Bottlenecks, Momentum Breaks, and the Next Phase of AI

[0:00] uh in Maine for the week. We'll be up
[0:02] here uh for at least a couple months
[0:05] beginning soon. Uh holiday weekend
[0:10] there. There I'm going to cover a lot of
[0:11] important stuff today uh in terms of
[0:13] some of the things that again uh that
[0:17] I'm going to keep talking about just
[0:18] because I think we're we're at and
[0:20] getting close to an inflection point. Uh
[0:22] and I want to make sure this is clear
[0:24] every single week. Uh from a trading
[0:28] perspective,
[0:29] we're in the middle of a secular trend
[0:32] in AI and not for any moment are you
[0:35] going to see me talk about the fact that
[0:37] this is not going to happen. Uh rather
[0:41] than worry about not having the
[0:44] financing,
[0:45] uh this is a bubble and once it collaps
[0:48] the whole economy is going to go down,
[0:50] you're not going to hear any of that. Uh
[0:52] this is a different time. It's being
[0:54] funded with companies that have the
[0:57] money and the revenues are coming.
[0:59] They're just on delay, but they're
[1:02] contracted and the momentum is uh it's
[1:06] never happened before. So, there are no
[1:09] free lunches. Uh markets are all about
[1:12] riskreward. We've had a huge run as I'm
[1:15] going to go through and I am seeing some
[1:18] global and sector correlation breaks
[1:20] which usually are warning sign that uh
[1:22] you should be careful. We had a very
[1:24] volatile week for momentum. Momentum
[1:26] shifts are usually around regime shifts.
[1:29] Uh I have a lot of podcast stuff in here
[1:32] to to listen to because I think we have
[1:35] to start building in that uh it's been
[1:37] now two months of talking about the
[1:39] straight of horses and how important it
[1:40] is and there's still no ships going
[1:43] through in any meaningful way and
[1:45] inventories are being drawn out. So
[1:47] again this podcast is about the factual
[1:49] side and now we have inventories
[1:50] breaking through the bottom end. We did
[1:52] have the Google IO. There's two reasons
[1:54] that this is important. One is to show
[1:56] you what went on last year. Another is
[1:57] to connect it back to crypto briefly. Um
[2:00] for the RAAS, uh I do and I am going
[2:04] through with Morgan Stanley on index
[2:06] launches. Uh I'm not planning to do an
[2:10] ETF on the AI thematic portfolio. But uh
[2:13] the 100 name portfolio I am working on
[2:16] getting uh two things. One as an index
[2:18] that's tradable not just for mutual
[2:20] funds, RAAS, but for hedge funds,
[2:22] options, everything along those lines.
[2:24] But most importantly in my opinion um as
[2:27] I go through this you have to understand
[2:29] that uh the outperformance AI is is
[2:32] driving the entire market. Uh in my
[2:34] opinion it's driving the entire economy.
[2:35] I'll debate any economist on that who
[2:37] wants to play the math game. If it
[2:39] wasn't for AI and the buildout, if it
[2:41] wasn't for the profit margins exploding
[2:43] for these companies uh and getting the
[2:47] wealth effect, I don't know what would
[2:48] be happening uh in the economy at this
[2:50] point that wouldn't involve needing the
[2:53] government to a greater degree. Uh I
[2:56] also have worked on and I'll show you
[3:00] the results of creating a concentrated
[3:02] portfolio uh which will go up at some
[3:05] point this week. Uh I've had a lot of
[3:07] RAAS that have discretion looking for
[3:09] something that's not a hundred names.
[3:11] Well, now you have the index uh able to
[3:14] be traded for anyone who has the ability
[3:16] to do ISDAS. But then you're also going
[3:18] to have a concent two concentrated one.
[3:20] One's of 10 names uh ones of 25. I spent
[3:23] a lot of time to basically put together
[3:25] a list that replicated to the highest
[3:27] degree I could uh what the portfolio is.
[3:30] So let's just go through this quick. the
[3:32] S&P
[3:34] uh up 88 basis points for the week.
[3:36] Rough start, good finish overall, strong
[3:40] week Q's again. And by the way, in the
[3:43] S&P, I didn't mention it, but you can
[3:45] see it here. Um we're extended. I mean,
[3:48] this is this is a lot of weeks in a row.
[3:50] We're going to get a correction at some
[3:52] point. I think it's going to be more
[3:54] than a oneweek correction
[3:57] uh when we finally do get it. Uh it
[3:59] wouldn't surprise me if it started next
[4:01] week. As we get into uh close to the end
[4:03] of the month, we've had a huge month and
[4:06] I think mean reversion should be on the
[4:08] top of everyone's mind. Uh especially
[4:10] since momentum showed weakness last
[4:12] week. Q's up 1.2%, IWM up 2.8%.
[4:17] The hyperscalers,
[4:20] they're basically unchanged now. So
[4:22] again, this is Meta, Google, Amazon, and
[4:24] Microsoft as an equal weight basket. Uh,
[4:27] and like I've said, I think these guys
[4:30] are the key to the market. As much as
[4:32] the AI trade is driving the alpha, as
[4:35] long as these names are bouncing since
[4:36] they are the spenders,
[4:39] everything's okay. Well, they broke
[4:40] below the 20-day moving average. It's
[4:42] unchanged pretty much over the course of
[4:44] the last three weeks. So, I would watch
[4:47] because this is telling me the market's
[4:49] tired. At the same time, Goldman make
[4:51] all time made all-time highs this week.
[4:53] uh with the S&P sitting basically at
[4:56] all-time highs still. Uh when that's
[4:58] going on, nothing bad is happening. So,
[5:01] you're gonna have to watch. I've shown
[5:03] this before. I'm going to show it again
[5:04] because this is the environment that I
[5:06] think we're going to be in. And this is
[5:08] the regime shifts that we're going to go
[5:10] back and forth with. We're not jumping
[5:11] into here. Whether it's positive or
[5:14] negative, it's over here. And if you go
[5:16] through this, we've got rising oil
[5:18] prices. We've got, as I'll show you,
[5:21] high mo high volatility with inside
[5:23] factors, specifically in momentum, not
[5:26] so much in the VIX and other places, but
[5:29] when things do go, I think it'll be
[5:31] correlated. As we've seen, you've got
[5:33] stock market multiple compression.
[5:35] You've got trap money and forced selling
[5:37] with inside the credit markets, and now
[5:39] you've got interest rates rising. So,
[5:42] again, this is a version of the 1970s,
[5:44] which I think is going to be a part of
[5:46] what you should expect.
[5:49] Last week I showed this which were
[5:51] basically show me uh the two times I put
[5:53] a monthly chart and a weekly chart uh of
[5:56] RSI. So monthly RSI, weekly RSI both
[5:59] which were above 70 last week. Uh on
[6:03] Thursday I ran it. So, Friday is when I
[6:05] did the video and I just highlighted
[6:07] that this historically is not a great
[6:10] time in terms of what goes on when you
[6:12] get momentum up there and that it either
[6:14] means we're at a top or we start to kind
[6:17] of build a top and then we head lower
[6:21] and it takes a while to get back up to
[6:22] the highs. Well, we did have a sharp
[6:25] move lower in momentum this week. The
[6:28] S&P didn't really follow. Uh this is on
[6:30] a close basis. We actually got all the
[6:32] way down here on Tuesday morning which
[6:36] meant we had given up a significant
[6:38] portion of the last month's uh momentum
[6:41] rise. And here you can see the rate of
[6:43] change. So I put these arrows in just
[6:46] because we've been on a trend now and I
[6:48] think this is going to happen because I
[6:51] believe we're at the beginning stage of
[6:52] a secular trend which mean momentum is
[6:55] going to continue to work uh over the
[6:59] course of time here. I don't think the
[7:00] names are going to change that much.
[7:02] I'll go through why that is. Uh because
[7:04] we're at the beginning of a buildout for
[7:06] the agentic world. This is a
[7:07] once-in-a-lifetime opportunity over the
[7:09] course of the next 3 to 5 years to pay a
[7:12] sec to play a secular trend which will
[7:14] remind us somewhat of the commodity bull
[7:16] market that occurred in China when China
[7:18] was building out from 2002 to 2007.
[7:22] Nothing could stop that train. Um and it
[7:25] ended when the US blew up. Uh but if you
[7:29] remember everything that was going on,
[7:30] it was a lot of money being spent. So we
[7:32] are seeing continued unwinds. This is a
[7:35] two-day rate of change that the numbers
[7:38] are just getting bigger on both sides.
[7:40] And I think that's what you're going to
[7:41] have to get used to as this trend
[7:43] continues to extend. Um we've got a
[7:45] historic breakdown in correlations
[7:46] between software and semi. So, if we're
[7:48] going to get a momentum sustainable move
[7:51] the other direction, it's going to
[7:52] include software that is either not
[7:54] going down uh or is going up while semis
[7:58] are coming down. And that's what I'd be
[7:59] looking for. I wrote this for 22V this
[8:01] week. Um
[8:03] it basically just went through what I
[8:05] see as the issue. And again, I'm going
[8:06] to say it again and again. I don't think
[8:08] this ends because of a bubble. I don't
[8:09] think this ends in the way that most
[8:11] people are writing about.
[8:13] uh meaning I I I've done this for a long
[8:16] time in terms of thinking about the
[8:18] things in go. When we had Amazon and we
[8:21] had the Mag 7 that were going up, it
[8:24] didn't ever end. You just got periods of
[8:26] time where they'd have corrections, but
[8:28] nothing stopped them from demolishing
[8:30] the other companies and creating moes
[8:32] around it. It just continued. So, I
[8:34] think this is going to be in there, but
[8:36] the risk is that at some point we hit
[8:39] bottlenecks. Uh hoarding happens. we
[8:41] don't have the physical world like we
[8:43] did with software where nothing could
[8:45] stop the growth. You can stop the growth
[8:48] uh basically by not being able to build
[8:50] the stuff anymore and by holding all
[8:51] this inventory. So that is the risk that
[8:54] I think will happen. Now I wanted to
[8:56] just highlight that one of the
[8:57] correlation breaks that's happened
[8:59] there's two parts to the AI trade. Um
[9:01] when I put together the thematic
[9:03] portfolio of 100 names it was really
[9:06] broken down into two sectors. Um and if
[9:08] I said two sectors, it's obviously heavy
[9:11] semiconductors,
[9:13] but then so technology is part of it,
[9:15] but the other side of it is industrials.
[9:18] Um and even with the power side, the
[9:20] majority of the names, if you guys have
[9:21] gone through the power side, they're
[9:22] industrial names are things like
[9:24] Caterpillar and Eaton. Um so momentum
[9:27] unwinds and industrials are very
[9:28] important. And this is what happened in
[9:30] the industrial momentum side. This is
[9:32] through Thursday. Some of the charts on
[9:34] this stuff are through Thursday. Friday
[9:36] was not a big day for anything. Uh so I
[9:39] didn't want to go back and make all
[9:40] these changes for nothing since I put
[9:42] some of these slides together on
[9:43] Thursday. But momentum in particular on
[9:45] industrials which is still down here.
[9:47] SMH has gone back to the highs. So this
[9:49] is the correlation between these two and
[9:51] this is the problem. So you're getting
[9:54] the long side of momentum in
[9:56] industrials. This is the long side of
[9:58] momentum. The Morgan Stanley
[10:00] Industrials momentum. This is the actual
[10:03] Morgan Stanley momentum side. So this is
[10:05] both of them. You can see this head and
[10:07] shoulders pattern. This is what happened
[10:09] to the overall momentum in industrials.
[10:12] It stayed down and we have a MACD sell
[10:14] signal on this in terms of what's gone
[10:16] on. So I think momentum um is showing
[10:20] signs that maybe we've run too far and
[10:24] we need some kind of profit taking. And
[10:26] whenever we get to the end of the month
[10:28] when you've had a huge month for things,
[10:30] I kind of worry about people taking off
[10:32] positions either in that week or the
[10:34] following week. So I think the risk is
[10:36] there for for more momentum uh pain. Uh
[10:39] I just wanted to include this. I did
[10:40] this at the beginning of the year. This
[10:42] came from a Michael Seymble piece just
[10:44] he brought up the fact that 40 names
[10:46] were driving earnings grow uh driving 70
[10:49] to 80% of earnings growth. Not the
[10:51] excess earnings growth, the actual
[10:53] earnings growth. uh that was coming into
[10:55] the year and they were related to AI and
[10:57] capex and I basically did this to say
[11:00] that um correlation spikes in stress
[11:03] periods you're going to get all of these
[11:04] moving together meaning all of the
[11:05] things creating the alpha will move
[11:08] together and that's one of the issues
[11:09] that you look for is if that's the case
[11:12] you have to start looking for breakdowns
[11:14] and correlations and here's the first
[11:16] chart of a broad-based global breakdown
[11:19] so each one of these is a different
[11:21] market around the globe the S&P is back
[11:24] up near the all-time highs here, but the
[11:27] Russell is is is is still below the
[11:29] highs. It did bounce and it's getting
[11:31] closer to the highs, but you have this
[11:33] situation where the US markets are here.
[11:35] These other markets here, it's European
[11:38] markets with the CAS and the DAX. It's
[11:40] Asian markets with the Hang Sang, the
[11:42] Indian market and Jakarta and Australia.
[11:46] So, you got Indonesia, Australia, and a
[11:48] lot of these. Now, this is all
[11:49] normalized over the course of uh the
[11:52] last year, meaning they're all together.
[11:56] These are big divergences that are going
[11:58] on. If you look at Indonesia and again,
[12:00] it went down and then you had and this
[12:02] is the key point. When you have oil
[12:04] going higher and now you have rates
[12:06] going higher around the globe, you're
[12:07] starting to get these divergences. So,
[12:09] initially the oil market, it was a fall.
[12:12] We get the bounce in the US market to go
[12:14] up. The Jakarta market goes up. All of
[12:17] these markets go up and now they're
[12:18] starting to roll over again. And that's
[12:20] the risk that I see. I see contagion
[12:22] amongst the down move. I see everything
[12:25] having a hard time going higher. I
[12:27] talked about the importance of
[12:28] industrials. This is the world
[12:29] industrial index. Again, it peaked and
[12:33] barely bounced this week. So, the
[12:35] industrials around the globe are showing
[12:36] weakness. And it would make more sense
[12:38] that they would be the first group to
[12:40] show the weakness over the
[12:43] semiconductors. So if you're going to
[12:45] have a problem with rates going higher
[12:47] and with oil going higher, you would
[12:49] expect it to be more of an Asian and
[12:51] European thing than a US thing. We're
[12:52] seeing that. You'd also expect the
[12:54] industrials to be to be hit hard.
[12:56] They're the manufacturers and very
[12:58] important to the data center side.
[12:59] They're very important to all this. And
[13:01] if they're not getting in what they need
[13:02] on the natural gas and the oil side or
[13:04] any of the chemical derivatives of that,
[13:06] they're going to start to have issues.
[13:07] And I always believe that the market is
[13:09] going to show where the pain uh is going
[13:11] to be beforehand. This is that world uh
[13:14] industrial index overlaid with the S&P.
[13:16] So again, the S&P has gone to new highs,
[13:19] gone higher, while this world industrial
[13:21] is lower than it was basically before
[13:25] the the war started. So you had gone up
[13:28] into the war, then you came down, then
[13:29] you went up and you're below it. So we
[13:32] never took out the highs and then we
[13:34] failed there. So the industrials are
[13:35] being impacted. Here's that now with the
[13:38] Cosby machinery. My the the yellow name
[13:43] here, which is up at the highs. This is
[13:44] my thematic portfolio.
[13:47] And the red line here is indicating
[13:49] where we saw the breakdown in the Cosby
[13:51] machinery index and the long side of the
[13:54] industrial momentum side.
[13:56] Basically the industrial winners are
[13:58] under pressure. The thematic portfolio
[14:01] as a whole is up. A lot of those are are
[14:04] US names. I don't have any real names in
[14:07] the industrial side from Asia. So what
[14:09] you're left with is you're getting a
[14:12] breakdown here in these uh these names.
[14:14] This correlation has gone down. And to
[14:16] make it look a little bit more let's say
[14:18] severe with inside the country. I've
[14:20] shown this in the past couple weeks.
[14:21] This is through Friday. So this is the
[14:24] um the construction side in Korea, the
[14:27] construction sector. This is the
[14:28] machinery sector. They both peaked at
[14:31] different levels, went down and this is
[14:33] SKHEX. I've talked about and I did a
[14:36] webinar this week for subscribers where
[14:37] I went through how important Korea is.
[14:40] Korea is an important part of the
[14:41] market, but it's not just them. This is
[14:43] another market that's been important
[14:45] that's also showing signs. This is
[14:47] Japan. Um Japan is one of the markets
[14:50] that stands out the most to me. uh on
[14:52] the Asian side they or on the the AI
[14:54] side they come up in the chemical side
[14:56] they come up in the semiconductors they
[14:58] come up in all parts of this uh
[15:00] Mitsubishi from the gas turbine side
[15:02] they are heavily involved in multiple
[15:05] levels and the topics
[15:07] basically has rallied but now all of a
[15:09] sudden you're getting this divergence
[15:11] again just like we saw there it has been
[15:13] very correlated to my thematic portfolio
[15:16] but beginning towards the middle of
[15:17] April the thematic portfolio continued
[15:19] to rip higher a A lot of this has to do
[15:22] with what's going on in Korea. A lot of
[15:24] it has to do with SKHEX. A lot of it has
[15:25] to do with Micron. A lot of it has to do
[15:27] with the power semis. All of this stuff
[15:30] has been kind of impacted by what's
[15:33] happening in Korea. And this is when you
[15:35] got the breakaway. So in Japan, we've
[15:37] kind of broken away. Now, if you go
[15:39] inside the Japan market and look at the
[15:41] machinery side, you have the same chart
[15:42] as you do in Korea. Basically went up
[15:45] and has gone back down. Very correlated
[15:47] to the thematic AI side. Worse than that
[15:50] is the construction side. So, not only
[15:52] does Japan have the issue right now of
[15:55] oil, it also has the issue of long-term
[15:58] rates going higher. And this is another
[16:00] part of the equation that people have to
[16:01] start factoring in. Remember, since the
[16:03] straight of Hormuz was shut, we've not
[16:06] only had oil prices go up significantly,
[16:08] we've seen a Fed, which was expected to
[16:10] have three rate cuts over the course of
[16:12] the next 12 months, and now we have one
[16:13] rate hike over the next 12 months. So,
[16:16] we've shifted 100 basis points out of
[16:18] that. We've seen 10 tenure rates
[16:19] basically go up uh about 75 basis points
[16:23] from where they were before the war. And
[16:25] so in countries like the UK and Japan,
[16:28] you want to watch to see how things like
[16:30] the construction market are doing, which
[16:31] are obviously more sensitive to
[16:33] long-term rates. We're breaking below
[16:34] the 200 day moving average. And to put
[16:36] that in perspective, since Chat GPT came
[16:38] out, which was in here, except for this
[16:42] panic during the yen unwind and during
[16:44] liberation day, we haven't been below
[16:46] the 200 day moving average. Well, we're
[16:48] below it now as of Friday. Um, and this
[16:51] is the topics overlaid with those. So,
[16:53] the topics has hung in. And again, when
[16:55] I start seeing correlation breaks on
[16:56] what had been driving the market, I
[16:58] start paying attention. So, you've got
[16:59] more there. This is the machinery side
[17:02] in Japan relative to the my thematic
[17:04] portfolio. So the AI side. Now if you
[17:07] take it even further and you go in the
[17:08] US and you say, "Okay, here's two-year
[17:10] rates overlaid with the S&P." So the S&P
[17:12] is ignoring again the rate side. It's
[17:14] also ignoring the currency side. The
[17:16] dollar has been strengthening
[17:17] continuously since we took out the rate
[17:20] hikes, the rate cuts. So we now have the
[17:22] dollar going, which is generally a
[17:23] tightening liquidity situation. You also
[17:26] have oil. Despite everything happening
[17:28] in the straight, all these promises
[17:29] about the deal, we don't have anything
[17:31] yet. And if you're living in facts,
[17:33] eventually you can't keep this game
[17:35] going on. You start to become an
[17:37] inventory problem. You go from a supply
[17:40] uh glut to a supply shortage at some
[17:42] point. And based on everything that I'm
[17:44] going to show you and everything I've
[17:45] heard, I listened to probably three or
[17:48] four podcasts this week to listen to
[17:50] four different people on the oil patch
[17:52] that have been around a long time. And
[17:54] the message is pretty similar. and
[17:55] rather than the doomer side which was
[17:57] coming out initially uh after uh which
[18:01] was focused on how bad it'll get. Well,
[18:03] now we're in the reality side where the
[18:05] inventories are being drawn not just in
[18:06] the US but across the globe and it
[18:08] matters.
[18:09] Google IO this is a chart I just wanted
[18:12] to show because I want you to pay
[18:14] attention to here. Um last year at this
[18:17] point in a span of a year we had gone
[18:19] from about 9 trillion to 480 trillion.
[18:23] Now we've done seven times that growth.
[18:25] This is in tokens process. So just think
[18:27] of this as compute usage. And just look
[18:30] at this parabolic move. So this was
[18:32] reported this week. The reason I want to
[18:34] bring this up and this is the other
[18:35] reason why I think people should be wary
[18:38] of just where we are now. I wrote this
[18:41] paper on May 15th a year ago. I could
[18:44] not get a single person to care about
[18:45] this at that time. Uh not one person. I
[18:48] spent a lot of time with institutional.
[18:51] Everyone was busy debating tariffs.
[18:53] Nobody cared about how inference was
[18:55] made. It was mentioned. And this was
[18:58] from the Q1 earnings reports where I
[19:00] talked about how many companies
[19:02] specifically mentioned what we just saw
[19:06] here. This is May of last year. So we
[19:08] had seen this dramatic uptick. No one
[19:11] cared a year ago. And I highlighted
[19:13] these are the names if you want to play
[19:15] the inference breakout. Nvidia, AMD,
[19:18] Intel, Qualcomm, Lattis, Micron, Western
[19:20] Digital, Seagate, Sienna, Marll, every
[19:23] single one of them. Okay, I created an
[19:26] equal weight basket off today of them.
[19:28] Here are the returns of them since I
[19:30] wrote the paper
[19:33] and now people are bullish these names.
[19:35] Yes, this is the reason why when you get
[19:38] through everything and you're seeing
[19:40] kind of a bubble formation, this is my
[19:42] inference side. So if I create an equal
[19:45] weight basket of these names, these 12
[19:47] names that I mentioned, this is what the
[19:49] chart looks like. Okay, so a year ago,
[19:53] here's where we are.
[19:56] This is why when I say I've bailed out
[19:57] of my micron, a lot of that has to do
[20:00] with why wouldn't you get out of a chart
[20:01] that looks like this? Yes, it probably
[20:04] goes higher and it may go higher because
[20:05] there's no way you can pick. I've been
[20:07] out of this and in this area, but here's
[20:10] the thing. If you want to judge retail
[20:12] into this, this is the Roundill DRAM uh
[20:16] ETF which was created in April which has
[20:18] $10 billion. I listened to Yan Van X
[20:21] speak today or speak yesterday on
[20:23] Compound in France talk about SMH which
[20:25] has been around forever has about 60
[20:27] billion AUM. This thing I I I heard is
[20:30] up to 10 billion in AUM uh at this point
[20:34] in two months. This is retail and it's
[20:36] very correlated to what I wrote about
[20:38] again back here. So, I think this is an
[20:41] indication that maybe everyone's jumping
[20:43] into something that's already had its
[20:45] move and everybody's in at this point.
[20:47] And even though I don't think
[20:48] institutions are in a big way, I do
[20:50] think that you got to be careful right
[20:52] now that if it turns the other direction
[20:54] for any reason, and I'll go through some
[20:55] of those possible reasons, it's an
[20:57] issue. Um, here's another problem right
[21:00] now is I don't use a lot of different
[21:02] things for sentiment, but I do think
[21:04] this Goldman Sachs Rick Goldman Sachs
[21:07] risk appetite indicator is one that I
[21:09] look at just as a general gauge. And
[21:11] again, it's at the highest readings
[21:13] since 2021. So, don't that's no reason
[21:16] to get bearish, but it is a sentiment
[21:18] indicator that I like to look at when
[21:20] they post things in X about it. Um, this
[21:23] is the reason everyone is so positive.
[21:25] And again, I've said it before, I'll say
[21:27] it again.
[21:29] Maybe we just did two quarters of orders
[21:32] and maybe the next one's going to come
[21:34] in here. When you're up at this high, so
[21:36] this is net income for Micron.
[21:38] Obviously, they've been able to price
[21:40] this at any price. They've had huge
[21:42] orders, but I mean, I don't want to go
[21:44] count the stuff of what you're talking
[21:46] about, but you're talking about
[21:47] basically the last however many years of
[21:49] total net income. That was for the
[21:51] quarter.
[21:52] So again, we'll have another reading in
[21:55] I think it's six weeks. But this is the
[21:58] other issue. Um profit margins in the
[22:00] S&P are almost entirely seven stocks.
[22:03] And again, profit margins are good by
[22:05] the different sectors, but the majority
[22:07] of the profit margins which are gear
[22:09] have taken this S&P up to these levels.
[22:11] It's being driven
[22:13] basically not just by the MAG 7 and
[22:15] that's where the focus of this is
[22:16] because that's what BFA did. But if you
[22:18] add in Micron and you add just put
[22:20] Micron in there and SanDisk and Western
[22:22] Digital,
[22:24] this this is a lot of them. This we
[22:26] don't have margins going up for the S&P
[22:28] at this point and especially with oil
[22:30] going higher. It's the only thing I'm
[22:32] saying is if all of a sudden the margins
[22:34] went down, which wouldn't be bad, but if
[22:36] it it went down because we had hoarding
[22:37] in the first order in the first quarter
[22:40] and there was so many shortages and then
[22:41] all of a sudden the demand starts to go
[22:43] down because people have either
[22:44] overbought or the costs have gotten too
[22:45] high, you'll see profit margins come
[22:48] down. And if profit margins come down, I
[22:50] just think the quant strategies and
[22:52] everything will see margins come down,
[22:54] which is usually a bad indicator for the
[22:56] market as a whole. and you might start
[22:58] to get some unwinds of positions or a
[23:00] change in certain kind of risk. Warren
[23:02] PI's put this out this week. This is
[23:04] another thing that again um the forward
[23:07] sales growth is projected at 18% over
[23:10] the next 24 months. Again, I think the
[23:13] orders are going to come in. I just
[23:14] think you're going to have lumpy
[23:16] problems along the way because you're
[23:18] going to have bottlenecks. Um I don't
[23:19] know how much you can order just keep
[23:21] ordering, ordering, ordering if you
[23:23] don't take delivery of stuff. We already
[23:25] see that delay in terms of the RPOS for
[23:29] Oracle. We see that for the RPOS for
[23:32] pretty much everyone. I mean, in Nvidia
[23:35] used to have a big receivable issue.
[23:36] They still have big receivables, but
[23:38] they're changing it to make sure they
[23:39] get cash. And that's because everything
[23:41] was in the future. Same thing goes with
[23:44] Microsoft, Google, and Amazon cloud
[23:47] where a total of $1.4 trillion. They're
[23:50] going to get it at some point. They
[23:51] don't know when. Um they're hoping it
[23:53] comes soon. Wall Street Journal had an
[23:54] article today on how the AI chip mania
[23:57] sews the seeds of its own destruction.
[23:59] Gets into the cyclicality side. But
[24:01] again, it just shows the analyst
[24:03] forecast income for this year. I showed
[24:05] you the 13 billion
[24:08] and then for for next year again a risk
[24:12] is that supercharged profits attract new
[24:14] rivals to enter the market. I am not
[24:16] worried about competition by itself. And
[24:19] I want to make this clear. The
[24:21] competition side has a bunch of issues
[24:25] with it, but most of them have to do
[24:28] with the fact that I think the delays
[24:30] that are likely to happen prevent the
[24:32] capacity from coming online quickly
[24:34] enough because the orders are just too
[24:36] big. So regardless of whether it's a
[24:39] timing issue or it's an actual
[24:40] competition issue, there's a lot of
[24:42] things that can go wrong a year from
[24:44] now. Not that it's not gonna still be a
[24:47] good investment, but you're going at a
[24:49] pace where an eight bagger in a year at
[24:52] some point it's going to slow down and
[24:54] at some point if it slows down the
[24:56] question is do people want to be long a
[24:58] stock that they only think is going to
[24:59] go up 50% when now it's bigger size.
[25:01] Remember I've shown you that Nvidia's
[25:04] multiple has collapsed over the course
[25:06] of the last two and a half years. Uh and
[25:09] their earnings have been great. Um this
[25:12] got everyone excited this week. um
[25:14] probably led back to the bounce. Uh this
[25:16] is the dollars to be spent in the Vera
[25:19] Rubin on memory 435%
[25:23] just to give you an idea of where the
[25:26] change is as you go through and how
[25:28] people get excited on the numbers that
[25:30] are just heavy growth periods in this in
[25:32] terms of dollar size
[25:34] and again that assumes that Ver Rubin
[25:36] and the buildout actually
[25:39] happens in the time. Um, so Zero Hedge
[25:43] kind of went through it. They're
[25:44] focusing on the memory cost in Ver
[25:46] Rubin, but the US power grid can barely
[25:49] keep up with the current setup. How are
[25:51] we going to get them more? And that is
[25:53] the issue. And remember, Nvidia was
[25:55] hunting through Korea to try and find
[25:57] help on the industrial side and we have
[25:59] the industrial side breaking down right
[26:01] now. So this is the connectivity between
[26:04] Vera Rubin and the needs of the Korean
[26:06] side for the Vera Rubin side. They're
[26:09] all connected. Um, data center pipeline
[26:11] and gigawatts has tripled during the
[26:13] past year. That's great. We get more and
[26:15] more orders for things, but that just
[26:17] means the bottlenecks continue. Gartner
[26:19] AI spending forecast. Remember, it was
[26:22] only six months ago that people didn't
[26:24] think we'd get any spending, that it
[26:26] would be cancelled. And now what we're
[26:28] talking about for 2027, the total AI
[26:31] spending will be 3.5 trillion. Now again
[26:34] this is three and a half times the size
[26:36] of what's expected for next year about a
[26:38] trillion dollars for the just for the
[26:41] data center side. So again the AI
[26:43] infrastructure spending service spending
[26:45] software spending cyber like these are
[26:48] big numbers guys and this is the reason
[26:49] why even if it comes in at 2.5 trillion
[26:52] this is this number is is more than 10%
[26:55] of GDP. So again these are these are big
[26:58] numbers and orders. So, inside the
[27:00] biggest bet in corporate history,
[27:04] I thought this was interesting just
[27:05] because of this chart. So, first of all,
[27:09] the year-over-year growth in capex
[27:10] expected to be here. Now, again, they
[27:12] keep revising these numbers higher. So,
[27:14] who knows how high it'll be, but over
[27:16] the course of the next eight years, the
[27:19] reason people are now fully excited. So,
[27:21] when I wrote the inference piece a year
[27:22] ago, I want you to remember no one
[27:23] wanted to talk about this chart. Think
[27:25] about it. A year ago today, nobody
[27:27] wanted to talk about this chart. And
[27:29] this chart is now higher than it was. So
[27:31] last year they doubted that we could do
[27:33] a linear line that looked like this. Now
[27:35] they're doub they're jumping in because
[27:38] they they accept a higher number. I've
[27:40] just seen this too many times in my
[27:42] career where there's going to it's just
[27:44] not going to work out that easily uh in
[27:46] terms of this. And I think we're
[27:47] starting to see the pain points because
[27:49] rates are moving higher, oil's moving
[27:50] higher. We've already had input costs
[27:52] between micro memory uh CPUs, silver,
[27:56] everything that goes into building. And
[27:58] guess what? The most important thing
[27:59] about this $8 trillion, the most
[28:02] important thing, this is what it looks
[28:05] like in terms of the buildout that needs
[28:06] to happen. It's that complex. We have a
[28:09] ton of things that need to get built.
[28:12] How far into the buildout are we? By the
[28:15] end of this year, we will have only
[28:18] spent about 18% of what's expected. The
[28:22] physical system is already stressed
[28:24] before the largest dollar years even
[28:26] arrive.
[28:28] Is this likely to be smooth? No. The
[28:30] chart is smooth. The buildout will
[28:31] almost certainly not be. And that's
[28:33] because this is not a software capex. It
[28:36] is a physical world capex cycle. It
[28:38] needs HBM chips, racks, liquid cooling,
[28:42] copper, fiber, substations. If any of
[28:45] them aren't around,
[28:47] it is about the sequencing. Meaning you
[28:50] can't add in, and this is what my paper
[28:52] was about. You have all these issues
[28:53] that are happening. You've only done now
[28:55] through Q1 of this year 12% of that
[28:58] spending. We are barely through it and
[29:00] you have multi-year delays in so many
[29:02] things it's impossible to name at this
[29:04] point. So the risk is that the buildout
[29:06] becomes lumpier, more inflationary, more
[29:09] volatile than the market is modeling.
[29:10] That's what I believe is going to
[29:12] happen. So all I have to do is upload
[29:14] the chart and say what does this say?
[29:16] What is the risk in it? I didn't even
[29:18] ask anything about it. I just said how
[29:19] far into it is and only being this far
[29:21] when we already have bottlenecks. What
[29:23] does it mean? Bottlenecks create cost
[29:25] inflation. Cost inflation creates
[29:26] delays. Delays create revenue
[29:28] recognition risk. We are already seeing
[29:30] all this stuff, guys. And that's the
[29:32] risk is that you're taking Q1, which is
[29:35] only basically
[29:39] two months after people thought it was a
[29:41] bubble and didn't want to be involved in
[29:42] it. The time when I started hearing
[29:44] people accept memory was in October of
[29:46] last year, and that was because the
[29:48] memory prices had gone higher. Before
[29:49] that, everyone was still worried about
[29:51] what was going to happen with tariffs.
[29:53] So, you've had this huge move in things.
[29:55] You've had all this stuff go through for
[29:57] a variety of reasons which I mentioned
[29:58] in the paper. And I just think we're not
[30:00] taking into account what's going on. A
[30:02] must-read interview with a Seaman's
[30:04] employee explaining just how big the
[30:06] demand is for energy equipment right
[30:07] now. And again, 12% of the way into
[30:10] this.
[30:12] Are these decisions speeding up or are
[30:14] they slowing down? Decisions get made
[30:16] very quick. Decisions are quick, but
[30:19] there will be delay in the production
[30:21] and delivery. Is any of this surprising
[30:23] to you? I would say yes. It's not only
[30:25] me. I'm working with some of the people
[30:26] who have been here in the business for
[30:28] the last 40 years and they say also
[30:30] cannot get a proper explanation of how
[30:32] these things are happening. Just imagine
[30:34] we are getting orders double the size of
[30:36] your whole factory can produce in a
[30:37] year.
[30:41] I've shown the Dodge momentum index. Um
[30:45] it's, you know, it slowed down obviously
[30:48] in the winter time, but now it's
[30:49] starting to pick up again. Uh but labor
[30:51] shortages, high material costs, and
[30:53] supply chain disruptions are weighing on
[30:55] owner confidence in the near term.
[30:58] Year-over-year uh DMI, the Dodge
[31:01] Momentum Index, was up 14.1%
[31:04] when compared to April 2025. The
[31:06] commercial segment was 37.2
[31:09] but only 5.8% when data centers are
[31:11] removed. Again, 37 down to six. Data
[31:15] centers are all of it. Um, this is the
[31:19] reason why I think people should now
[31:20] start paying attention to oil. So, Art
[31:23] Burman, uh, I spent a lot of time with
[31:26] Art this week listening to him, but
[31:28] also, you know, I tried not to get into
[31:30] the the the oil side in terms of the
[31:33] details other than just to mention that
[31:34] we still haven't gone anywhere, but
[31:36] sentiment just doesn't care about it.
[31:38] And now we're at the point that people
[31:39] should be caring because we're starting
[31:42] to have draws. The shortage hits US
[31:44] inventories. So, we've been releasing
[31:46] things on the SPR. That's part of the
[31:49] draw. Everything is happening. US total
[31:52] crude inventories fell 17.8 million
[31:54] barrels. That's commercial and SPR
[31:56] stocks combined. On that basis, it's the
[31:57] largest weekly fall since data available
[31:59] starting in 1982.
[32:02] Warren pies last month was the largest
[32:04] cumulative draw of total US petroleum
[32:06] inventories ever. Gasoline inventories
[32:10] are now the lowest since 2014.
[32:12] distillate inventories, the lowest since
[32:14] 2003. Next few weeks is a crucial window
[32:17] for oil fundamentals. Now, you've been
[32:19] hearing this for a while, but again,
[32:21] when you start seeing what is happening,
[32:24] which is the inventories are being drawn
[32:26] down,
[32:27] the inventories are being drawn down,
[32:31] you start running into a situation where
[32:33] this person highlighted that before
[32:35] Ukraine, Russia, we actually had low
[32:38] inventory. So, and that was because of
[32:41] COVID with all the shutdowns. We had a
[32:44] supply galut. So, we had room for this
[32:46] to go on. The question is, is the war
[32:49] actually going to end? We've been in a
[32:50] ceasefire for a while. And yet, you
[32:53] don't have anything changing. Trump went
[32:54] to China. Nothing seems to have changed
[32:57] this week. Uh there seems to be a couple
[32:59] more tankers going through. They seem to
[33:01] be China. So, maybe China's going to get
[33:03] their oil, which would help. But oil's
[33:06] imports, China's oil imports have
[33:08] plunged 20 30%.
[33:11] The global onshore crude oil inventories
[33:14] exchina. So everything exchina, I mean
[33:17] again, you can see the years here. We're
[33:19] really drawing down. So I highly
[33:21] recommend listening to this uh interview
[33:24] with Art Burman. I think he's very, I
[33:27] don't know, honest. He seems to talk
[33:29] about this in a way that is not
[33:31] hyperbolic. He really does go through,
[33:33] like I said, I listened to about four
[33:35] this week, but Art was the one that I
[33:37] think went through why this is so
[33:39] serious and I think he highlights a lot
[33:41] of points. We get caught in it's open or
[33:43] it's not open. The reality is it's
[33:45] probably partially open. The question is
[33:47] it partially open. We don't know the
[33:49] demand destruction. We don't know if
[33:51] it's going to be easy to build up
[33:52] things. So, the reality is it's probably
[33:54] a bigger problem than people realize.
[33:56] And most importantly, people don't seem
[33:57] to care about it anymore. I'm not going
[33:59] to read all of this right now, but you
[34:01] can go through this to um see it on your
[34:03] own. In terms of uh another podcast
[34:06] worth listening to, Luke Goman
[34:09] was on TFTC this week with Marty Bent,
[34:11] who I'll be on with soon. Um and Luke
[34:15] brought up another point that I just
[34:16] think is worth people thinking about
[34:19] again. Um people have forgotten about
[34:21] the sovereign debt problem. Uh he says
[34:24] the Iran conflict has pulled forward the
[34:26] bond market problem.
[34:28] uh I don't want to talk about the AI
[34:30] buildout, but he warns that current
[34:31] valuations enthusiasm assume the the
[34:34] economics will work smoothly even though
[34:36] the whole buildout could be vulnerable
[34:38] to competition, debt financing issues or
[34:40] supply shocks. I I'm the supply shock is
[34:42] the part that I'm I'm going to agree
[34:44] with. I'm not worried about the debt
[34:45] financing uh unless again oil starts to
[34:49] go higher and I think it will impact all
[34:51] of these things. If oil doesn't sit here
[34:53] and we're draining inventories,
[34:56] we should have a spike then. Um, if it's
[34:58] been held in check because we've been
[35:00] drawing or we've been using oil that was
[35:02] in storage, we should be in a situation
[35:05] where if it doesn't get resolved now, it
[35:07] could be an issue. So, think about the
[35:09] riskreward of things when again you're
[35:11] seeing signs right now that it's already
[35:13] impacting supply shocks and we already
[35:15] have supply shocks. This is just an
[35:17] add-on. Um,
[35:20] yeah, I'm not going to go through the
[35:22] rest of this stuff in here,
[35:24] but it was a good interview. Um, CPI,
[35:27] we're now starting to get the Cleveland
[35:29] Fed inflation now casting. Uh, and
[35:32] again, CPI is looking at point4 or 0.5
[35:35] again. PCE is looking at point4 again.
[35:39] That would put both the CPI and the PC
[35:41] above four.
[35:43] And I've talked about the historical
[35:46] returns in the equity market. So, is 4%
[35:49] the line in the sand? I don't know. But
[35:51] rates are moving higher. And it just
[35:53] seems like if we get another surprise in
[35:55] CPI and PPI, and you start seeing that
[35:58] flow through, all of a sudden, you end
[36:01] up in a situation where you're starting
[36:02] to get a little bit more antsy in terms
[36:04] of the Fed. And we'll see how the long
[36:06] rates go. Here's the PPI on things that
[36:08] have nothing to do with oil. This is the
[36:10] input cost. So, again, you've already
[36:12] seen this massive rise. If we don't
[36:15] continue to see the rise, is that mean
[36:17] prices are starting to cause demand
[36:19] destruction? Um, demand destruction does
[36:21] happen at some point, but also
[36:23] bottlenecks cause demand destruction.
[36:25] Truckload. I've talked about this. Um,
[36:27] again, people were fighting this back in
[36:29] here and yet here we go. Uh, I don't
[36:33] know how inflation is going to come down
[36:34] when you've got truckload spot rates
[36:36] heading higher. Welcome Kevin Worsh
[36:40] sworn in today. The Fed will have to
[36:42] raise interest rates in July, says Edard
[36:44] Denny. I don't believe that's going to
[36:46] happen, but if it does happen, again,
[36:47] you've got the uncertainty at a at
[36:50] regardless. Um, Donald Trump couldn't go
[36:54] without speaking on the day that Worsh
[36:56] comes out. We're going to grow our way
[36:57] out of it. We're going to grow the
[36:58] national debt so fast. Economic growth
[37:01] doesn't mean inflation. You don't have
[37:02] to stop the world because you're doing
[37:04] well. So it sounds like we're going to
[37:07] go with the inflation as transitory
[37:08] routine and more importantly because the
[37:10] debt problem is so big, we're going to
[37:12] try to focus on nominal GDP. Well,
[37:15] that's good. The issue is will long-term
[37:17] rates be able to stay in check and
[37:19] that's where the issue is going to come.
[37:20] And that is where I believe the major
[37:22] inflection point on the regime will be.
[37:24] So to get a regime shift, you have to
[37:26] get a momentum unwind because the
[37:27] momentum unwind would probably mean the
[37:29] AI names get hit. If the AI name gets
[37:31] hit, all the markets are going to get
[37:32] hit. Whether it's a 10 percent
[37:34] correction, 20% correction, 30%, 5%, I
[37:37] don't really care. But at some point in
[37:38] there, if rates are going if rates are
[37:40] going higher, oil's going higher, and
[37:42] you have the market going down, I think
[37:44] people are going to be confused on what
[37:46] to do. And right now, everything's
[37:47] quiet, the VIX is low, but I think
[37:50] people should be worried about the
[37:51] whatifs more than they are. Um, I got to
[37:54] give these guys a shout out. I I've
[37:55] started I've, you know, I've I've given
[37:57] up a lot listening to most macro um
[38:00] economic podcast. I've talked about it.
[38:03] I just find it to be way too doom and
[38:05] gloomish. Uh not focused on AI, still
[38:09] calling AI a bubble, saying the economy
[38:11] cannot go with AI doing this, focused on
[38:14] the delinquencies and all stuff like
[38:15] that. Compound and friends does a really
[38:17] good job of just talking about what's
[38:19] going on in the market for the day or
[38:20] for the week. I think they have a good
[38:22] grasp on the general um things happening
[38:25] and what's important. And so by
[38:26] listening to them, they also have good
[38:28] guests. And this week, I thought they
[38:30] did they had two good podcasts. So, I'm
[38:32] gonna show you know both of them in
[38:34] terms of this. They talked about Nvidia.
[38:36] They talked about Google IO. They talked
[38:38] about Gavin Baker. Um, Gavin Baker gave
[38:41] an interview. It was a great interview.
[38:42] It was on another one on Patrick
[38:43] Oonessy, which I'll get into in a little
[38:46] bit. But the point that they made, which
[38:48] I thought was interesting, it's not a
[38:49] bubble, it's a wave. Bubbles imply
[38:51] something that suddenly disappears. see
[38:54] this type of logic to me. Um, and I give
[38:56] them credit. Both Michael and Josh do a
[38:57] good job of being rational. Uh, and the
[39:00] fact that Michael is a big Knicks fan
[39:02] and wears Nick's paraphernalia all the
[39:04] time, uh, means a lot to me as well. But
[39:07] they've done a good job of just going
[39:09] through it. So, I highly recommend if
[39:10] you're looking for people to listen to
[39:13] regularly, uh, to find kind of under the
[39:16] hood things going on, I would give these
[39:18] guys a shout. I like what they do. They
[39:20] also have my boy Adam Parker on there
[39:22] all the time. and he seems to be
[39:24] spending a lot more time. This is
[39:25] something I wrote about in a paper
[39:27] called the art of unlearning, but they
[39:29] talked about Adam Park's idea that
[39:30] investors may need to unlearn the past.
[39:33] Uh I wrote a paper on that. I was going
[39:34] to show you guys some of the details. If
[39:36] you guys want to see it, just let me
[39:37] know. I wrote it in 2016. I was
[39:40] rereading it this weekend. I will
[39:41] highlight that in 2016. Um Dr. Michael
[39:45] Bur, uh I used him in there to say he
[39:48] has to unlearn. It's amazing how many
[39:50] times a guy can call for a bubble. He
[39:52] called for the collapse of markets. It
[39:55] wasn't even just that. That was in 2016.
[39:58] So again, if you're paying for his
[39:59] Substack and you're listening to stuff,
[40:00] just remember he's only interested in
[40:02] one thing, which is the markets don't
[40:04] make sense mathematically. And that was
[40:06] the point of the art of unlearning is
[40:08] that you're going to have to start
[40:09] putting math onto everything when you're
[40:12] dealing with AI and when you're dealing
[40:13] with the debt, the size of the country,
[40:15] and the printing press. Exponential
[40:17] innovation. The printing press has
[40:19] created a different thing. Um, in terms
[40:21] of Adam Parker software versus semi-
[40:23] divergence, that's where that chart was
[40:24] from that I highlighted. Um, I just
[40:27] wanted to show this on the bottleneck
[40:29] side just because, um, I I I I really do
[40:32] think it's important to kind of pay
[40:33] attention and for you subscribers, I
[40:36] have it in there. Um, the grid problem
[40:39] is an issue. So, again, I've shown this
[40:42] before, but if you haven't seen it, it
[40:44] is up on on the website. The data
[40:46] centers and getting them done. This just
[40:48] shows all the different places of of the
[40:50] weakness, but I just want to show you
[40:52] the numbers. The probability of being
[40:53] realized in terms of getting built by
[40:55] 2030. Look at these numbers here. And
[40:58] this is just half of them across the
[41:00] country. And this goes through all of
[41:01] the issues. I weight them and use all of
[41:04] these. We have so many acute issues,
[41:07] high issues, um, to get the buildout
[41:09] out. It's just a lot, guys. And so to
[41:12] think that this is going to happen like
[41:14] without a problem is not there. Um Adam
[41:16] also kind of went through
[41:19] um basically saying that investors are
[41:21] using mental models that were built for
[41:23] the last regime. I couldn't agree more.
[41:26] Um and he says capex heavy businesses
[41:28] may deserve a rerating. That's already
[41:30] happened for some of the industrials. I
[41:32] think that's happening for the semis and
[41:34] I think that's part of the issue
[41:35] especially for Micron. Um capex is not
[41:38] wasteful overhead. It is the toll to
[41:40] participate in the new growth cycle.
[41:43] Again I couldn't say it any better. this
[41:45] is why your capex is my opportunity is a
[41:47] big focus on it. So again, big shout out
[41:50] to those guys. And here was number two.
[41:53] Uh Yan Vanek was on. If you guys have
[41:55] never heard Yan Vanek from Vanek, uh
[41:59] I've now I was at an Anthony Pompiano
[42:02] event where he spoke after I did and I
[42:05] found him to be have great macro
[42:07] comments and he was on the show today or
[42:10] yesterday compound and friends. their
[42:12] second episode and I thought he did a
[42:14] great job and Van has the SMH. Uh they
[42:18] acquired it uh I think after the great
[42:20] financial crisis when Meil Lynch was
[42:23] closing things down or Bank of America I
[42:25] guess it was Meil Lynch. Um he talks
[42:28] about
[42:30] the issue with memory stocks and he
[42:32] basically brought up on the question of
[42:34] whether memory stocks is a bubble
[42:36] especially compared to Nvidia. Um,
[42:38] memory looks more bubble-like than
[42:40] Nvidia because memory does not have the
[42:42] same moat.
[42:44] Memory, you need my product. I can so
[42:46] I'm I can raise price. Nvidia, I can
[42:48] give you more intelligence. Blah blah
[42:49] blah. Then he lists, and this is the
[42:51] part I was most interested in, the
[42:54] possibility of things that could be
[42:56] questions over the course of the next
[42:57] year and break the memory trade. Chinese
[43:00] memory supply could eventually become a
[43:01] competitive threat. Model efficiency
[43:03] could reduce memory intensity. We've
[43:05] already seen that with TurboQuant. the
[43:07] market ignored it so far. Third,
[43:09] customers will eventually respond to
[43:10] higher costs. This gets back into your
[43:13] costs and he gives the specific example
[43:15] which I think is important telling his
[43:18] CTO to stop spending so much on compute.
[43:21] I think the spending on compute is
[43:23] becoming more of a story. The AI stack
[43:26] itself could change and that's what
[43:28] brought memory to what it is. Once we
[43:30] got to inference and we started getting
[43:31] into Vera Rubin, which was a new thing
[43:33] at the beginning of this year, which
[43:34] hasn't happened yet, we started buying
[43:37] memory because it meant the long-term
[43:39] makes it more likely that it's going to
[43:40] be needed. Well, maybe some other change
[43:42] will happen. Memory is the bottleneck
[43:45] trade. Nvidia is the platform trade and
[43:46] I completely agree. And as someone who's
[43:48] been in the trade, I just think the
[43:50] riskreward has changed and maybe it does
[43:51] double from here. I like Marll to do
[43:53] more than what I think memory will do. I
[43:55] think Marll is more important to the
[43:57] future and more of a moat. I could be
[44:00] wrong on it. Uh but I leave that up to
[44:02] you guys. This is the cost side to AI
[44:04] and how much it is starting to go
[44:06] higher. Again, you can see chat GPT5 uh
[44:10] up here. Um again, I'm going to say that
[44:12] from my perspective in terms of the best
[44:15] model, I have not used Claude more than
[44:17] a couple times this week as I end the
[44:19] week. I have used ChateBT and Codeex the
[44:22] entire week. I have used Gemini
[44:24] occasionally. I have used perplexity
[44:26] occasionally. I have uh become very very
[44:29] uh sick of of the compute situation in
[44:32] claude and have spent most of my time on
[44:34] on chat GBT this week. We'll see if I go
[44:37] back to it next week. Here is the token
[44:39] cost. So Bloomberg has this index. I
[44:42] think most importantly you see that this
[44:43] started to raise in February which means
[44:46] again the compute needs the shortages
[44:47] are there the costs are going higher.
[44:50] Gavin Baker again whenever he speaks on
[44:52] any of these things I think he's worth
[44:54] listening to. He said it again which he
[44:56] said multiple times and I'll just say
[44:59] this one of the things he said is it is
[45:01] amazing that you get to listen to let's
[45:04] just take Jensen Yuang but all of the
[45:05] leaders in AI speaking multiple times a
[45:09] week. I completely agree. I've created
[45:11] notebook LMS of Jensen Yuang so many
[45:13] different people to have knowledge
[45:15] brains of this. They're handing out free
[45:17] alpha. They are telling you how things
[45:19] are going on these podcasts, what they
[45:21] need for the future. He talks about that
[45:24] and then he basically goes through the
[45:25] investor questions. If AI can materially
[45:27] improve efficiency faster than demand,
[45:29] then the market may be overestimating
[45:30] long-term demand for memory, GPUs, and
[45:32] power. So again, the efficiency side is
[45:35] going to happen. That I'm sure of. Um,
[45:37] and the reason I say I'm sure of it is
[45:39] because really smart people that have
[45:41] been on interviews, particularly Sergey
[45:42] Brin, have effectively guaranteed it.
[45:44] And I think Google, which had the
[45:46] Transformer paper, probably knows more
[45:48] than anyone that they're they're getting
[45:50] into this. But also, I think you're
[45:52] hearing more and more about if there's
[45:54] continual learning, does continual
[45:57] learning arrive soon? He treats this as
[45:59] potentially explosive.
[46:01] I think we've learned our lesson with
[46:03] how Opus 4.5 came out of nowhere and
[46:05] then the agentic world flew in.
[46:07] Eventually, we're going to have a change
[46:11] in one year is a long time in AI. to bet
[46:14] on next year's earnings for memory.
[46:15] Guys,
[46:17] sorry, you're living in a linear world
[46:19] where continually and uh recursive
[46:22] self-improvement is going to change
[46:23] things. And even if it doesn't change
[46:25] things into next year, if all of a
[46:27] sudden there's a cliff because they
[46:28] figured out some way to use onetenth the
[46:32] memory, it's going to be an issue. So
[46:35] again, these are all would a could a
[46:36] shouldas, but I've been living in the AI
[46:38] world, it seems, a lot longer than most
[46:39] people. These models have moved faster.
[46:42] They're changing
[46:45] unbelievably. I've gone from not using
[46:48] Claude to using it all the time and now
[46:49] I'm back to chat GPT and I love 5.5. By
[46:53] far my favorite experience in terms of
[46:55] AI and I really view Claude as being
[46:58] something a little bit older at this
[47:00] point because it's slower and I haven't
[47:01] seen the improvements and the model is
[47:03] not as smart as 5.5. It's just not. Um
[47:06] the profits are still going to energy
[47:09] data centers chips and models. The app
[47:11] layer is not dead, but it's much more
[47:14] selective than many hoped. And that gets
[47:16] into the software side where the value
[47:19] is acrewing. This is not, you know,
[47:21] surprising, but this is basically the
[47:23] five layer cake.
[47:27] Google IO, the keynote speech. Uh, by
[47:29] the way, all these podcasts, you know,
[47:31] I'll have a summary up on the subscriber
[47:33] website with the link, but if you just
[47:36] want to go through and read the summary
[47:38] of each one of them, and that way you
[47:39] can upload them into, uh, your LLM of
[47:43] choice, uh, just so you guys hear this,
[47:45] I've done a lot of things that I've gone
[47:48] through on the video side. I will
[47:49] release them on the payw wall soon. So,
[47:51] the videos that I'm doing to show you
[47:53] how to use the the tools that I've put
[47:55] up there, the model portfolio to use the
[47:58] uh actual um uh transcripts that I'm
[48:03] putting up there uh or at least the
[48:05] links and showing you how to do the
[48:07] transcripts. All that stuff will start
[48:08] going up next week and the week after so
[48:10] you guys can start using it. You have to
[48:12] be using it regularly. Don't just watch
[48:14] this video and take what comes out of
[48:17] it. upload it to notebook LM first of
[48:20] all if you guys haven't done that and I
[48:21] will show you how to do that as well. Uh
[48:23] but I would absolutely highly recommend
[48:26] just taking some of the stuff and even
[48:27] if you take a screenshot of this and
[48:28] upload it into chatgpt 5.5 you'll have a
[48:32] uh a good experience. Uh I just wanted
[48:34] to bring up in the Google IO that if you
[48:37] followed what they said last year about
[48:38] the token side which was really the
[48:40] focal point. the models were just
[48:42] getting smarter but the focal f focal
[48:44] point was token usage was going up which
[48:47] was again big for their cloud business
[48:49] everything now is about agents and as
[48:51] someone who has his own business his own
[48:53] LLC
[48:54] um and it's on Google work workspace I I
[48:58] mean I am happy as can be that I'm going
[49:00] to be able to use Google in the way that
[49:02] they're saying basically as long as your
[49:05] work lives in in Gmail drive docs sheets
[49:07] calendar and chat this is going to be
[49:09] different than what a normal chat
[49:11] chatbot can do. So, I was listening to
[49:12] it because this is all about agent. It's
[49:15] going to be a better experience. They
[49:17] also talked about the consumer agent
[49:19] part. This is the part that I'm telling
[49:21] you in the same way that Inference last
[49:24] year, if you would have followed it, a
[49:26] year from now, you're going to look back
[49:28] and go, "Oh my gosh, between the Stripe
[49:31] annual newsletter, the Coinbase earnings
[49:34] report and Google AI agent or IO agent
[49:38] for both the consumer and the business.
[49:40] How did we not buy Circle? How did we
[49:42] not buy Ethereum? How did we not buy
[49:44] Bitcoin?" That will be in May of next
[49:46] year, guys. And whether it starts in a
[49:48] week, a month, three months, or six
[49:51] months. Remember the memory trade took a
[49:53] long time. And I saw Micron bang up
[49:56] against 110 for weeks. Back down to 60,
[50:00] back up. I think cryptos in the same
[50:02] place. I don't think people have
[50:04] connected yet that the AI infrastructure
[50:06] is to trade right now, but starting at
[50:08] some point, the handoff is going to the
[50:11] redesigning of the global finan
[50:14] financial system. For those of you who
[50:16] got to watch the webinar this week that
[50:17] I did with uh the Kraken uh team and uh
[50:23] and uh the other groups that are
[50:25] involved with the spa. Uh again, I think
[50:29] this is part of what's coming very very
[50:31] soon and I think you guys should be
[50:33] doing your homework on the space. So the
[50:35] Google AI IO side for the agentic side
[50:38] was very important. And again, I'm not
[50:40] going to read all of it, but you just
[50:41] have to go through and just see what
[50:43] we're talking about in terms of doing
[50:45] things across the ecosystem and what the
[50:47] agentic world is going to do, especially
[50:49] when combined with search. For those of
[50:51] you Bloomberg users, I highly recommend
[50:54] going through and watching um Sundar
[50:56] Pachai and at least learning about
[50:58] Google Finance and what you're going to
[50:59] be able to do because we're getting into
[51:02] the point where with agents you're going
[51:04] to be able to have 20 30 40 50 60 70 80
[51:08] 90 agents working on a problem for you
[51:11] all with different specialties doing
[51:12] different things. And that's what leads
[51:14] to the tokens going exponential.
[51:18] And here is that page from the Coinbase
[51:21] side. And whether it's tokenization,
[51:24] whether it's stable coins, whether it's
[51:27] just the onchain economy. Again, guys, I
[51:30] think you're missing the parabol the
[51:32] parabola that's happening. Uh, one group
[51:34] I've had a lot of people ask about um
[51:36] why it's not participating.
[51:38] I want to just highlight. So, this is
[51:40] Vistra, which is basically unchanged
[51:42] since the end of 2024. Uh, its multiples
[51:45] were too high back then. there was less
[51:47] names to invest in on the AI trade. When
[51:50] we hit October
[51:52] right here, um we had a lot of names and
[51:54] this is when the broadening out
[51:55] happened. This is it relative to Nvidia.
[51:59] These two names to me are think of them
[52:02] as the defensive side of AI since I
[52:05] think we're very close to some kind of a
[52:07] momentum unwind where the names are
[52:09] running out of steam. This is kind of
[52:12] like the inverse of Micron. Now it
[52:14] doesn't have the parabolic downside but
[52:17] I think people were looking for beta. So
[52:19] with inside AI these are the defensive
[52:21] areas. So if you want to play something
[52:24] for the next whatever months I would be
[52:26] increasing size into some of those names
[52:28] like Vistra here is a chart with an IP.
[52:31] I created uh an index of 10 names that
[52:35] are IPs. So the power independent power
[52:38] producers are at least close enough to
[52:40] the IPs including Vistra nine others and
[52:43] you get this thing here which is
[52:44] consolidated. I think uh
[52:48] for all you cyclical defense uh watchers
[52:52] I think if defense starts working in AI
[52:55] you're going to see Vistra in some of
[52:56] these IPs. This is where I'd be moving
[52:58] some of my money in to reduce the beta.
[53:00] Um, for those of you who keep asking, if
[53:02] you go to my video and you go down here
[53:05] and you look, it will tell you where you
[53:07] can go. Subscribing to my Substack for
[53:10] free is not getting you the subscriber
[53:12] payw wall. So, I'm only bringing this up
[53:14] as I start to go through the
[53:15] conversations for the RAAS, for the FAS.
[53:18] I've been getting so much progress.
[53:20] People are asking for things. I'm
[53:22] spending a lot more time. I'll be down
[53:23] in the New York Stock Exchange in a
[53:24] couple weeks on tokenization and also
[53:26] talking to them about doing some stuff
[53:28] with them. But I'm really trying to help
[53:30] people with the benchmark arbitrage. So
[53:32] for any of the FAS and RAAS that want to
[53:34] have a conversation, that want to see if
[53:36] there's ways we can work together. Like
[53:38] I said, I've got Morgan Stanley creating
[53:41] uh the ability of of doing this from an
[53:43] ISTA basis. Uh I've had ETF uh
[53:46] reachouts, but this stuff is moving so
[53:48] fast that I don't want to wait six
[53:50] months on something. Uh here's this
[53:52] thematic portfolio made new all-time
[53:54] highs in this again. So you got the
[53:57] momentum trade. The good thing is not
[53:59] all of the thematic portions, the themes
[54:02] made new highs. But again, here is the
[54:07] alpha that's been created since the
[54:08] launch of Opus 4.5, which happened in
[54:11] here. So again, you've had basically
[54:14] five down weeks. Only one has been
[54:16] greater than 2% and you can see how many
[54:18] have been above. Again, it's a hundred
[54:20] names.
[54:22] It's across multiple sectors. This is
[54:24] not just semiconductors. You've got
[54:26] power, you've got the whole rack, you've
[54:28] got chemicals, you've got optical, and
[54:29] you've got package, plus five macro
[54:31] satellite names, which just means five
[54:33] other names that I've written about. So,
[54:36] not all of them made new highs. Most of
[54:38] the ones that went up to the highs were
[54:40] related to the non-memory semiconductor
[54:42] names. So, Marll made new highs. Um,
[54:44] Intel had fallen. Uh, I mentioned on the
[54:47] the subscriber webinar that when Intel
[54:49] went under 110 this week, I started
[54:52] buying some because I believe CPUs and
[54:54] optical are still good trades over
[54:56] memory. That makes people crazy because
[54:58] Intel's up five times, but I'm doing
[55:00] this based on the demand surprise versus
[55:02] the supply. Um, this is where we get
[55:05] into the concentrated. So, I've had a
[55:07] lot of people saying, "Hey, I can't
[55:08] trade 100 names. I'm a subscriber. How
[55:11] can I do something better?" So, I
[55:13] created a 25 name ticker. This is
[55:15] basically out of the five themes uh plus
[55:17] one name from the uh macro satellites of
[55:20] the five took 25 names and this is the
[55:23] overlay between the thematic portfolio
[55:26] and that 25 name. So you can see it's a
[55:30] direct overlay uh in terms of
[55:31] correlation. It has more beta so it's
[55:34] outperformed uh by a decent amount and
[55:36] that's because it has 25 names as
[55:38] opposed to 100 names. Uh but it's got
[55:42] stuff from every single grouping. Uh
[55:44] this is the performance on it in terms
[55:46] of the spread to the S&P. The reason I
[55:48] wanted to show this is
[55:51] this basket is up 68% year to date. The
[55:55] S&P is up 9% year-to date. If you had as
[55:59] an FA or sorry, as as someone who has
[56:02] discretion or someone in your portfolio,
[56:04] whatever you have in the S&P 500 in
[56:07] spies, if you had 10% 90% in this, uh,
[56:11] you got 8% and then if you have 10% in
[56:14] this, you have another close to seven.
[56:17] That's the way that I think about these
[56:18] things is I think you need to create the
[56:20] alpha in your portfolio by overweighting
[56:22] some percentage of this. That is the way
[56:25] also with Bitcoin and things like that.
[56:27] In this case though, I think that's
[56:29] where the conversations I've been having
[56:31] is the benchmarks are wrong. I don't
[56:34] think the hyperscaler spenders which are
[56:36] huge companies or the software companies
[56:39] which are huge companies deserve to have
[56:41] the same waiting that they had in the
[56:43] past because they are being disrupted by
[56:45] AI or they are the spending side. You
[56:47] want to be long the receivers not the
[56:48] spenders. Because 25 names I knew would
[56:52] be too much. I created a 10 name
[56:54] portfolio as well. Again, taking them
[56:57] across each of the themes basically to
[57:00] try and get to the point where you have
[57:02] at least one name in every theme to get
[57:04] it correlated. And this is what I came
[57:06] up with. So, this is the three of them
[57:07] overlaid. Now, uh again, the green one,
[57:10] which is the one with the least names,
[57:12] is going to have the highest beta. Uh I
[57:16] just want to remind you guys this goes
[57:18] up every week. the exhaustion side, the
[57:20] technical side, and the fundamental
[57:22] sheet. It helped out dramatically over
[57:24] the course of the last week from a
[57:25] trading perspective. Remember, we had a
[57:27] significant momentum unwind. This is
[57:30] where we stand on the exhaustion side
[57:31] now. We've reset most of these. So, if
[57:33] you were trading, you got in at good
[57:35] entry points because we had a huge
[57:38] amount of exhaustion names and now
[57:41] they've all run off. This is how you can
[57:42] use these things to at least have a
[57:44] sense. The technical pattern on all of
[57:46] these are are are still strong and I
[57:48] think it's going to remain that way. The
[57:50] only thing that's weak is the macro
[57:51] satellites and that's because Bitcoin's
[57:53] below the 200 day moving average. Silver
[57:56] not trading well. Um you have Eli Liy in
[57:58] there. You have EWZ in there. I forget
[58:01] the fifth one. Oh, Palanteer, which is
[58:03] not trading well. So those five names,
[58:05] you know, they're mixed. The rest of
[58:07] them, they're well above the 200 day
[58:09] moving average. This is the percentage
[58:10] of names in the thing that are above the
[58:12] 200 day, above the 50-day. The average
[58:15] RSI is now at least reasonable. The
[58:18] relative strength is fine. Again, I
[58:20] think there's a risk for all of these
[58:21] and a correction. But from a trading
[58:23] perspective where you guys don't care
[58:25] about the next three months, you care
[58:26] about the next day. At this point, the
[58:29] trend is still very, very strong and you
[58:31] should be trading exhaustion to get out
[58:34] or at least reduce and rotate and then
[58:36] when they reset, be in a position to buy
[58:38] things. Believe it or not, Micron went
[58:40] from above 800 down to about 650 and
[58:43] then closed the week somewhere around
[58:45] above 700. You can make money trading
[58:47] stuff that that wide. So, that's the way
[58:49] I'd use it. I'm going to show you guys
[58:51] how to do this, but I showed on the
[58:53] webinar webinar side. If you upload all
[58:56] of those Excel sheets into one LLM and
[59:01] then you go go through the three files
[59:03] and find me the names that meet this
[59:04] criteria and you pick the criteria, I
[59:07] gave an example. It must meet all three.
[59:09] I want only ones with PEG ratios below
[59:11] one. So they have some kind of a
[59:13] valuation, fundamental valuation put in
[59:16] there. Technical names with a score of
[59:17] 85. The charts are strong. I only want
[59:20] good technical scores, but I want cheap
[59:22] names that are in there and I want names
[59:24] that have low exhaustion. If you would
[59:28] have done that this week, you would have
[59:30] done well in terms of names that have
[59:32] gone. In terms of at that time, there
[59:35] were five names that met the criterion.
[59:37] Optical and interconnects,
[59:39] four in semiconductors, four in the
[59:41] whole rack, one in chemicals in this.
[59:43] There were 15 overall. That's the way
[59:44] I'd be using this stuff if I were you
[59:46] guys. Um, that's it for this week. I'm
[59:49] going to leave Chile Maine and head back
[59:52] uh on Monday. I'll see you guys next
[59:55] week from uh