Transcription
One in the summertime, a relatively calm week. Earnings continue to come through, and that will be a theme for the rest of the year, regardless of what you're reading and hearing. The earnings are going to be the major story. I'll get into some specific things on AI and some things that I've talked about in the past regarding the earnings and the profit margins.
China came to the market today after I had heard people talk about how China wasn't in an AI race with the US. For those of you who are listening to people say that, that is not true. And yes, I've read the Deep Seek interview. I've read the Xi Jinping interview. They're in a race for a lot of good reasons, too. If you believe that the dollar reserve status has been a benefit, owning AI and crypto becomes kind of the most important thing in the world going forward. So, everyone from a national basis that can compete would compete, and that really makes it a two-horse race between China and the US with two different strategic approaches. I'll go through some of that.
Google earnings, I'll get into the nitty-gritty of it and just go through the facts in terms of not just the negative free cash flow, which makes its way around Twitter for the quarter, but also the incredibly rising backlog in the cloud growth. Short-term noise versus long-term flow. I'm going to try to give you guys a perspective on what you should be thinking about in AI, regardless of whether this slowdown that we've been in, in terms of the post-parabolic move consolidation, whether it lasts another six weeks or six months, doesn't really matter. It doesn't change the long-term flow of AI.
A new prompt on the Mosaic side that I spent some time working on this week. Kimik 3, I go into more details. I released a new paper on Vera Rubin, which is now coming out. You've got the rise in oil prices again. So, we get to talk about that and the Clarity Act. But, let's get started.
So, for the week, after all the stuff, whether there was a big up move, big down move, the S&P was down 61 basis points. So, again, what I like to look at is a massive move up. And we've now had basically two up weeks, then two down weeks. Two up. We're consolidating a big move higher at this point. That's the way the weekly moves are going. That may change, but for right now, I'd go with that. Qs down 1 and a half percent. IWM, the fourth week in a row with a sub 1% absolute return move. So, it's been down four weeks in a row, less than 1% each move. You don't see that kind of movement anywhere. So, vol's been coming down, and we've seen small caps and equal weight do much better than the broader market at this point. Which to me is a symbol of earning strength, and breadth, and broadening out.
Momentum, tech momentum, after just a gruesome four weeks, we got a bounce, 8% for the week. Vol for the week was less. Here's the move. We went up, then we came back down. This is the tech mo chart. Industrial mo went right back down towards the lows. The white line here is mo in general. After you have a massive move of this size, with the vol the way I've shown it, I find very little chance that we go straight back up. If we did go back up, I think it's going to take some time. I think it'll take the whole second half of the year. I believe we'll be going through a period of chop where the companies that beat, like the Intels, the Microns, the stuff like that, they'll find a bottom and they'll be able to build something. And then you'll have some other companies that don't meet the bar.
On the flip side, what helped mo this week on one side, tech mo, was the fact that Google sold off. That was really the biggest part of the jump here. The Google falling, because again, the hyperscale has been on the negative side. As you look at the rest of mo, again, very quiet, not much going on. This is really an AI side. This is the equal weight ETF, which as I mentioned, you can't even tell there's anything going on there volatility-wise. We're just kind of bouncing around. Thematic portfolio was up 1 and a half percent. Again, a lot of chop, a lot of names up, a lot of names down. I think you're going to be seeing that. Relative to the hyperscalers, the thematic portfolio, relative to hyperscalers, actually had its biggest week since late March. That was bigger than all of these in here. So, it just highlights to you that this was a bad week for the hyperscalers.
Thematic portfolio, I showed this last week. We kind of went under the 60. Basically, we're right around the 38% retracement. Could it go lower? Sure. I don't think it will at this point. I think we're going to spend time chopping around here, and if we do, it'll be minor.
In terms of the breadth from a technical basis, again, this was a warning sign starting to see the change up here. Well, we did get to the highest level since July 10th, and basically, except for that one print on that one day on July 10th, we're starting to see the breadth improve. This is something to watch, especially if we come back next week and we get higher on the 20-day. That is going to be the first sign. Again, the upward sloping 200-day is still fantastic. Above the 200-day, still great. We are still in a bull market in this. And last week I showed you the history of me and head and shoulders and structural bull markets, which is some of the most powerful signals come from breaking the neckline and then turning back up. Now, we've done that so far. Give it some more time and see what happens.
There's a lot of leverage still in the system, so it is highly possible, even if it's for a day print, that we break down further towards the 200-day. As I mentioned, I started buying some last week. I didn't buy any this week, but I did buy last week. And I will continue to look to buy as we consolidate around here because I believe that there are the bearishness in AI is growing again. And we go through these periods, and when people get bearish on it again, and I listen to the arguments and I hear the stuff people are saying, and when I hear things like Anthropic is done, just because their growth rate is going to slow, that's when I start getting more interested in things.
In terms of the earnings, this is going to be the major story. You're not going to hear me talk much, and I'll get into the earnings growth, but 86% of, we're almost a third of the way through, have reported and been above. They've beaten estimates. 3% have been the same. 11% have gone. So, we're on an 8:1 ratio of beats to misses. The 1-year average is 80%. The 5-year average is 78, and the 10-year is 76. Again, the earnings are just coming out across the board. You're going to see this repeatedly from me.
The profit margins. Now, again, this includes Google, which had a one-time off. Well, not one-time, because it also impacted these numbers here. Without the Google number, we're at 14.4%. If you take the Google number out, we're down here closer to closer to 14%. So, we are still progressing higher when it comes to S&P margins. And that's the major story.
During the week, I did do an update video for the subscribers, and it was on China. Last night, China sent an important signal. Capital markets have become critical to achieving its AI ambitions. It's one of the reasons why I want to make sure that you guys recognize this is a race. This is important to understand. Even though there's two different goals, and I do agree that China's more focused on open-source and this concept of bringing abundance to everyone, where the US is still focused on capitalism, which I think it's going to run into problems in terms of making money off AI the way that people think. So, I am not a bear on Anthropic, but I do believe their growth rate will never get to the levels that people had talked about just a month ago. And I think that's an important shift.
So, China unleashes broad state support to halt the tech sell-off. This is important for you guys to think about because when it comes to the bottom of AI, China was heavily involved in supporting it because they've got their own issues if their stock market goes down. China's in the same position as the US, where the stock market is becoming more critical, not because it's a huge part of the investment side there, but because of the capital needs, and they need a stable market, and they don't have a housing market, because house prices continue to fall in China every single month, and that leads to instability if they don't have some asset class going higher, and stocks are a little bit easier right now.
China state-owned funds buy nearly 9 billion shares to support the market. They're changing listing standards, just like the US did. So, on the IPOs and getting listed, they are changing to again get people to get more involved in these stocks very quickly. It's much easier to support them if they're able to get into indices earlier. Chinese stocks extend rises, state-backed insurers tout long-term investments. The latest sign of government support, five state-backed insurers say they will continue to boost their equity holdings.
Now, a lot of this comes from a very important AI company, the most important AI company in China, Deep Seek. Founders Fund slumped 16%. So, remember, it's a hedge fund. So, it was down 16% last week. Deep Seek in China's AI boom are increasingly powered by state money. This was in May. China's Deep Seek closed over $7 billion funding on June 16th with an unusual deal structure. And the reason it was unusual is because China's state-backed. Basically, the government is the only outside party given voting rights. This is why you can't separate the race. China's heavily involved, and so if the most important AI company inside China falls and China isn't supporting them and they need to raise more capital, which I'll get into, and not just them, everyone.
China's AI companies rush to raise funds and close gap with the US. Chinese AI developers are racing to raise money through share offerings or loans, believing they need a bigger war chest to keep up with US competitors. Capital is critical. China's Deep Seek to raise fresh capital. They just did a raise, the government was involved, $74 billion, and this is a month later. Moonshot of fame Kymeta K3.0 seeks $50 billion valuation in pre-IPO round. So, they're going to raise money. And again, I said this last week. Anthropic is a two, three trillion-dollar valued company now. This is a $50 billion valuation. No comparison between being able to support AI from a big basis. And I'll get into more of Kymeta K later.
All of this capital needs, well, as I showed last week and we'll continue, you've got the hyperscaler bonds are declining. That's got everyone freaked out. It's hitting global bond markets. That's got everyone freaked out. I'm not worried about that. SpaceX sets unlock $116 billion. This is all just to show that again, when I went through the reasons for the mid-cycle slowdown, one of the things that was on the headwind list was the massive amount of capital raises that were going to come, not just from what's happened so far. Google raised $80 billion. You've got all the Chinese companies raising money. You've still got OpenAI. You've got Anthropic. SpaceX was done. It should not be surprising to anyone that the market's having trouble as all of this liquidity is being soaked up by issuance. It takes time to digest it, and it comes during the summertime when there's not much volume or liquidity as it is, and most people are not around.
Eric Schmidt says cash, not energy, is the real limit on AI growth. I've seen this chart a bunch. Phil Rosen, my buddy, again, I think this is going to persist for a long time. I don't think this is anything new. But it fits into this. So, this is from Sundar after the results, and I just want to make sure. There's you can go through and talk about the free cash flow. You can talk about the fact they're not going to make it. This is a seven slides of just honesty on what is happening. The first part is you cannot ignore that their revenue, not their earnings, the revenue growth, 24% at the aggregate level. 17% still in search. 82% in cloud. And 13% YouTube ads. Now, the reason I bring that up as you go through this is because they are absolutely growing as they're spending money. Their issue is not their business growing. It is the backlog because they don't have capacity. We continue to be supply constrained. Evidence of rapid adoption. Demand still outpaces that investment, even after large capacity. The company has already spent heavily, and it is still capacity short. They're expanding third-party capacity to bridge. Again, you think about why Meta is going through. Meta can't use it. They don't have the cloud yet, so they're willing to sell it out to people. Because other people can't fill it. This is exactly what Meta should be doing. If they can't use the compute for the next six months, why wouldn't you use it? You're getting great margins on it. They're willing to rent more expensive external compute. Short-term, expensive cap capacity deal can still be very attractive if they do it outside. So, they raised their CapEx, so good again for all the memory and the chip names. They're going to continue to invest as long as they see an attractive return.
On the direct ROI C question, Google's become more bullish on the opportunities over the year ahead, and that's for good reason. First of all, the cloud is growing at 82%. The backlog was up another $50 billion. They expect to recognize just over half of that, or let's say $260 billion over the next 12 months. So, if they're going to spend $200 billion and they're expecting over the next two years to get $200 billion, $260 billion, you guys can start to do the overall math over, okay, they have a lot of revenue sitting outside, and very few people talk about the backlog like it's not going to happen. I get into the risks later, so let's just go through this.
So, the free cash flow was negative. Operating cash flow was still $39 billion, but CapEx was $45 billion. This was the first quarter where it was negative. They still have a huge cash flow. But as they continue to spend money, they're depending on the revenues coming in to keep this fairly close to nothing, so they don't have to go into their cash reserves. Regardless, this is not some crazy story. They have the ability to raise money. They're not trying to manage towards near-term positive free cash flow.
So, here are the revenues. And what I want you to look is you see the parabolic move. This is not a short move. That's why it's 82% year-over-year growth. So, Bob Elliot, you have to think about what they paid to get it. Since Q1 '24, Google has generated $30 billion excess earnings in their cloud unit, but has paid $140 billion in CapEx to get it. Okay. So, I take that from go through and grok, and basically ask it for the Bob Elliot argument for apples-to-apples, since CapEx is an investment for the future, shouldn't you include the backlog growth as well, as this is what the CapEx is for? And basically it goes through, and you can't ignore it. The exploding backlog supports the thesis that returns will materialize over the 5-to-7-year depreciation window if conversion rates hold. So, again, this is just a case of investors having trouble. This is a backlog that grows from $106 billion to $514 billion in 1 year. I mean, that's unbelievable. To go from $106 billion to $514 billion in 1 year is unprecedented, makes it hard to believe. Elliot-style concerns reflect difficulty pricing unprecedented AI-driven acceleration versus traditional ROI models. Yet, the backlog in this, basically, investors often lag exponential curves, and sustained execution removes doubt.
Now, to give everyone the honest benefit of the doubt, let's just go to Fable 5 and let the smartest LLM on the street go through and basically tell me what's going on. And what I wanted it to do was basically go through the details and then give me a probability that this is going to work out well. First of all, it says the backlog is the whole ballgame. I didn't ask it to say that, it just did it. They are not spending on faith. The incremental spending is chasing demand they can already see. And the unit economics are improving as they scale, not deteriorating. They're literally renting Nvidia chips from Space Act for roughly $920 million a month. This is why if you don't go in and just type in, "Give me the details on," and all I asked it was, "Earnings Google report. Let's go through the details. Let's go through the backlog. Let's go through the cash flow. How should we interpret it?" And then at the end, give me a probability that this makes it.
One honesty note on the corner. The reported $9.11 EPS included about $99 billion of gains on equity stakes. The operational EPS was closer to $2.85, slightly below consensus. The great thing about this, it's just giving you the honest thing. And here's what it said. Should they keep spending? Yes. I put roughly 70 to 75% odds that this CapEx program proves value accretive over the three to five-year horizon. This is the logic. Now, it gives you the downside, too. Why it's only 70-75. So, this brings up, and this is why guys don't make everything some kind of absolute. It's all thinking in bets. One, chip obsolescence. Two, off-balance sheet commitments. So, basically they're going through in the obsolescence. The fact that these won't last long enough, and their lives are only really four years, then six. Most people are talking about them being longer than six, but let's just assume that that is a risk, which has been brought up. The off-balance sheet commitments. If you've read this this week about 1. blah blah blah trillion of off-balance sheet for all these companies, all mentioned in here. Three, the backlog concentration and conversion risk. If you believe that that backlog of $514 billion is mainly OpenAI and then for Empire, then yeah, they have a concentration risk. If OpenAI goes bust, then they lose that. So, there are those risks associated with it. It's brought up. The verdict flips not on any single FCF print, but on whether contracted demand continues. This is the thing. It's all about the demand side, guys. This is what goes through Jevons paradox. This is what I'll spend some time on.
Now, I want to make sure this visual is there because I think your focus needs to be on the long-term flow. So, when I spend weeks saying sentiment has finally gotten to push back on me on memory and says there's no way to stop this. When I finally say, "Okay, Korea is breaking down and we're seeing the breadth break down." The only two companies that are going higher when it was all of machinery, all of construction, all of those things from mid-May. This is the short-term noise. Markets do not go straight up ever. Even if you're in a secular bull market, which we are for AI, there is no way to say this isn't a secular bear bull market. The only thing you can say is that it gets too far ahead. We need to have a big correction. I don't think we're going to have a big correction, and that's the reason why, if we do, I'll start seeing it on some of the signals within there. I saw it on the one side, and we do have a lot of issuance going on.
So, let me just use an example. Here's the S&P since over the last six, five and a half years. We've had a return since 2019. This includes COVID. 15% annualized. Seems like an easy bull market when you look at the chart. Here's the rolling 63-day. This is 22 trading days a month. This is the rolling three-month returns during that period. Look how many times they not only go negative, they go negative double digits. One, two, three, four, five, six. Every year we get the rolling three-month down to -10%. Even in a secular bull market of 15% a year. Now, I think this is going to be well above the 15%. The reason I created this is because for the next three to five years, I expect this to outperform the S&P by 15% a year, minimum. So, if that's the case, here's the rolling 63-day on the thematic portfolio. Could we go down another 20% based on the S&P? Yes. I don't think that's going to happen in this because I think a lot of this first wave was something different. Do I think once we make another leg higher it can? Yeah. Do I think we'll go negative during this period? Yes, I do. If I go back and look at the Mag 7, you'll see the same thing.
Three podcasts for you guys to either listen to. I will put all these in a subscriber side in terms of the summary so you don't have to, but there's a prompt room. But I listened to these three: the All-In one, Moonshots one, which I thought was fantastic. I love Emad Mostaque in terms of having his view on not only the technology from it, building an AI company and selling it, but also from a hedge fund perspective in terms of the economy, a macro economic viewpoint. And then Steven Ho, who did Forward Guidance. Those three were all interesting. Now, if any of you go to conferences and you go for a multiple-day conference and you hear a couple companies and then a couple macro strategists, they all talk about the same theme. I wanted to put something together, which I did first in here. I asked, I put all these transcripts into one context window. And the first thing I said was, "Go through all of them and give me the overlapping conversation from these three. And what are the most important points?" And what it gave me was, we're reaching model abundance, meaning the models just keep getting better. They're converging on very similar capabilities, and the price is dropping dramatically. We still have compute scarcity. So, all of these different things, these are the quotes. Intelligence can be copied. Power can be copied. Powered compute cannot. Jevons paradox, cheaper intelligence means more intelligence consumed. AI adoption moves from experiments to workflows only when it is cheap, private, and controllable. This is the enterprise adoption moving from token maxing to token efficiency, and then the inevitable winner of this thing is routing. Which means it will include Anthropic, but not only Anthropic. We're going to head to routing to the best place. So, that's what you take three of them. They're all good. It's like having a brainstorming session with a total of, let's see, that was probably nine people. And you get that cross-side.
Now, what I did next is using my thought process. I wanted to create a mosaic method prompt. So, this is basically creating a mixture of experts for forecasting. I'm putting this up on the subscriber side, so the prompt will be there. That shows you what it is. The human brain is a serial processor. Listen to three podcasts, and you experience them one at a time. By the third, the first one is decayed into a few vivid moments, and the connections between conversations, the exact place. This gets back to that conference thing. The most powerful thing is to have a prompt that basically goes through and gives you an output based on what you heard in a summary where all of these little nuanced points that keep coming out in there, you're connecting back. So, think of this again as using AI as the conductor of a symphony, and each of these podcasts end up being one of the people in the seats. So, instead of the first way I did it was just say, "What are the most important points that come out?" This is a prompt that literally goes through and takes the nuances in all of them, even if it's just a minor point, and then gives you a mosaic. Everything that I like to do in this as I go through these data is look at the mosaic. If I think there's going to be a bear market, then there needs to be a recession. If there's going to be a recession, there's certain things I want to see happen, like credit spreads widening, like profits already coming down, like profit margins already coming down, things like that. Not when they're all pointed higher, because there's not going to be any job firing of any big way if profit margins are going higher. It's simple things like that.
So, if you guys use it, one thing I would recommend, and I say this just because of the time of the year, sorry, I'm having trouble getting this. All right, there we go. Took a little while. Um, it's fantasy football time. Using this mixture of expert approach, the mosaic approach, fantasy draft prep has the exact structure this message was built for. An ocean of expert talk, a visible market consensus to beat, and a scoreboard that settles who was right. Think mosaic, you're going to have a bigger picture. This is one way to think of things in a mosaic perspective.
So, I wanted to give you the visual of that first context window of, we've got model abundance, we have Jevons paradox speeding up because we're having enterprise adoption, we're figuring out the routing structure. All of this means demand continues to go exponentially higher. The problem is physical infrastructure cannot scale at software speed. We will forever be short compute. I don't know when we're going to get to that point because as I go through the rest of this, and you guys have to trust me on one thing, if you hear there's going to be an overcapacity for a period of time, we could have overcapacity. If OpenAI is not generating the revenue and the market just says, "We're not going to give you any more capital," and no one wants to go through it, it could be a scary time. Again, I don't think that is going to happen because they are growing the revenues and they have a massive cash situation, but as you go through this, we are at a point where this is about demand outpacing efficiency gains.
So, I want you to think about the intelligence commoditization loop. The stuff gets cheaper. Jevons paradox drives higher demand. Demand outpaces efficiency gains. The physics wall come in, and this is all before video. I just want you always to remember how much more video, seeing the world, humanoids, full self-driving. We are still at the beginning stage, and as I kind of go through the agentic side, we are nowhere on the agentic side either in terms of adoption.
So, we've got Kimmy K3 and QE for the mind. I wrote this this week. If you didn't get a chance to read it because of the summertime, definitely read it, especially when I go through this part of QE for the mind, which I'll get into in a little bit. This is really important. This is the key thing. Kimmy K3 needs tons of compute. It is not a small model. It needs a lot. I can't use it. Nobody sitting at home can use it unless they have a lot of hardware. So, this is not disrupting the business. I used Anthropic every single day, OpenAI every day, Perplexity every day, Grok every day. Every single one of them, Gemini, every single day. Never once Kimmy K3. Kimmy K2.5, yeah, because it fits on my hardware. This is you still need the compute power. Kimmy K3 may be an important inflection point for AI, potentially negative for OpenAI and Anthropic while being net positive for essentially every other company in the world, and I do mean literally, yes. That is the QE for the mind part, which I wrote about.
Has received far more love than we expected. Over the past 48 hours, demand has pushed close to the limits of our current capacity. We're adding capacity as fast as we can. We will reopen new subscriptions. You can't be a company and be running out of compute. Sorry, it doesn't work that way. China's Moonshot pauses Kimi subscriptions amid hot demand. IPOs coming.
Aaron Levie, who I've referenced a bunch in here, he posts a lot, but I think this is important. He referenced this Daniel Kahneman talking about AI's biggest winners have the lowest margins. Low margin businesses may have become some of the biggest AI winners, and that's because even at a cost reduction of less than 1% of revenue, you can increase profits by more than 25% because existing margins are so thin. The main challenge is employee adoption rather than asking workers to learn the tool, they should embed agents. Agents are coming, and this is the stage that we're at now. The central takeaway is that AI should be deployed as infrastructure, not merely software. I want you to think about this because this is coming now. And what that means is we haven't really seen the profit margin benefit for the majority of companies, and some of the biggest revenue companies in the world do not have this benefit.
So, I wrote this paper in November of last year called QE for the Mind. And I talked about how artificial intelligence is flooding the economy with intelligence liquidity. This is what I was talking about. Under QE and ZIRP, companies use cheap money to engineer earnings. The playbook was simply, you borrow 2%, buy back stock yielding 6%, and watch your earnings per share rise, even if profits stood still. This was financial leverage. But now, a mere 5% reduction in expenses, go back to what I just showed you in terms of that post, yields a 20% EPS jump. You get far more leverage with reducing your expenses through AI and how it flows to the top to the bottom line.
So, then I just went into Fable and said, "Give me five S&P 500 companies that fit the paperwork." Now, this is the paper, Daniel Cornum's paper. And it just brings up UPS, Cisco, Jabil, Centene, and Walmart. You can go read what you want, but you can see these companies, they have a lot of revenues, but they have a lot of expenses, a lot of employees, they have a lot of friction, they have a lot of internal problems. So, if you can start to use agents for the workflows, it's going to have a big difference.
And here's the profit margins measured in the way one way that just uses the GDP statistics. So, this is government. This is corporate profits from the GDP numbers related to nominal GDP. So, this is another way to look at profit margins, and we're at all-time highs. And again, the reason I brought this up with the recessions, profit margins weaken into. So, the dot-com bubble, we had already seen this happen. Profits were already coming down. So, could this happen again? It could, but right now we're nowhere near there. And you see the lead time on this. So, we'll look at profit margins. It'll be a good gauge. Earnings, year over year, they're just growing. And again, these levels are isolated to basically times when we've come out of a recession, except for the internet boom. Everything else, you come out of recession, you have this boom. This was the housing boom in the China side. But look what happens before you go to a recession. Same thing. It's not just profit margins that go down, profits go down. Well, profits are accelerating right now. That's why it's the critical part for you guys. If you're going to get bearish on this, you're taking a low probability bet. Is it possible? Yes. Is it probable? No. And if you're betting against bonds because the hyperscalers' bonds are widening, they are the best balance sheets in the world, bar none. And you just saw what Google's numbers were. Don't look at this stuff in that way, and don't let this hype that goes through X's permabear bear porn get you in this position. We were in a slowdown. Everyone got too optimistic. Follow the technical charts. Follow the weekly stuff. There's no reason to chase this thing at this. This is right now in a quote-unquote mini bear market. We have 30% of the names below the 20-day moving average. Let's see gradually this stuff start to go higher.
Just another side of where we were in the earnings last week. And again, I'm not going to read this off, but it's just to highlight that the earnings have been explosive. Estimate revisions factor. So, just to show you how important the earnings are, even on a different basis, estimate revisions will show you weakness before meaning when the stocks are going down. So, I believe the stock market discounts. If we're seeing the factor of estimate three-month estimate revisions come down, meaning they go down, then it's fine. Even with the S&P side, the recent side, we've seen a momentum. We haven't seen a change. So, this is overlaid with earnings per share. So, forward earnings are way, way up from where they were. This is showing you as of this year. This is a quarterly number at 305. We've taken this number up significantly. So, estimate revisions are going to be the thing to watch as a factor. And this is where we are all the way back to 2010. It's almost a direct overlay. So, we've been in this part since the iPhone came out that estimate revisions have been the dominant factor to follow. And right now they're just saying that the stock market's fine.
This is another version of the consensus beats and the surprises, again, 93% as of Wednesday. The numbers I got from FactSet for Friday were less. Again, we don't have enough compute. So, if you want to find a place, again, the hyperscalers are spending money. People don't like the free cash flow. I don't blame them. There's a 25% chance this won't work for them. They need to see their multiple compress. For memory and all these things, these guys are going to receive money. It's just a question of when we actually hit peak earnings and when capacity is going to show up. I've seen no proof yet that anything can keep up with the demand. So, let's look at what's going on to see if anything is showing that compute is becoming available. GPU availability tightening back up. So, going right back down to where it was after a small time where there was some compute.
Be aware, I don't know if anyone has proven this to be correct or not, but Nvidia, according to the information, would be able to make 1,000 Vera Rubin racks per day, which is $630 billion per quarter. Now, I'm bringing this up because Vera Rubin driving performance for the lowest cost token worldwide. We are in the Vera Rubin spot. Hits the gas, next gen. We are absolutely in full production. There's a lot of speculation that they wouldn't be doing this. So, I finally released the Vera Rubin report that I spent all in total, I would say three or four months, really working through it, trying to find companies that were not part of the 100 name portfolio. I think there's 25 that are not connected to it. They're still going to be correlated, so I wouldn't expect you to go in there and we go through it, but this will give you guys, especially hedge funds, mutual funds, some more names to go through on the 800 VDC. What I did for this was I spent the time trying to find ones that would have a step up. This is the critical thing for the rest of this year. It's the reason why Marvell is still my biggest position. I believe we're going to see a step-up function in optical and photonics. I believe that we haven't seen the earnings flow through yet on these. I'm looking for beats that are going to be bigger, where we're at the catalyst point where you can look for where they have to revise numbers up significantly. Whether or not that happens for Marvell, we'll see. But at this point, I did buy another optical name this week as well, AAOI. And again, for various reasons, it's part of the portfolio, but this is really me scanning and looking for things that technically look like they've had a big enough correction and paying attention to data. I would spend the time reading this report and going through it.
For those of you who've been waiting, I did post part three and part four of the research skill architecture, the way that I go through and do the work for the Vera Rubin report. A lot of the idea for the mosaic thing came from the mixture of experts approach that I take in the deep research thing, which you guys have seen. These prompts will be up this weekend as well. I just wanted to bring Steve Hoe up again because I think this one, for those people who really want to listen to something, if you've looked at the token index, he's part of the company now, Silicon Data, that does that token index. He covers that. I've talked about how that's a ridiculous index to get bearish on. But he does go through everything. AI compute is becoming financial market. Compute demand should fragment over time. Enterprise AI likely settles around orchestration routing. We've talked about that. Token index is not showing demand collapse. The market is shifting from token maxing to token efficiency. I talked about that. Cheaper models, they pressure the frontier, but they expand the home, the total market. So, you can see how when you listen to Gavin Baker and you listen to the All-In pod and you listen to this, this is the same story. So, don't sit there and say cheaper models are going to destroy Anthropic. That's not the case. It's going to expand the total market. The pie is going to get bigger faster. And the orchestration level will be the smartest intelligence. So, if enterprises start doing more, it just everyone gets a bigger piece of the pie. It's just that Anthropic doesn't grow as fast as it was. There's competition now. And I talked about that with Pomp and how by 2030, I'm definitely worried about what will happen to all companies that are public companies, including the infrastructure companies. So, this is not a permable thing. Actually, I think I'm one of the more bearish people when we get to 2030, but that's down the road, and that may change if AGI doesn't come as fast, and if we get to a point where the governments are just making sure that the public markets keep going. Remembering demand remains structurally strong. Next phase is enterprise ROC. And this gets into what I wrote about, which is the QE side. I really think people need to focus on that.
I just wanted to highlight this. Michael Dell put this out a year ago, and this is becoming more important when I talk to people because people really believe we're running out of compute. We're basically in here for the most part, for the agentic side, so the reasoning side. Video generation. I just want to show you how much video it takes because every time I hear people, I had written a paper on VLMs, so video language models. Video generation. This is how much more it is versus baseline chat. So, chatbot to here. 10 seconds, guys. So, how can a movie be? How much more compute would it take to build a movie, a full two-hour movie, which we know is coming? The compute demands will be forever.
AI headwinds. Again, I showed this before. I'm just putting this up because all of these. There's going to be a cyber issue this year. We still have the data center backlash. We've got no power. We've got cooling issues. We've got memory bottlenecks. We've got government open source competition. All of these are going to be there, guys. That's why it's short-term noise versus long-term secular boom. As long as earnings are growing, all of this stuff is doing is stopping it from becoming a bubble. That's what that stuff is doing. The friction prevents it from getting too far, too fast.
I showed this. This is the software world. We are in this world. There is going to be a lot of things. And if every time these stocks come down, you start bailing out and start worrying and you start focus that Iran and the US are at war again. Oh my god, oil's going to go higher. How many times are people going to fall for? They go up, and then all of a sudden, when we need them to go down, they go back down. We've got volatility. We're always going to have volatility. Credit spreads. We're going to have periods where credit spreads widen. Right now might be one of those. Right now we're seeing it with the hyperscaler spreads going. The world does not end during these periods. It creates a correction, but the stocks usually lead. And then by the time things get better, the stock market's already gone up.
So, here's the oil price. This is the December contract for this year. So, we went almost back up to the highs. And again, I want you to look at this. We spent 2023 to 2024 around the midpoint of say $68. Now we're at $78, and people are panicking. When it could be $200, that was worth it. It's not worth it. Gas prices. Yes, futures are up here. Gas is continues to rise. So, are we going to see a bump up in the headline CPI? Yeah, I don't think it's going to be dramatic, but could we see a bounce back up? Of course.
Here is the 100-day rolling average, or the 100-day rate of change on crude. So, you can see what's happening as it is. Like, we already took the CPI up on the monthly number. There's we're not going very far, guys. Gas at the pump didn't come down all the way. Here's what we're looking at right now in terms of the inflation print for July, which would come out in the second week of August, I believe. So, the first time we're going to see a CPI be non-zero is going to be mid-September. Do you really want to bet on that? By the time we get there, maybe oil is back down. That's why this is happening. So, the first time that oil went higher, we saw 1-year inflation swaps go up with it. Then, inflation swaps led the way lower. They haven't budged, guys. Smart people are not paying attention to what oil is doing and saying that inflation is going higher.
Here are 2-year break evens, TIPS. Again, same story. War, come down. All right, we're not falling for this again. Oil's going to have to go a lot higher. Here are those two swaps. Here's year-over-year CPI over the last five years. I don't know how many times to show this. So, if you're worried about oil and the impact it's going to have on the Fed and all this stuff, stop.
I just want to show you guys this because now more and more people are jumping into the Fed, and they're going to have to raise rates. And like I said, they may raise rates. We're now at 35% for July. It's not going to kill the market. Um, here's wages. Sorry, here's 2-year rates. So, this is 2-year rates going higher. Here is wages. Median. Here's the relationship over time. You can't get past this. The labor market is not strong. Not when it comes to wages. Not when it comes to labor participation rates, labor force participation rate.
So, the Fed is heading into one of the most unpredictable meetings in years, and that's because, well, no more forward guidance. Everyone's freaking out. That means, oh, he could go. He might go. He might. 36% chance. We now have back up to one and three quarters by the end of the year. Here are rates around the globe going higher. It's not just a US thing. Growth is good. CDS on Oracle blowing out, so that gets people all scared. $7 billion guarantee. I can't even tell you how small that number is in a world of hundreds of billions of dollars.
So, let's move on to the Clarity Act. Bestin says, "I think we're at the one-yard line on the Clarity Act in the Senate." Still below 50%, so they're not listening to Bestin. David Solomon says, "It's time to advance the crypto bill." My buddy Chris Perkins, taps new leaders to bridge crypto, AI, and traditional finance. Bank of America, and someone he used to work with at Citi, Sandi Cull, wrote a very important crypto piece. And written many pieces on the importance of agentic AI to crypto and the blockchain, and I'm going to keep telling everyone who is not spending the time on crypto by the time a year from now goes on, you will have spent way too much time playing catch-up. Go spend the time understanding why there is no way to get around that globally the blockchain and crypto is being accepted by Japan, by Korea, by every single country. And now, as the Clarity Act is being attempted to get through, everyone's in a race, just like in AI, to see who's going to own something that is equally as important as AI, which is the global financial system. Pay attention to crypto and the blockchain.
Remember when Mark Zuckerberg said AI agent tech progressing slower than expected? Getting back to Sandy Cole's paper. I want you to think about that in the context of this movie, 50 First Dates with Adam Sandler and Drew Barrymore, and the reason I bring that up, if you remember it, this is memory. So, this is an analogy of if you want to guess why the agentic world is so much in this point. She is a context chat window, only remembering things within inside that one context window. What has to happen is, in the agentic side, the amount of memory that's necessary to store all this and know everything that he knows, every experience he's had, everything he's seen, everything he's connected, songs that he learned when he was two years old, the amount of memory that's necessary for that from an agentic side is massive. And that's why it's taken so long for the consumer agents to actually be there, because you need to understand the person. If Alexa's going to talk to me, I don't want to go back. I want to be able to say, "Not what's a recipe for this?" I want to say, "The last time we cooked this together, what did I do?" Think about all the things in life that you're just going to put into AI, so you never have to remember anything again yourself.
S&P just launched its first-ever crypto index and Bitcoin. I went through that whole thing with 50 First Dates, because I want you to understand that when we get to the point of the consumer agents, the crypto side is going to explode in terms of volumes. Once we get there, and this stuff starts growing, it's going to be big. So, despite everything that went on, the Clarity Act falling back down to the 30s, all this, Bitcoin had another quiet week. So, it's held the gains. It's still stable around $64,000. If you look at it, this is Bitcoin, this is Ethereum. They're both hanging in there. They've bounced off the lows. They are not out of their bear market. It is still a bear market until we can break the 200-day moving average on these.
Most importantly, I am showing this more and more. This is Dogecoin. Yes, this is a meme coin, but the most important meme coin. And to me, when people want beta in a market and they jump into things like GameStop, this is a signal that retail is active. The energy in retail is still nonexistent until we start seeing Dogecoin trend positively. And here's one way to measure it. We've been below this the 20. This is the 20-day moving average on Dogecoin. We've been below it since the middle of May. It's now 67 days. This was through Wednesday. 67 days. The longest in its history. It matters. We close above it, I think it matters.
Finally, one more. Add Russia to the list. Parliament passes law regulating the crypto market. That's it for this week, guys. Remember, if you want to get some more details as an RIA, reach out to Mark Whaling. That's MWhaling, mwhaling@22vresearch.com. We're doing a lot more for RIAs, but most importantly, everyone is starting to reach out on the crypto side. I will be launching the new crypto stuff like I've talked about in September, late September into early October. I'm looking forward to spending a YouTube a week on that for people who want it, but more importantly, I think everyone should be paying attention to what's happening in crypto. All right, guys. I'll see you next week.