] and it also changes the business cycle.
[22:32] I'm building a business. I have a
[22:35] company.
[22:36] I've managed hundreds of people over the
[22:39] years. I'm not hiring anyone. There is
[22:43] one person who works with me to help
[22:45] grow the business aside from all of my
[22:48] digital employees. This is a completely
[22:51] new world with high margin businesses.
[22:54] established companies are going to have
[22:55] a harder time adopting which is why I am
[22:57] negative on all Fortune 500 companies by
[23:00] 2030 but for the time being we are here
[23:03] to make money guys and the way to do
[23:04] that is to follow this so what happened
[23:07] and what did semi analysis I've directed
[23:10] you guys to their fantastic agentic
[23:12] traffic has surpassed human traffic
[23:15] across the worldwide internet for HTML
[23:17] web pages guys it's already started
[23:21] transactions same thing are going on
[23:23] this way. Whatever. Raul Pal and I have
[23:26] been involved in macro for a long time.
[23:28] Raul still does GMI.
[23:31] He's obviously jumped into crypto and
[23:33] for some people they've stopped
[23:34] listening to him. Go listen to this
[23:37] 40inute podcast and see if you
[23:40] understand what he talks about. He
[23:42] breaks down how AI agents will create a
[23:44] vast invisible economy operating at
[23:46] machine speed and that crypto rails
[23:48] especially major layer level ones are
[23:51] the infrastructure needed for agents to
[23:53] transact, settle and coordinate value.
[23:55] This is the reason why I am buying
[23:57] Bitcoin. This is the reason why I have
[23:59] been this is the reason why I believe in
[24:02] building out for you guys a crypto
[24:06] YouTube to basically create the
[24:08] ecosystem. Now, Bitcoin, as I said last
[24:11] week, is in a bare market, and until it
[24:13] crosses the 200 day moving average,
[24:15] there's no need to be involved in it.
[24:17] I'm buying it because I believe we're
[24:18] going to get there, and I'm buying small
[24:19] bits and pieces on the way down, but
[24:22] especially because we're here and
[24:24] because of this, and again, we've got
[24:26] divergences here on the downside for
[24:27] now. So, we'll see what happens. If I'm
[24:29] right about the equity market being
[24:31] fine, if I'm right about three-month
[24:32] bills remaining under inflation, I think
[24:36] we've now done a correction. And I'm not
[24:37] going to spend a lot of time on this,
[24:38] but I want you to think about here was
[24:41] the ETF
[24:43] in for Bitcoin. Here was Trump taking
[24:45] over. This whole period of time for me
[24:47] now that we look back and we're going
[24:49] through it was kind of a sell the news
[24:50] event. If we get back above the 200 day,
[24:53] I think it's an important thing. But
[24:54] we're at the 200 weeks. Let's go back to
[24:56] Charlie Munger. If all you ever did was
[24:58] buy highquality stocks at the 200E
[25:01] moving average, you would beat the S&P
[25:03] 500 by a large margin over time. The
[25:05] problem is very few human beings have
[25:07] the kind that kind of discipline. So
[25:08] here's what I'm going to say. Bitcoin is
[25:10] obviously not a highquality stock.
[25:11] Charlie Mer didn't believe in Bitcoin. I
[25:13] do. I believe it is the only highquality
[25:17] stock that will exist for certainty in a
[25:20] decade. Regardless of that reason, the
[25:23] 200E moving average is where we are in
[25:25] Bitcoin. So this is the part that
[25:28] matters. Few human beings have that kind
[25:30] of discipline. So, just when everyone is
[25:32] getting bearish, just when everyone is
[25:34] throwing in the towel, this week I had
[25:37] Patricia, I had Jessica, I had Rick, I
[25:40] had Tad, I had a bunch of people inside
[25:43] the community along with Jordy all
[25:46] nibbling down here and buying a little
[25:47] bit. It's a very different world than
[25:49] talking to people on Wall Street like
[25:51] during the bottom of Liberation Day with
[25:53] Bitcoin. The people that have been
[25:54] involved, they've been paid off
[25:56] handsomely for buying during times like
[25:57] this. I just thought it was ironic that
[25:59] all on the same day, whether it was
[26:01] yoga, whether it was a friend, whether
[26:03] it was a trainer, it didn't really
[26:04] matter. Everyone was nibbling. My
[26:06] thematic portfolio was down 4%. And
[26:09] again, 1 2 3 4 five at least greater
[26:15] than 3.89% with all those weeks in a
[26:18] row. Again, this just creates
[26:20] opportunity. Here is the Morgan Stanley
[26:22] one that was created equal weight as of
[26:25] last week or two weeks ago. And this is
[26:27] it versus the hyperscalers. This is my
[26:30] favorite trade.
[26:32] This was down, but it was up for the uh
[26:33] it's up for the month. But more
[26:35] importantly,
[26:36] the hyperscalers
[26:38] were down big on on Friday. And there's
[26:40] a reason that's important. For those of
[26:42] you with the subscriber list, put all of
[26:45] the sheets. I ran this and just said,
[26:47] "Hey, go through and find your top 10
[26:50] names that you think I should buy based
[26:53] on the PEG ratio and based on the
[26:55] pullback that's happening from a
[26:56] technical basis." And again, it combined
[26:59] the technical side. It went through it.
[27:00] It combined everything and it said here
[27:02] are the three. There's two buy on
[27:03] weakness names, meaning these names have
[27:05] been weak. Their technical scores are
[27:07] fairly low. And then you've got one
[27:09] where their technical scores are high
[27:11] where because of the PEG ratio and
[27:12] because of their earnings growth rate
[27:14] because of the I would stick long and
[27:16] you can look to buy them. That's the way
[27:17] I'd be using this stuff. Now let's go to
[27:20] my DRAM situation. So it really bothers
[27:23] me when people are referencing this
[27:25] Goldman Sachs on memory DRAM to remain
[27:27] in under supply until at least 28 2028
[27:30] blah blah blah blah blah. Okay. Now
[27:33] again
[27:34] here are the facts
[27:37] behind Goldman Sachs and Micron. Hold
[27:40] hold hold
[27:42] I don't give any credibility to anyone
[27:45] talking about memory because as I repeat
[27:48] I was abused most of last year
[27:50] throughout the year even into September
[27:54] the only thing that time this changed
[27:55] was October and that's because DRAM
[27:57] prices were not only up but they were
[27:59] still continuing every week. That's when
[28:01] the most sophisticated tech people
[28:03] started to at least nibble, but it had
[28:06] already gotten away from people. And
[28:07] then by the time people got on board, my
[28:09] issue is this chart. All of the stuff
[28:12] that Goldman talked about is based on
[28:14] this chart. And I'm telling you guys,
[28:17] although I agree with this, you cannot
[28:20] translate token consumption into memory.
[28:23] There are many things that can go wrong
[28:26] along the way for memory. many in terms
[28:29] of not needing as much and that's just a
[28:33] reality. So when you start extrapolating
[28:35] something all the way out here and say
[28:36] I've done the numbers to go through it,
[28:38] you're really going the wrong direction.
[28:41] You're really trying to do something
[28:42] that most people uh just really uh make
[28:46] a mistake with on the bearish side too.
[28:48] So this is another one that comes out in
[28:51] terms of the capex
[28:55] right here at this point again before it
[28:59] got to this because this includes the
[29:00] final quarter. You have to remember that
[29:03] people didn't believe this was going to
[29:05] happen. So now they're believing it up
[29:06] here and now they're believing this is a
[29:08] guarantee. The reasons that people were
[29:10] negative on this, some of it still is
[29:14] true. The bottlenecks are here. If we
[29:18] don't do seven, if we do 785 this year,
[29:20] but we don't do this number until this
[29:22] year or we do only twothirds of this
[29:24] because we don't have the data centers
[29:26] to build it, the problem is going to be
[29:28] is we've already built in this whole
[29:30] future. We've already built it in,
[29:32] meaning we're we're ahead of the game.
[29:34] We have no cyclical shelf. And for those
[29:36] people who say, "No, Micron is still
[29:38] cheap. SKHEX is still cheap." I get it.
[29:41] But the reason they're still cyclical is
[29:43] because there can be solutions. They can
[29:47] be algorithmic. They can be efficiency
[29:49] gains. The problem is if this bottleneck
[29:51] delays the ability to build it out,
[29:54] there will be solutions. The fact that
[29:57] memory is so expensive will lead to
[29:59] solutions. That's to a degree what Sarah
[30:01] was. And if you don't think there's risk
[30:03] of this, Meta is building dozens of
[30:05] massive tents at campuses across the US,
[30:08] sticking billions of dollars of chip
[30:10] inside and powering them off off-grid
[30:12] turbines. The IRS has officially entered
[30:14] the Mad Max phase. I've heard this so
[30:17] many times now. I've heard it from
[30:19] people this week, not in X, but people
[30:21] saying the hyperscalers are in shock at
[30:24] what they have to do now to get the
[30:26] power.
[30:28] JP Morgan on 27 data center buildout.
[30:30] The latest analysis based on satellite
[30:32] images shows that over 60% of data
[30:34] capacity planned for completion in 2027
[30:37] has not begun. This is the key thing.
[30:40] Anthropic urges, global pause, and AI
[30:41] development flag self-improvement risk.
[30:43] Now, you can read this and you could go,
[30:45] "Oh my gosh, a pause in AI development.
[30:48] That's what caused the market to sell
[30:50] down." That's not the important thing in
[30:52] there. Two bearish implications for
[30:53] memory makers from the anthropic cell
[30:55] improve self-improvement article. AI
[30:57] developers could slow because of safety
[30:59] and government's concerns. So, that's
[31:00] one side. You've also got Trump posting
[31:04] maybe Bernie Sanders is right. David Sax
[31:06] is posting it. Maybe we should give
[31:08] people part of the AI companies. I don't
[31:11] know what that would mean to them being
[31:12] as aggressive and building out if the
[31:14] government is taking a stake. I don't
[31:16] know. It's one of the risks that I've
[31:17] talked about in here with the
[31:18] hyperscalers is should they be trading
[31:20] at the multiples they are when it is
[31:22] obvious that the government either this
[31:23] administration, the next one or someone
[31:26] will be involved with memory names. It
[31:28] is impossible for me to not believe that
[31:30] somewhere in South Korea there won't be
[31:32] a politician that realizes we should be
[31:34] taxing these. There are a lot of
[31:37] negatives which should lead to multiple
[31:39] compression, but this is the one that is
[31:40] the biggest negative for memory.
[31:42] Recursive self-improvement could allow
[31:44] AI to solve its own memory bottleneck.
[31:45] So, I'm going to go back. I'm going to
[31:47] just make sure you guys read this flag
[31:49] self-improvement risk.
[31:52] The easiest way to solve the memory
[31:55] problem
[31:56] is for RSI. The bearish memory risk from
[32:00] RSI is two-sided. Safety concerns could
[32:02] slow the build out. While recursive
[32:04] self-improvement itself could accelerate
[32:05] the discovery of memory efficient
[32:07] architectures that reduce the amount of
[32:08] HPM needed for pure intelligence. When
[32:11] this starts happening, you won't be able
[32:13] to get out fast enough. So that is my uh
[32:16] logic that is my not even guess that is
[32:20] the reality that you have to build into
[32:22] the riskreward which is why for me and
[32:24] Micron that risk is far greater than
[32:29] investing in Marll. The optical side is
[32:31] in the very early stage and that is why
[32:34] Marll which Jensen Yuang said will be a
[32:37] trillion dollar company. It's 250
[32:38] billion. Give me that one all day long
[32:40] guys over Micron at this point. That is
[32:42] the reason why more firms turn to
[32:45] deepseat. This is the other issue for
[32:46] memory.
[32:48] More US firms turn to China's deepseek
[32:50] over pricey Silicon Valley.
[32:55] In fact, I didn't read the bottom part
[32:57] there for you. Chinese artificial
[32:59] intellig took the top spot on a major US
[33:01] business spending index in June as more
[33:03] companies swap out expensive American
[33:05] options like open anthropic in favor of
[33:07] more. If we get any sign on those
[33:11] parabolic charts for anthropic of a
[33:13] slight peak, a slight peak, if all of a
[33:17] sudden in July you look and it's the
[33:19] same RR, what's going to happen to
[33:21] memory? What's going to happen to the
[33:22] entire AI thought process? I'm telling
[33:25] you right now, do not take these and
[33:27] extrapolate them. There are a lot of
[33:29] issues involved. Do I think S&P earnings
[33:32] are going to be good? Yes. Do I think
[33:33] they're going to continue at around
[33:34] where they are? Yeah, we just blew out
[33:36] numbers. we're probably likely to beat
[33:38] numbers. But do I think that that is a
[33:40] place to have your money? No. I want to
[33:41] move to the application layer. Now, the
[33:43] application layer is the benefits that
[33:44] are coming from recursive
[33:46] self-improvement. For those of you who
[33:48] don't follow my Substack and my um
[33:51] subscriber at this point, I'm trying to
[33:53] help you navigate through exactly this
[33:55] stuff. For all of you on the
[33:56] institutional side that have now jumped
[33:58] in, trust me, you've jumped in at a much
[34:01] more dangerous time than it was in
[34:02] October. And that is the reason why I'm
[34:04] trying to talk like this. Edge open
[34:07] source substitution risk. Users migrate
[34:09] towards cheaper deepseek like open
[34:10] source specialized or ondevice models
[34:12] reducing dependence on centralized
[34:14] frontier model. Everyone that I talk to
[34:16] that has a brain is figuring out ways to
[34:18] do this stuff. I already have a Chinese
[34:20] model on mine. The only reason I don't
[34:21] have deepseat yet is because the
[34:23] hardware isn't there for me to have it.
[34:25] But the hardware will be there. That's
[34:26] why I want to be on the hardware side.
[34:28] But it's not the memory side, guys. It
[34:30] is just the hardware.
[34:33] first podcast to listen to Noris Bank
[34:35] Nikolai
[34:37] interviewed
[34:41] the head of IBM
[34:44] he says that some of the AI
[34:46] infrastructure buildout is a bit ahead
[34:47] of what the world can tolerate the next
[34:48] few years he estimates that one gawatt
[34:50] of AI data center power requires roughly
[34:53] 60 to 80 billion dollars of
[34:55] semiconductors to populate it just of
[34:57] the semiconductors okay remember it was
[35:00] only 50 billion for the buildout out a
[35:02] couple months ago. He does not think the
[35:05] revenue pool is large enough yet to
[35:07] support all the capex. He's doing all
[35:09] this based on a very simple equation of
[35:11] this is how much it costs. This is how
[35:13] many gigawatts we need. This is how much
[35:16] infrastructure would need to be built
[35:18] and do we have enough revenue. Now a lot
[35:20] of these companies have to build the
[35:21] revenue out of for no other reason for
[35:22] the cloud side to be able to get the
[35:24] clients assuming they can get them. Some
[35:26] will disappoint, many will thrive. In
[35:28] other words, the first phase was about
[35:30] securing chips, power, land, and
[35:31] capital. The next phase will be about
[35:32] proving utilization, pricing power,
[35:34] customer ROC, and payback. Eli Liy,
[35:38] other places in there that are
[35:39] benefiting from the ROIC. It's going to
[35:41] be a very different world coming out.
[35:42] Frontier models may become more
[35:44] commodity-like than people expect. If
[35:46] customers can move between models, which
[35:48] I do every single month, it seems like I
[35:50] am off of Claude as my dominant one, and
[35:53] I've moved to GPT 5.5. And everyone who
[35:56] argues with me, I think at this point
[35:58] you have just become too connected to
[36:00] Claude and many people were too
[36:02] connected to Chat GPT. I like to shift
[36:04] based on what's going on and I use all
[36:06] of them every single day. The capital
[36:08] cycle of the adoption cycle, he says, is
[36:10] still early, no longer the first inning,
[36:12] but maybe the second inning.
[36:14] The Bears are wrong if they think AI is
[36:16] just hype. But the Bulls may also be
[36:18] early if they assume enterprise adoption
[36:19] will happen instantly. I could not agree
[36:21] more. I think everyone is starting to
[36:22] get disappointed with the cost.
[36:25] There's no doubt that this is becoming
[36:26] more of an issue. AI might compress the
[36:28] cycle again, but he still thinks it will
[36:29] take years, not months. That is the key
[36:31] tension. The capex cycle is behaving as
[36:33] if demand is immediate while the
[36:35] enterprise adoption cycle still has to
[36:37] pass through the trust integration
[36:39] measurement and it's trying to do it
[36:42] with cost rising rapidly. The biggest
[36:45] investment implication is that the
[36:46] market may move from a capacity landra
[36:48] phase to an ROI ROI discrimination
[36:50] phase. So that's where I'm going to
[36:52] leave that at this point. Jensen Yuang
[36:54] if you didn't believe uh Krishna's uh 60
[36:57] to 80 for this chips Jensen says the
[37:00] cost per gigawatt is going higher 80 to
[37:02] 100 billion per gigawatt into AI factory
[37:04] from 50
[37:06] each one of these was a 20 to 30 then it
[37:09] was 50 to 60 and soon it will be this
[37:11] all of this while the adoption is way
[37:13] behind the cost
[37:16] way behind
[37:18] Sam Alman says AI budgeting has recently
[37:20] become a huge issue for some companies
[37:22] never came up last last year. Company's
[37:24] AI bills are bigger than ever and coming
[37:26] due.
[37:28] So, I wanted to bring this up not
[37:31] because I agree with Michael Bur. I want
[37:34] to make sure that all of you that are
[37:35] now bullish because you've had the
[37:37] Goldman Sachs research on what DRAM is
[37:39] going to look like going forward and
[37:40] what the capex numbers are going to be
[37:42] in the same way that he's using math in
[37:45] exponential time to argue for what's
[37:49] going on. You cannot go the other
[37:50] direction as well. Extrapolating 5 years
[37:54] in when AGI is now being said by
[37:57] Deisabis who as of a year ago was saying
[38:00] it's not going to happen before 2030.
[38:01] I'm leaning towards the early 2030s. Now
[38:04] he's publicly saying 2029. You've got
[38:07] anthropic saying recursive
[38:08] self-improvement. This is what I talk
[38:10] about in Eli Liy. This is what you want
[38:13] to be long now. You don't want to be
[38:15] playing the math game on capex, either
[38:17] on the bearish side or the bullish side,
[38:19] guys. Alphabet's record-breaking $85
[38:21] billion raise for Google AI business is
[38:23] a hell of a good signal. Hell of a good
[38:26] signal. The good signal is there's
[38:27] enough cash to do this. So, anyone
[38:30] worried about what's happening, they're
[38:32] buying this for the cap, they're doing
[38:33] this for the capex. The fact that
[38:35] they're front running the IPOs that are
[38:38] coming out, which are also in the game
[38:39] of getting the money for capex, is
[38:41] unbelievable. the fact that it was 85
[38:42] billion and barely budged anything and
[38:45] that is bigger than the I I think all
[38:48] the companies in the S&P except about
[38:50] the top 200 160
[38:54] the number 200 market cap is less than
[38:57] the raise that they did and it had zero
[38:59] impact and then on Friday Metaways big
[39:02] equity raise after Blockbuster Google
[39:03] and if this is going to happen guess
[39:05] what you know it's going to happen
[39:06] before the IPO does uh Goldman Sachs
[39:09] analysts are predicting this is what the
[39:10] IPO numbers is going to look like. We've
[39:12] already seen what the debt side is. The
[39:14] wall of supply, Paul Tudtor Jones
[39:16] talking about it. I'm not going to go
[39:17] through all the numbers, but let's just
[39:18] say there's a lot of stuff going on. At
[39:22] the same time that that's happening,
[39:24] this is the reality of what the World
[39:27] Semiconductor Trade uh group is saying.
[39:30] The WSTS forecasts that we're going to
[39:32] go from 90% year-on-year growth
[39:39] to 27
[39:41] And this is assuming no real
[39:42] bottlenecks. So the question is if
[39:45] that's the case, what's going to happen
[39:47] to these names? If the second derivative
[39:49] goes, how much of the market understands
[39:51] this? Now again, if you're a mutual fund
[39:54] and you're buying these with the
[39:55] expectation that three years from now,
[39:56] you're going to get money. They're in my
[39:57] thematic portfolios and they're going to
[39:59] remain there. Uh I think in the near
[40:00] term, they're going to need to be priced
[40:02] properly for the risk that I'm measuring
[40:04] out. I don't know how long that's going
[40:06] to take. It's either going to be through
[40:08] time or price. If Micron goes to 500, I
[40:11] will be involved in starting to buy a
[40:13] little bit back. If it goes to 400, I'm
[40:14] absolutely going to be buying it. I
[40:16] don't think it's going to do that. I
[40:17] think more likely we're going to look
[40:18] back four months from now, we're still
[40:20] going to be around the same price. I was
[40:22] wrong getting out at 700ish. It went up
[40:24] to a,000. That's a, you know, a big
[40:27] move, but Marll went up the same amount
[40:30] and is now outperforming with the move
[40:32] that we had there. The whole point was I
[40:34] didn't want to be in those things. Now,
[40:36] at the same time, your boy did buy
[40:38] silver and bought Bitcoin. That's gone
[40:39] down. So, not all of the things that I
[40:41] did worked out. Um, I'm I'm going to
[40:44] give these guys another shout out.
[40:45] Compounding Friends. This is without was
[40:47] without Josh Brown, but I think this is
[40:49] a good one to listen to on the topic of
[40:50] bubbles. Um, Michael Batnneck. I my
[40:53] appreciation for these guys just grows
[40:55] every week just because they get on
[40:57] there and they are not they're they're
[40:59] trying to do what I'm doing, which is
[41:00] provide facts to push back on extreme
[41:03] things. Not real things. It's extreme
[41:06] things like saying this is a bare
[41:07] market. The market's going to collapse.
[41:09] This is a bubble. That's an extreme
[41:11] statement. Agreed. Retail's over their
[41:14] skis a little bit here. There's too much
[41:15] gambling going on. What they talked
[41:16] about here is Michael argues that
[41:18] experience can actually hurt investors
[41:20] during a major regime changes because
[41:22] people often become experts in an
[41:24] earlier version of the world. I
[41:26] completely agree with that statement.
[41:28] The deeper point Michael is making is
[41:30] that experience is not only w is not
[41:31] always wisdom. In markets, experience
[41:33] can form can become a form of anchoring.
[41:36] Investors live through a traumatic
[41:37] period, learn a set of lessons, and then
[41:39] carry on those lessons forward as if
[41:41] they were universal laws. But markets
[41:42] are not static. I've talked about this.
[41:44] They're behavioral. They are biological.
[41:47] They adapt. The world changes, the
[41:49] dominant companies change, the source of
[41:50] earnings changes, and the cost of
[41:52] capital changes, and the economy
[41:53] reorganizes around new technologies.
[41:56] The investor who has seen this movie
[41:58] before may actually be watching the
[42:00] wrong movie. pattern recognition can be
[42:02] become pattern imprisonment. I think
[42:04] that's a good line. All right, another
[42:07] podcast, another keynote speech. Go
[42:09] listen to Jensen Yuan. Again, there's
[42:11] nothing new in it, but I think you need
[42:13] to listen whenever he speaks. Um, Frames
[42:15] AI is the new industrial system. AI
[42:18] factories that turn electricity into
[42:19] tokens. Ver Rubin has announced uh is
[42:22] launched. Uh, Nvidia and Microsoft are
[42:24] reinventing the PC era. So again, you're
[42:27] going to start getting more hardware out
[42:28] there, which should make it easier to
[42:29] run these bigger models, which means
[42:32] more on the edge side. So again, we
[42:35] start moving to the edge. Go back to
[42:36] what Gavin Baker said as the really big
[42:38] negative for the capex trade is if all
[42:40] of a sudden everything moves to the
[42:41] edge. Well, if the Chinese models keep
[42:43] staying up with the US models, and trust
[42:45] me, they're right there, everything's
[42:48] going to everything's going to shift.
[42:52] All right, Ver Rubin, you guys should
[42:55] make sure you're spending time on it.
[42:56] That's why I'm going to do the 800 volt
[42:59] DC. Uh that'll be a thematic uh index as
[43:03] well. I'm also going to try to do one on
[43:04] the application side. It hopefully will
[43:07] be diversified, but it may be
[43:08] specifically towards certain um by
[43:11] sector. In this case, healthcare is the
[43:13] one I'm focused on. Uh Dell connected to
[43:16] Ver Rubin and the first shipment was
[43:18] confirmed.
[43:20] Dell AI factories in the
[43:21] industrialization of intelligence. If
[43:24] you guys didn't read this, I would
[43:25] highly recommend reading it. I think
[43:28] Dell's earnings were the and and the
[43:30] company as a whole are one of the most
[43:32] important companies to pay attention to
[43:36] because of this part. So the chip part,
[43:38] the building of the data center part is
[43:40] not as interesting to me as the agentic
[43:43] side, which is about turning tokens into
[43:46] intelligence. And the intelligence is
[43:48] done through the agentic world. I think
[43:50] you have to again spend time on that. AI
[43:53] agents are moving closer to the edge.
[43:55] Dell gave a presentation on this.
[44:00] If you didn't believe in Dell last week,
[44:02] well then we got ULIP Packard Enterprise
[44:03] Company this week blowing out numbers
[44:05] and here's what happened to their stock.
[44:07] So if you thought Dell was a bubble,
[44:09] then you clearly thought that HPE was uh
[44:12] as well. And the what I wanted to show
[44:14] with this chart is okay. So here's the
[44:16] last decade now of Hulip Packard.
[44:19] This is not a bubble for one reason.
[44:21] Here's the total annualized equity
[44:25] return of 15%. It is still underperform
[44:27] the S&P even with this bubble-like move.
[44:29] The hardware side eventually this will
[44:31] be massively outperforming. I'm sure
[44:33] you'll see this at I would expect this
[44:34] to be at some point over 30%.
[44:38] eventually. Um, if you want to go
[44:40] through what they said, elevated prices
[44:43] to persist well in 27 for DRAM, uh,
[44:46] memory bottleneck is a multi-year issue.
[44:48] Uh, and these are the reasons why that
[44:51] people are involved. So, this whole
[44:52] setup for the DRAM thing is where this
[44:54] is going to be. But he also talks about
[44:57] they're explicitly saying they can push
[44:58] multiple rounds of price hikes into a
[45:00] tightening memory environment. So
[45:01] they're passing along the memory side
[45:03] and I think that's what makes them very
[45:06] very interesting to look at from what's
[45:08] happening. Um meeting European customers
[45:11] uh did not want the product because not
[45:14] one of them said they did not want the
[45:16] product because it was too expensive. So
[45:18] there's HP just again to look at it in
[45:20] the bubble side after not moving. Uh
[45:23] Marll Matt Murphy was on stage with
[45:26] Jensen Yuang.
[45:30] They talked about the partnership. They
[45:32] talked about all these things. Customers
[45:34] do not have to buy everything from
[45:35] Nvidia, but Nvidia still wants to be
[45:37] inside the data centers when custom when
[45:39] uh customers build custom asex. Um this
[45:43] whole thing of listening to it
[45:46] is very important because it goes
[45:48] through as the copper wall moves inside
[45:50] the rack. Murphy argues this will create
[45:51] an explosion in optical demand. Now the
[45:53] key thing is here Murphy argues this
[45:56] will create an explosion in optical
[45:58] demand. The next wave is co-ackage
[46:00] optics or CPO. This is why I keep saying
[46:03] this is the early stage of something
[46:05] that does not have as much competition.
[46:08] This is not just the fiber side. This is
[46:10] the actual movement side, the chip side.
[46:13] I wrote a paper again on March 30th on
[46:17] this. Again, the price was around $90 at
[46:19] the time. a powerful thematic trend that
[46:22] remains in the early index. This is two
[46:24] months ago.
[46:27] Here's the chart. This is before Jensen
[46:30] Yuang spoke. Then he said it was going
[46:31] to be a trillion dollar company. We
[46:33] finished the week at 260ish.
[46:36] Uh so up again. It was up almost 30% for
[46:39] the week. And then after market hours,
[46:40] it was trading up another 7%. Marll will
[46:43] become the next trillion dollar company.
[46:45] It's about 250 billion.
[46:48] Marll stock was just added to the S&P
[46:50] 500
[46:52] energy. I was buying Chevron and Exxon
[46:54] this week. Uh partly because I think the
[46:56] stocks are completely mispriced relative
[46:58] to what I'm about to show you in terms
[47:00] of where we are with Kouchy and
[47:02] everything on the prediction markets. Um
[47:04] I think these stocks are going to be
[47:06] necessary. I don't envision these being
[47:08] doubles and triples in a year, but I
[47:11] think energy at times you want to have
[47:12] it in there, especially now because if
[47:14] we don't get the straight solved in some
[47:16] way and it's still closed, these are
[47:18] going to be a hedge in the portfolio.
[47:20] Even if we get something that looks like
[47:22] some ships are going through, the
[47:24] reality is we're not going to get back
[47:25] to normal for a while. My guess is after
[47:27] looking at surveys on oil and how scared
[47:29] people are to be long, we're probably
[47:31] going to see oil prices either migrate
[47:32] higher or stay around the same level.
[47:34] And I think energy is going to become
[47:36] more attractive because the chip names
[47:37] are going to be more volatile. These are
[47:39] defensive safe names. I bought Exxon. I
[47:41] bought Chevron.
[47:43] Fluence had a big week up 44%. For those
[47:47] of you I've done calls with lately, I've
[47:49] talked extensively about batteries and
[47:51] the fact that this is going to be a
[47:53] place that two years from now you're
[47:54] going to wish you spent a lot of time
[47:55] on. Um because of all the problems we
[47:57] have, because Meta is using tents,
[48:00] batteries are a necessity. They are a
[48:04] necessity. I wrote a piece on this uh
[48:07] three weeks ago and the reason that this
[48:12] was such a big issue is because of the
[48:15] deal. Fluence energy will integrate its
[48:17] battery energy storage products into the
[48:18] data center designs for flexible power.
[48:20] It did a deal with Seammens and Nvidia.
[48:24] So they already mentioned their earnings
[48:26] which I highlighted that they have two
[48:27] hyperscalers
[48:29] uh involved. We signed master supply
[48:32] agreements with two major hyperscalers.
[48:34] The first one expected during the third
[48:36] quarter and then the seaman's
[48:38] relationship adds another layer
[48:39] validation. This is before the
[48:41] announcement. So this was me picking out
[48:43] what was already going on. I've had a
[48:45] high hit ratio with Nvidia announcing
[48:46] things like with Marll the day after I
[48:49] wrote the paper. I write the paper on
[48:50] Fluence and Nvidia does a deal with them
[48:52] in two and a half weeks. This is just me
[48:56] going through the earnings and paying
[48:57] attention to everything going on. I just
[48:59] wanted to show this because
[49:01] we've got CNI loans going up. We've got
[49:04] demand for loans going higher. We have
[49:07] the ISM PMIs going higher. I just want
[49:09] to make sure you guys realize that we
[49:11] are in the very early stage of this
[49:13] whole thing. I hear it from all kinds of
[49:15] places relentlessly. The nominal GDP
[49:18] side is there. And for anyone who writes
[49:22] negative things about what's happening
[49:23] in the economy, they are using the old
[49:26] cycle of capital and labor. Remember
[49:29] what I said, AI agents do not buy
[49:31] houses. They do not buy autos. Autos and
[49:33] housing and all the things that go in
[49:35] those things are very important to the
[49:36] way the LIE works. All of the people
[49:40] that were trained on housing matters. It
[49:42] doesn't matter anymore. This isn't a
[49:45] bare market in housing. This is called a
[49:47] stuck market. And I think it'll be stuck
[49:49] forever. Real estate stuck, private
[49:50] credit stuck, stuck, private equity
[49:52] stuck, VC stuck. I think all of them are
[49:55] stuck. Which is why tokenization will
[49:56] unleash twothirds of dormant assets and
[49:59] lead to another phase of what Raul Pel
[50:01] talked about. If you want to learn about
[50:03] the software side of the future, read my
[50:06] extensive multi multi-page paper on Eli
[50:09] Liy and start listening to everything
[50:12] I'm talking about in crypto. If you want
[50:14] to stick with the hardware trade
[50:15] forever, good luck because we are moving
[50:17] into the application side right now. Rio
[50:22] is the story. So, here's the paper I put
[50:24] out on Friday. I have spent about 10
[50:27] months working on this. It is very hard,
[50:30] as all of you know, to get up to speed
[50:31] on what's happening in healthcare. But
[50:33] GLP1s, I wrote papers about them at the
[50:36] hedge fund. So, I'm going to just give
[50:39] you the extreme hot takes that are in
[50:40] this paper. First, I believe by the end
[50:42] of this decade, Eli Liy could be the
[50:44] largest company in the US. Larger than
[50:46] Nvidia, larger than all of them.
[50:51] Eli Liy is becoming the most important
[50:53] AI company in the world. Peptides are
[50:57] this decade's API keys.
[51:01] So, the Mag 7 needed API keys to get
[51:03] there where we are. This is human
[51:06] software, guys.
[51:08] Eli Lily made new all-time highs
[51:12] on Friday.
[51:14] Lilyod is an early sign that the
[51:16] enterprise AI cycle is moving from
[51:18] generic models trained on the internet
[51:20] to specialized models. Read that line as
[51:23] many times as you need to to understand
[51:26] the difference between this and a
[51:29] general LLM.
[51:32] Eli Liy has a lily pod with a thousand
[51:36] plus Blackwell GPUs
[51:41] trained on their data, not on the
[51:43] internet. And that may become the
[51:45] defining mega trends of the return on
[51:47] investment phase of AI. This is the AI
[51:50] factories. This is it. and Eli Lily,
[51:54] as I write in the paper, go read all of
[51:56] the things that they've done, all the
[51:58] partnerships they've done, all the money
[52:00] they're spending right now, and where
[52:01] that money is coming from. They have the
[52:04] cash. Their revenues are growing 55%
[52:07] year-over-year.
[52:08] Hardware is now more of a risk.
[52:12] S&P 500 is up nine days in a row, nine
[52:14] weeks in a row. The reason I want to
[52:15] show this, this is not some ab some you
[52:17] uh some thing that you should fade.
[52:20] Historically, when you go through what
[52:22] happens in returns when you get this
[52:24] kind of a strong market, nine weeks and
[52:26] nine days on the back of earnings means
[52:29] this is something explosive. I don't
[52:32] know how explosive.
[52:34] I just know not to fade it.
[52:38] Crude, I'm going to finish up here. We
[52:40] cannot ignore this. It just continues.
[52:42] And unlike the tariff situation, I do
[52:45] not believe that the White House meant
[52:48] to be in this situation. And I do not
[52:50] think it is their choice to get the
[52:52] straight of Hormoose open. And everyone
[52:54] I think now gets that. So if Iran wants
[52:58] to keep playing a game and keep this
[53:00] closed, I don't know why that isn't a
[53:03] risk for the market that's greater than
[53:04] the way people think about it. So again,
[53:06] the reason I'm buying energy stocks
[53:08] along with Eli Liy, I like that Lily was
[53:10] up. I like that Exxon and Chevron were
[53:12] barely down on Friday. I think there's a
[53:14] risk here that as the trend and the
[53:16] charts keep going this direction. If you
[53:18] don't have energy stocks in there, you
[53:21] are playing a big risk with the other
[53:23] side of the equation because I think
[53:24] once you get a downtrend in a lot of the
[53:26] AI names, as I said before, if oil is
[53:29] going higher and rates are going higher,
[53:31] we have a new Fed chair coming in, he's
[53:33] going to be sitting in a difficult
[53:35] position. I don't know what's going to
[53:37] happen. Japanese crude oil reserves. So
[53:40] this is not just a US problem.
[53:43] Here are the probabilities on the
[53:44] straight of hormuz.
[53:46] When will it return to normal? Meaning
[53:48] before it was and again before October
[53:52] it's now down below 50%.
[53:55] So again normal is a different thing.
[53:57] We've taken off some of the supply. So
[54:00] I'm not bearish on the market. I don't
[54:02] think the straight of hormones matters
[54:04] nearly as much. And for the smarter
[54:06] people out there that I've heard speak,
[54:07] the pipelines and all the things that
[54:09] are being built will be ready in 3 to 5
[54:12] years. So the straight of Hormu's
[54:14] importance will go down. So for those of
[54:15] you who are extrapolating out, I think
[54:17] that means that Iran does have a reason
[54:20] to negotiate because if you can go
[54:22] around them, then they're really losing
[54:24] everything in including their oil. So I
[54:27] don't want to get bearish on this. But I
[54:28] do believe from an inflation perspective
[54:30] and the fact that it probably doesn't go
[54:31] back to normal quickly, you have to
[54:33] build that into what you're doing in
[54:34] your portfolio. And I think energy
[54:36] stocks and the fact that we're going to
[54:37] have a bid underneath is going to be
[54:38] there. If you want a good um balanced
[54:42] look at the oil market where they talk
[54:45] both about the negatives and the reality
[54:46] of what's happened and the positives, I
[54:48] think it should be here. I'm going to
[54:50] end on this note. In the same way I
[54:51] talked about recursive self-improvement,
[54:53] I am not worried at all about energy 5
[54:56] years from now. I'm not worried about
[54:57] nuclear. I'm not worried about any of
[54:59] those things because recursive
[55:00] self-improvement will have the ability
[55:02] of solving problems at a level which
[55:04] will allow us to not need as much energy
[55:06] at the same time and make it easier on
[55:08] the supply side. That's where we are.
[55:10] That's my story for this week, guys. Uh
[55:13] thanks to everyone who helped this week.
[55:14] Thanks to all the subscribers. Go to
[55:16] ai.22vresearch.com.
[55:20] I also put in the description every week
[55:23] where you can go to it. Let's go make
[55:26] some money.