AI崩盘结束了，接下来会发生什么

We're making it through the summer. The rally continues into all-time highs, and vol falls off sharply.

The biggest question I've been asked, and as I mentioned last week, I'm spending a lot more time with RIAs and institutions. The biggest question I received after, let's say, Situational Awareness and the breakdown in factor volatility was how to hedge AI going forward. So, I'm going to go through that.

I put together a prompt for you guys on something connected back to the thematic portfolio so you can run that. I'm also going to start to hopefully give you guys some stuff on agents. Gavin Baker interview. I'll basically make sure that I emphasize this is a must-listen to regardless of your views on AI. I've said a lot of these things, but I think as I take you through it, it's really important to make sure that you at least hear what he's saying based on a trip out to Silicon Valley. I didn't get a chance because I record on Fridays during the summertime most of the time. That'll probably be the case going forward. After I recorded, the U.S. intervened in dollar-yen, and I'll give you my thoughts on that because there was a major reaction in gold this week. I think this is all going to lead into Bitcoin, and I'm going to go through why and how the Substack I wrote this week connects everything back to it.

1. Follow-through confirmed: The 4-day Qs rally near 10% matches famous bottoms since the GFC; the three-month pennant breakout exceeded the entire range; S&P +3.6% on the week, thematic portfolio +6.5%; new all-time highs in IWM, NYSE Composite, and equal-weight S&P.

Last week, I talked about the bounce that had occurred in the final 2 days after the Citadel buying from Situational Awareness, and how it reminded me of prior panic situations where we finally had a cleansing event. You can't, especially in the moves that we had seen—which were historic in terms of volatility and the factor drop—not look for follow-through days.

What I wanted to see were 2 conditions. First, hold the bounce pretty much across the board. Secondly, I wanted to see factor volatility and the volatility measures come down to indicate that people were either hedged or had taken their gross down. I had seen the data that gross was down, but we obviously got almost all of that.

On the Qs, I posted this on X, but the 4-day rally—which included 1 day last week and the first 2 days of this week—ended up getting close to 10%. You can see these 4-day rallies over time during the last 17 years since the Great Financial Crisis ended. A lot of these are obviously very famous bottoms, in terms of Liberation Day and, obviously, COVID.

This one here in 2022 was a bear-market rally in early 2022. The difference was that we were below the 200-day moving average. This is something I brought up before: bull markets stay above their 200-day moving average, which is pointed upward. That's what we did in 80-plus percent of the AI thematic names, which is why I was looking for where to put money to work. That's why we went through it that way.

In terms of the S&P's 4-day rate of change, the biggest one over the last year was 6%. Just to give you an idea of how vicious that rally was, we had been trading in a range. Every week that I came in here, I kept saying the same thing: this is a pennant-flag formation, or whatever you want to call it. To get bearish on charts like this is a very dangerous thing.

In particular, when you're seeing the Qs fall and factor volatility collapse, you're looking for contagion. That's very difficult to happen when you have earnings growth at post-recession-type levels. The range that we were in basically ran from May to July. We're talking about 1 to almost 3 months—a 3-month pennant. The rally we had was 6% bigger than the entire range that we saw during that period. That is not something to fade.

The S&P was up 3.6% for the week, the Qs were up 5%, and the thematic portfolio was up 6.5%. It was the second-biggest week of the year.

In terms of where we stand year to date in the S&P, I'll say this over and over again: this is a bull market. Every sector is up on the year, and the S&P is up 13%. There was a pennant formation in there for a solid 3 months, followed by a breakout higher. We not only held after Monday and Tuesday, but we made new all-time highs today.

The Russell 2000 made new all-time highs this week. The New York Stock Exchange Composite made new all-time highs this week. The equal-weight S&P made new all-time highs this week.

2. Volatility mean reversion: Factor vol, tech momentum vol, and single-name vol all collapse — funds are hedged; VIX at its lowest since January; no credit stress in CDX despite hyperscaler CDS chatter.

In terms of the KOSPI, because I had a lot of people reaching out saying this was still a problem, this is not a problem. The KOSPI is up 61% year to date. This is the KOSPI 200. The technology sector is up 116%. This is after a collapse.

I'm not going to go through it today, but on days now when SK Hynix is down slightly, the names and sectors that I showed—the machinery and construction side—have been leading on the way up and holding their gains. Again, this is what long-short is about. You get deleveraging on both sides, and at some point you start to go in the other direction.

This is the DAX breaking out to new all-time highs. This is the FTSE breaking out to new all-time highs.

Here's the big thing for the week. I showed you the market-neutral momentum. This is the sector-neutralized market momentum. This is market momentum, technology momentum, and industrials momentum. If you look at all of these, the drop-off in volatility over the last 6 days says that funds are hedged.

Here is the 7-day realized vol of tech-momentum vol, basically collapsing, which is indicative of what I just showed. The main point is that we've seen factor volatility drop off. That is usually what happens after this.

As I've said, it takes a long time, particularly when you're below the 200-day moving average. You've knocked people out. Anyone who's vol-targeted just cannot lever back up. Any of the pods that were hurt and had their risk management reduced are not going to immediately put risk back on.

It's the summertime. This is why we're going to rally on lower volume, and everyone is going to be put in a position of chasing. They don't want to chase. They're going to read about seasonality, September, the midterms, and all of this stuff.

In my opinion, I've heard enough podcasts this week from so-called market people who are in the game, I think, of being bearish. They remain bearish during the rally and are looking for a false breakout. We'll see if we get that, but this is really hard to be bearish on this stuff when this is going on.

The VIX EQ for single-name vol has fallen sharply, too, and it's basically back to where it was before this whole thing rallied—not where the panic started, but where it goes. We've now seen mean reversion on the vol side.

In terms of the VIX, you're at the lows since January. For everyone who reached out to me saying contagion was coming and buying S&P puts, you not only lost on the direction, you lost on the vol. The vol never really was indicative of the panic that we saw, with factor volatility at 45-year highs on some measures. We never got it in the S&P CDX.

3. Earnings engine: ~50% earnings growth, 15% revenue growth, margins expanding from 13% to 17%, PMI approaching 60, durable goods confirming, the AI capex certainty is driving everything.

I talked last week about the silliness in the hyperscalers and how people were looking at the CDS of the companies with the best balance sheets, especially when they go into X. Here's what the CDX index did: basically the same as the VIX. There are no credit issues at all.

In earnings season, I'm not going to show you this number, but you know it's up near 50% in terms of earnings growth. Sales growth, though—we're just killing it. This is across sectors. The AI trade, from an earnings basis, is taking everyone up with it.

Here's revenue growth, now up at 15%. I want you to look at the fact that this has basically gone on since Q1 2025. What happened in Q1 2025? Donald Trump took over as president, and sentiment changed on the business side. He put tariffs in place, and everyone got worried. But the AI trade was now in full force, and that's what is driving this.

This is the certainty around AI CapEx. These are the CapEx numbers basically driving this, and it's having a huge impact. Margins were 13% at this point last year; this year, they're 17%. Again, it's not just technology. Margins are increasing, and as AI adoption and AI agents advance, you're going to see more of this.

On breadth, I highlighted last week that it was a good sign that we basically had 2 prints higher than these 2, and we followed through this week. The 50-day is up to the highest level since July 14. Most importantly, the 200-day slope has 80-plus percent pointed upward.

This has been the best quarter in years in terms of earnings-estimate beats. If you don't believe the earnings, look at the revenues. They continue to move higher.

This is the PMI, which soared higher. I can remember that a year ago, PMI was below 50. I was writing papers on the PMI heading up toward 60 and what that would mean, all on the back of the AI trade. Now it's fairly consensus, but we're almost at 60. More importantly, this implies that we're going to get closer to it. We continue to move higher, and revenues keep going up.

We got durable goods this week. Non-defense ex-aircraft was again up at a level that is very close to 60. For anyone who has faded this, it's not just the earnings reports now. You're seeing it at the PMI level, in global stock markets around the world, and you're still seeing revisions go higher.

4. Structural regime change: AI compresses economic time; digital employees are coming fast; terminal value gets harder to estimate; ubiquitous AI creates crowding, bubbles, parabolas, and speed crashes. The speed crash has occurred, the bottom is in on a probability basis.

Here is the AI thematic portfolio with all 4 parts. Remember, you've got the Morgan Stanley one overlaid with my 100-name one. The Morgan Stanley one was just created later. For those of you who want to trade it, you can call up the Morgan Stanley desk. They have it. You’ve got the concentrated portfolio, which has 10 names, and then the more global concentrated portfolio, which has 25. Again, these were put together so that you can have an index. I think if everyone learned anything, it’s that indexes are really important when you have this much volatility.

I want to get into the hedging side now and some of the conversations. I showed this slide last week. I believe the market is changing structurally, and this is never going back to the way it was before because AI is compressing economic time. I think you’re going to see that with the rest of the year.

I’m going to go through the payroll numbers from yesterday, and I want you to reach out to every one of your favorite economists and ask them why they’re not seeing the weakness. This is not about people necessarily being replaced by AI. This is not about jobs being lost, but this is about wages. This is about the fact that digital employees are here, and they’re coming incredibly quickly because the adoption pace is picking up.

I’m going to take you through Palantir, which is a name that I own personally, but also a name in the model portfolio, which just showed their corporate numbers. Since we can’t get numbers from Anthropic and OpenAI, you can look at Palantir’s. These numbers are insane in terms of adoption. When adoption picks up, AI agents are going to come, and when the AI agents come, not only are you going to have less hiring, but you’ll start to see more job losses in public companies.

You’ll have job creation in private AI-native companies, which I’ll go through. Terminal value gets harder to estimate, which makes it more difficult when companies all of a sudden can go from trading at a 30 P/E to a 20 P/E in a very short amount of time. We’ve seen that in many, many places, and I think you’re going to continue to see it. I think Google’s kind of going through that right now.

I won’t read all of these, but all of these things include this one, which Gavin Baker brought up, too: the ubiquitous side of AI. The fact that everyone has access to it, everyone can create a backtest, everyone can create an optimized portfolio, and everyone can find the names that are working leads to more momentum. It leads to crowding, and again, it leads to bubbles, parabolas, and speed crashes.

Exactly what we had this year—what I wrote about, the speed crash—has occurred. It’s over. The bottom is in, as far as I’m concerned, on a probability basis. If we take out the lows, I’d be surprised, mainly because I continue to believe people are way too bearish relative to the reality of the earnings growth that’s coming. I also don’t believe you’re ever going to see that type of move again that we saw in the first quarter, but you never know.

5. The new hedging toolkit: Factor vol is structurally higher and index puts are broken; the framework combines a 50-day rate of change threshold on the 100-name thematic index with breadth breakdown the same signals that flagged the top before the Micron sale and rotation into silver, Bitcoin, and Lilly.

On the hedging side, how do you deal with bubbles, parabolas, and speed crashes? Well, those are all based on price and time, which I started to show you last week, but I’m going to go through one of the demos in terms of how I’m handling this. For the SC subscribers, this will become part of the way that I’m going to identify things.

I built turbulence models. The turbulence models, because everything was so correlated, never really showed leading problems. The breadth did, and that’s how I’m going to go through this. I try to learn from what happens and look for the signals that’ll warn you.

This is one signal that will warn you when factor vol continues to move higher and break away from index vol. The reason factor vol is so much different than, say, single-name equity vol, which also broke away and I showed last week, is that this has a long-short component to it, and this is directly related to gross leverage.

The great thing about factors is that, as gross leverage is going to all-time highs, I’m making the call that we are at peak gross leverage and that we will see leverage come down over the course of the next few years because of what I showed before: AI is compressing time. It’s making it more crowded, and it means the Sharpe ratio of this stuff is going to come down.

I think factor vol is structurally going to be higher. I don’t think we’re going back down here. Maybe we come back down into 25, but if we spend the rest of this year in the 30-to-25 range—and again, this is an equal weight of 4 separate factors—it’s not just momentum, because I have value in there, I have quality, and I have beta. I’m averaging them all in there.

You can see the difference in the breakaway from the S&P, which means using index puts is not going to work the way that it used to. So, here’s what I am using, and this is the price-and-time side. Right here, when we got to 50%, or almost 50%, this is a 50-day rate of change on my thematic portfolio.

For those of you in the AI side who are more tech-focused, more industrial-focused, or more utility-focused, you need this tool. You need to have access to this data, and the reason is because there are no indexes that are diversified to this level. This is global. This is across sectors. This involves 100 names: chemicals, power companies, and industrial companies. So, it is diversified.

A 50% move over 50 days—50 trading days, which means slightly more than 2 months. To be up 50%, you’re going to fall. Here are the historical levels. At 20%, I think the risk grows, and I think that’s when you need to combine it with this. These happen to peak at the exact same time.

Rate of change does not mean that the thematic trade goes down. It just means you start to run into problems. I’ve always used rate of change. In the past, before we had exponential innovation, my model gave a recession signal in September 2007. In September 2007, the warning went out that we were going to have a recession.

The reason was because the year-over-year S&P—the 252-day rate of change of the S&P—had basically started to turn negative. Once that happens, historically, that means we’re at a bad point. When you add in corporate profits and jobless claims, again, this is before exponential AI or before exponential innovation with software.

For me, this is the way I’m replacing it. I’m looking for the rate of change to be up at a level that doesn’t make sense, which shows the crowding. It shows the AI agents. It shows everyone pushing everything. Then I’m looking for breadth to be breaking down within the 100 names. You need a diversified portfolio, so this is the way I’m going to measure it.

I gave you guys the warnings back here. We didn’t make a high in the index until right around here. Then we didn’t violently fall on the memory names and stuff here. I talked about selling Micron in here. This is when I got rid of all of it. It was early. I rotated to silver, Bitcoin, Eli Lilly, and some other names.

The main point was those names in the month of July. Go look at what they did. Silver and Bitcoin—while silver was flattish, Bitcoin was up a good amount. Lilly was up. You’re trying to navigate where you’re still in a position to buy when you get the puke.

This is the way it looks, and this is the way I’m going to go through it. This is the whole year, actually. It’s the rate of change. Once you break above this level, the warning sign starts.

We broke above this level and started peaking in terms of the breadth. We were up at 90%. Believe it or not, we were above the 20-day moving average, and then gradually we broke down. You can see how this goes. Now, the 4-week average peaked here and rolled over.

As time goes on, I believe this is going to happen more often. I think this is the way you’re going to have to deal with it. I’ve had a lot of subscribers reach out about a bunch of names that just haven’t bounced yet: things like Chemours—I own this one—Fluence, which I own, and Entegris, which did go higher and I own.

Palantir—I didn’t include the price today—is all the way up here now. For all the Palantir haters, all the short people, and the people who didn’t go, you’re missing something on this one. It’s the only software name I own personally. Then you’ve got Cadence and Synopsys, which are part of the thematic portfolio and are software-design names. I’ll probably be talking about them soon because there’s been some stuff.

That’s the way that I’m working through things with people to automate on the hedging side: to use my index for that, but also to use the breadth and the different parts.

6. Agents and tools: The earnings mosaic prompt for subscribers, plus a real-world example, my son spinning out sub-agents from his AI chief of staff at his internship. If you're not using agents, you're falling behind.

This is the thing I was saying on an earnings mosaic. What I did create—and the prompt will go up over the weekend—is a prompt for you guys that will basically spit out something like this. You can use it in Gemini, you can use it in ChatGPT, or you can use it in Claude, Gemini, or whichever one you want—Grok.

Basically, you take my technical sheet, update it, put the technical sheet into ChatGPT, Claude, Gemini, or whichever one you want, and then just run the prompt. It will go out and get the earnings commentary from the week and put it together with a bunch of different things, including a look at the mosaic side.

This was from Eaton. You had Eaton report this week. You also had Schneider report. They’re both on the electric side, and they both happen to be participants in Vera Rubin. It will go through and give you basically a recap of that, with some of the most important commentary that was in there. I’m trying to help you guys, especially as time goes on. So, this is what it does. This is what you’ll see up there. I’m trying to keep you guys going with the AI side. Again, if you guys want access to the tools, reach out to Mark Whalen. Go to the website. We’re doing more and more of these, but in particular, I’m doing more and more presentations for RIAs where I’m also doing regular calls with them. They’re setting things up that way. So, you can reach out to Mark.

I did want to reference this because this is where we are on the agentic side. This is a text exchange with my son. He is entering his junior year of college, and he is not a computer science major. I put him through a Python boot camp when he was a junior in high school.

I have my AI chief of staff spitting out new sub-agents on a regular basis. This is what he is doing at his internship. He even said, “I prefer GPT. Gemini coding is insane, though. I got the information from YouTube and AI.” Your kids have to be doing this, guys. He is guaranteed to have a job at whatever firm he works for because he has already built a chief of staff and is spitting out new sub-agents.

7. Gavin Baker (must listen): No negative demand metric anywhere; Nvidia at its lowest forward PE in 10 years; B200 rental prices nearly doubling from ~$2 to ~$4/GPU-hour in seven months; acute compute shortage with only 250–500K agentic AI users out of 8 billion people. There will never be enough compute.

There is absolutely no reason why your kids cannot do this. Remember the knowledge-brain thing that I put in there. I will be doing AI agent stuff for the subscribers, and I will probably be going through some videos as we get into this year. I have a lot going on with the crypto side, but my goal is to really make sure that you guys realize that the agentic side is here, and if you are not using it, you are falling way behind.

So, let’s go through the Gavin Baker side. Again, this is a must-listen. The reason it is a must-listen is because very seldom do you find someone who is this knowledgeable about the AI trade from a very high level, which Gavin is. He speaks about it regularly, he is asked questions about it regularly, and he invests in it. It is very hard to find people who know this from the industrial side all the way through.

I am not going to read all of these. I am just going to go through this. The speaker argues that the AI sell-off is disconnected from the quantitative data. He says he has not found a real negative demand metric.

I just want to make sure that if you listen to this, even if you listen to the first 5 minutes, these are all quotes from the first 5 minutes. Nvidia, as we record this, is at the lowest forward P/E of the last 10 years. I have bought a decent amount of Nvidia over the course of the last 3 weeks. It looked good for a while, broke back down again, and now it is looking better.

I believe that with Vera Rubin coming out, with all the chatter around this, and with the fact that it is this cheap, but also with the token-per-watt needs and realizing how quickly they are coming out with this stuff, I just cannot see a sign, given what GPUs are doing and the price of them—which he goes through—that they are not going to benefit from this at this level. They have done their multiple compression. It has been significant.

It is not going to double for you, I am sure, in a short amount of time. But when you are looking at safer names that have already seen their multiples compressed, I still think Nvidia is one of the best ones. Not a single instance of deceleration. Nothing. In fact, every metric is accelerating.

Whether you look at GPU availability, GPU rental pricing, or the spot price of DRAM this month, everything is actually accelerating. When you go through all of these and hear what he says about Anthropic, what he says about GPU compute prices, and the anecdotal information that he brings out, he is basically saying that these companies—Microsoft, Meta, Amazon, the hyperscalers—are the critical part of this.

If you have been bearish on them, which most people have, you have called this a bubble and you have been wrong. You have said return on invested capital will not come through, and you have been wrong. Maybe you will be right. The CDS is wider for justified reasons because they are still leveraging up more and more, and I will go through the reasons why that does make sense in a second.

But you cannot argue the fact that at this point there is a chance that they could see significant revenue in the very short term, if for no other reason than that they are building so much compute and the value of that compute is going up. He talks about it here: “The same type of B200 cluster that rented around the mid-$2-per-hour range 7 months ago is now expected to rent for just under $4 per GPU hour.”

Remember all the stuff about how these chips will not survive their lifespan? We have a bottleneck in compute, so you have to recognize that all compute has value right now, regardless of where it is. Open source would love to be using this stuff. Let’s just forget all the things that you have heard on this. Do not get bearish on this because of something that is not true according to the data.

If you see this data start not to happen—and that is what he goes through—it is like the breadth that I have on the index. If 1 thing is going down, that is fine. But when people look at the token index and make that their focus, even the people at the place that provides it say it is not bearish on the market. It just means more and more of the usage is going to open source, which makes sense.

It does not mean it is bearish. It does not mean that Anthropic’s revenues are not going to grow. But very smart people that I have known in the business for a long time were saying that to me. We have reached a point where, if you can listen to someone like him and go through it, it is there. It is rewiring the market structure.

He got into the same thing that I talked about. He even brought up the probabilistic Bayesian interpretation of the future, which I write about and talk about all the time. AI compresses economic time, but it may also compress market time. The month of July for momentum was worse than any year in the data that I have. That is how we are compressing market time: bubbles, parabolas, and speed crashes.

Here are the key quotes from the transcript. You can pause it and look at them on your own, but here is the most important one: “We are in an acute compute shortage.” At the same time, 500,000 people—maybe 250,000—are using agentic AI. My son is 1 of them. He is spinning out sub-agents, but only 250,000 of 8 billion people are doing this.

What happens when we go from 500,000 to 1%, to 100 million, to 500 million? I say it all the time: there will never be enough compute. Ever. We will continually be building compute. We will be doing it in different ways, and we will continue to get more efficient. Everything will keep happening, but nothing is on the horizon that would make that go away.

8. Hyperscalers, memory, and disruption: The capex math, negative free cash flow for years means eventual utility-style multiples; Google's talent frustration (Jeff Dean) signals the bureaucracy problem; Palantir's 149% commercial revenue growth shows adoption; new business formations go parabolic; and per Elon Musk, memory is the limiting factor supply +20%/year vs. demand +200%.

When we see the prices of older GPUs start to head down, at least we are at a point where the newer stuff is saturating some of the demand ahead of adoption. But the adoption is happening because of agents, and once you set them off, they are tireless workers. It is exponential against the physical side. No amount will ever be enough.

The math behind hyperscaler capex: this is my post-earnings paper that went out this week on the hyperscalers and the reason why you should not be playing for them to break apart. I got a lot of questions saying, “Does this mean you are bullish on hyperscalers?” No, not relative to the infrastructure trade. Nothing changes on that front. I still think they are going to have heavy competition.

Here is the other reason why: the hyperscalers’ quarterly free cash flow, when we think they are going to go positive as a group according to Goldman Sachs. Look how far out this is. Here is the problem: they are going to keep spending money, hoping they get the revenues in the door.

I understand that they have the cloud revenues, but the cloud revenues are not going to be enough, in my opinion. You are going to need to have products. You are going to need to have other things, or you are going to get priced like a utility or a real estate company. Eventually, when compute is not as outsized in terms of supply versus demand as it is now, we could have a lot of innovations. We could have a lot of things happen on the algorithmic side. There could be breakthroughs in hardware.

When we go to humanoids, cars, and all of this, everything could change. I still believe we are going to need more compute and more hardware, but whether their businesses will be there is a question. That is why their multiples should be compressed and their CDS should be wider. There is a risk there.

It is not a risk that I want to go out there and trade. But if you are buying their paper and you want to buy some CDS, of course you do. If you are trying to short the AI trade and you want some exposure out there, of course you go buy CDS in this, especially when it just keeps moving.

I just want to bring this up quickly because I think this is a sign of the problem inside these companies. I also believe that what happened this week with Jeff Dean leaving, and with Demis Hassabis moving into another area, is a big deal.

Google is the perfect example of a company where they release AlphaFold, which makes them no money. If it makes them some money, it is coming through something. But given the number of people using it, releasing it for free has probably created some conflicts inside because they are spending tons of money and taking on tons of debt. They need actual consumer products for corporations. They need something to come through.

I think this is where the people who are trying to leave a legacy with this technology—the talent—are getting frustrated with the bureaucracy at big companies. I think you are going to see this over and over again. It is the reason why I am negative on all public companies by the time we get to 2030. The bureaucracy is going to be a killer. At the same time, you are going to have AI-native companies growing rapidly.

I just want to remind you about the memory side. I wrote this paper in June, “AI = Memory.” This was after I had sold my Micron. This was not to buy anything. It was just to let people know that memory will remain in a shortage, and it is something to watch when it comes to the trade. It is also something to watch closely to see if we ever get to the point where it catches up to compute.

So, on the SpaceX earnings call, Elon Musk basically reiterated what I wrote in that paper. The limiting factor currently is memory. Memory output is increasing by around 20% a year, so supply is increasing by about 20% a year. Demand is increasing by 200% a year, maybe higher.

If you've got demand increasing faster than supply, Economics 101 would suggest the price increases. Again, I'm bringing this up because you should watch memory, and more compute does not solve the bottleneck without more memory. The memory side is going to be needed, and that's what I wrote about: memory is needed for all parts of this when you start thinking about what AI agents need. Especially when you get into consumer agents, we need lots of memory for everything that we do.

Palantir shares jump on otherworldly sales, improved forecast. This is what I was saying. This is commercial revenue. This is the adoption side of companies. The growth rates are 54% and 64%, year over year. They're not only growing by this much; they're still growing faster, at 149%.

In Q2 last year, they had $306 million of revenue. This year, it's $764 million. Whether it's quarter-over-quarter growth of more than 20% or whether it's something else, it doesn't matter. These guys are growing rapidly, and everyone is focused on them. Commercial was not their business, so it's starting from a small base. This just shows the adoption pace that's going on.

They closed 220 deals of at least $1 million, 98 deals of at least $5 million, and 78 deals of at least $10 million. The only reason I brought this up is to show you that the bear porn just remains out there. Here are new business formations. This is what the story is, guys: these are AI-native companies. They're starting their own businesses. That's what I did. This is what's happening.

9. The debasement trade: 30-year yields at highs globally; the US props up the yen in the first coordinated intervention since 1998; nothing stops this fiscal train; gold's third-biggest week in 16 years (+7.2%) breaking above the 50-day, with silver and platinum following.

It's going parabolic with AI adoption. It all started with Opus 4.5: the rise of million-dollar companies with just 1 employee. This is what's happening to disrupt the profits of Fortune 500 companies eventually. It will take some time because these are small, but it's coming.

The age of abundant intelligence and scarce Bitcoin. Two printing presses running at once. One dilutes the unit you measure wealth in. That is the debasement of your money. The other dilutes the durability of the corporate claims you bought into to protect that wealth.

This one is from the government and the printing press. This is the AI printing press, which will disrupt public companies. That's in my Substack.

Thirty-year yields. I put this line here because, basically, we took it out. This was a big story. I'm writing a paper right now on JGBs as the widowmaker from the 1990s and how, every time I get a technical chart or hear a person saying, “We just broke out, this is going to collapse, rates are moving higher,” stocks almost always rally.

The reason is that the government doesn't really have a choice. If you're going to fight something and you want to be bearish on equities because of bonds, the government clearly is going to do something if bond yields go too high. They've done it repeatedly. They did that on Liberation Day.

We don't care if equities go down. Then, when the bond market went down, they cared that equities went down, and they had to do something. Here is the relationship between corporate tax receipts and household equity-asset values. If asset values go down in the stock market, you're going to have tax receipts go down at a time when we have 5% to 6% deficits.

That's why Lyn Alden wrote about why nothing stops this fiscal train. It is still in place, guys. We learned more about it this week because it's not just the U.S.; it's the whole globe. When you see these charts, you've got the U.S., you've got the U.K., you've got JGBs, and you've even got Bunds. They're all 30-year yields around highs of some level. And what happened?

On Friday night, Scott Bessent had to explain why he helped Japan strengthen the yen. I believe this is more of a signal. Scott's obviously a savvy market person, and he's trying to find a way to help the BOJ for a variety of reasons. One of the reasons is that they own a lot of Treasuries, and he doesn't want them to have to go into the market to sell the Treasuries in order to go out and buy the yen.

First coordinated move since 1998. Long view: U.S. props up the yen to support Treasuries. Then Bessent is urging the Fed to help defend the Japanese yen using FIMA and hoping to expand its capacity.

Then you've got The Wall Street Journal and Nick Timiraos: “Why Bessent is leaning on the Fed to help prop up Japan's currency.” I want you to remember that because last week Kevin Warsh spoke. Last week, you had situational awareness. You had Kevin Warsh. The market got angry at Kevin Warsh because he neither raised rates nor said he was going to raise rates.

He gave no guidance. Nobody knows what to do. He didn't even say, “I think we're getting the sense here that it's going to be really hard for Warsh to raise rates.” If the Fed's going in to help Japan, we need interest-rate differentials there, and he's actually going to the Fed and trying to see if they can raise the borrowing limit, it seems like the Treasury and the Fed might be a little closer than they used to be.

Then we start talking about the unintended consequences, and I think the joint U.S.-Japan support of the yen is unusual. The way it is being financed is unprecedented and adds liquidity when the punch bowl of the U.S. economy and markets is already overflowing. He pledged to do whatever it takes to help.

Again, I think the bazooka—he doesn't want to do anything—but I think he's hoping that we can stay away from 164 and maybe they'll raise rates in September. But if Warsh raises rates in September, then what does it really accomplish? That's one of the reasons why I'm shocked that, after these last 2 weeks, we still have a 45% chance of a rate hike in September.

That's why I like all the debasement trades, as I talked about a few weeks ago, thinking that people were getting way too over their skis on rate hikes. Remember, a lot of places were talking about 3, at least 2, before the end of the year, and now we still have 1 and change in there. We'll see where it goes.

The policy fails to tackle the underlying problem and can easily backfire, not least by drawing the Federal Reserve into exchange-rate policy at an especially awkward moment. That's the awkward moment I'm talking about. The Treasury has now spent all the ammo in its Exchange Stabilization Fund. It can probably find other ways, but unless interest-rate differentials change, it's all a waste.

So, we're back to the interest-rate differentials. Then a former Japan PM warns FX action won't save the yen without a 2.3 trillion yen growth strategy. This is running it hot into a scarcity of tools that the central banks can use. Can the U.S. Treasury save the yen? Kevin Warsh is facing more than a communication problem.

I wanted to bring this up again because I think last week was a big week: situational awareness, Kevin Warsh not raising rates, and then the Fed doing what they did with the BOJ—the first coordinated move since 1998. With all of that, at the end of a historic month where we had deleveraging of significant size, there are a lot of similarities between October 1998 and now, a lot of them.

I believe that, when we look back, as long as those lows are in place, that will be a time when we started to see a reality check: when the U.S. is intervening in the Japanese currency market to help its Treasury market, the fiat system might be in trouble, especially when I go through some of the charts.

We're at 45% as of Friday for the Fed to raise rates, and that's after the payrolls number. We're at 63% for the BOJ. Darius Dale put this out on the quarterly refunding: “The Fed has little choice but to continue monetizing U.S. sovereign debt over time.”

Basically, the refunding announcement. Then there's whatever people thought Kevin Warsh was. Again, this is why nothing stops this fiscal train. We just have too much debt and too big a deficit. Remember when Paul Tudor Jones went on and did the math and said there's just no way to solve this? They're running it hot, but they're not making any impact.

During Donald Trump's tenure as president, July was the biggest 30-year yield rise in a month. Back here is when he drew the line in the sand for the equity market. Here's what happens to the S&P after the bond market ends up going higher. The bond market goes higher. Guess what? You're starting to get the picture here.

The S&P has rallied when rates go higher because everyone starts getting bearish on it, and the Fed or the Treasury needs to step in. Nothing stops this train. We keep doing the same thing. Gold is waking up to the possibility that maybe Warsh is not going to be bad for gold. Maybe he's not going to raise rates.

Over the course of the last 16 years, this was the third-biggest week for gold. This was one of the largest printing-press weeks of all time. This is when gold was out of control. This was a big week in gold, up 7.2%.

Here's a chart. Gold broke through the 50-day moving average. Look how long we were below the 50-day. We broke below the 50-day over here and hadn't been back above since March. In fact, it was the third-longest time that we had been below the 50-day moving average.

Here are the returns 3 months out from the time of breaking back above. We'll see what this one is, but at the breakout, year over year, most of the time these things were down big. Gold is in a bull market. Again, I go back to what I said before: when you're in a bull market and you correct, then consolidate, and then break back above the 50-day, things can run pretty hard.

I think there is an absolute reason to be thinking that gold, silver, platinum, and Bitcoin—all of them—assuming that the Fed is not going to cut rates in September. Again, I'm surprised we have 45%.

We'll see what happens with the inflation data next week, but I think we're getting closer and closer to this, and the market is telling you the signals. People are saying, "Well, why don't I just short Treasuries and go buy gold and go buy this?" I think either way, right now I'm going to win because they're going to come in and stop it, but I don't think it can go down.

So, here's what we've got. Platinum breaks above the 50-day, and again, it's a nice rounding chart. We'll see if there's follow-through. Silver finally closed above the 50-day today. I have a good-size position in silver because of the AI side, but I like it now because of the debasement side.

I showed this last week. We're at the highest dollar-long position. So, the dollar's already broken down, but this is just fuel for the fire for gold, silver, platinum, and Bitcoin. And again, just a reminder: still 44%. Not just that, we still have more tightenings in. I think all of these should be coming out. Personally, that's just me.

10. Labor weakness and Bitcoin: Weakest six-month aggregate payrolls since 2012 ex-COVID, no job creation outside healthcare, wages back at 2019 levels; Bitcoin holds firm despite Clarity Act odds collapsing, Saylor selling, and a cold storage exploit resilience ahead of the agentic kick-in.

The reason is this inflation thing is driving me nuts. We have 1 inflation data point, which is upward. This is the Truflation core number relative to the CPI core number. You cannot get away from the fact that it has led, and it has gone. This is a violent move lower over the course of the last 4 weeks. So, the year-over-year Truflation number just continues to fall.

This is on the same scale, people. Where this was around the same at the beginning of the year, you've had a slight uptick here, and you've had this collapse. Go read how they do their work. If Kevin Warsh is looking for better data points than just taking the way something was calculated 50 years ago, at least give Truflation a chance, along with trimmed mean and the other ones that I've shown.

I think the inflation thing is a disaster. So, let's go the other way. I'm sick of economists and the payroll side. The job situation is not good. Yes, there are low claims, and part of the reason there's low claims is that the labor force is coming down, but they never talk about that.

So, the supply of labor is going down because the labor force participation rate has come down sharply this year. If there's less labor supply, shouldn't that mean wages are going higher? Here are the wages. This is the quits rate and the Atlanta Fed Wage Growth Tracker on both measures. We are back to 2019 levels. You can't tell me the labor market is good with those metrics.

Let's just take the average hourly earnings that we just got, which is now down to 3.2%. The white line is also where it was before 2020, overlaid again with the Atlanta Fed median wage growth right here. You can't tell me the labor market is good.

Let's go a different route. Let's now take the aggregate payrolls. This is the aggregate payrolls, which is taking weekly hours, average hourly earnings, and the number of jobs created. I've talked about the fact that the only jobs that have been created over the last 18 months are in health care. When you strip out health care, there is no job creation. That is not a good jobs market.

AI agents—or AI—are having an impact on hiring. It hasn't yet had an impact on firing. But again, if we're having job creation with entrepreneurs and we're gradually losing people on the other side, which is the way I think it goes, I don't think we're ever going to see big job losses.

I think the demographics will kick in. We'll see more people retire. As the stock market goes higher, more people will say, "I don't need this anymore. I've made enough money."

So, here are the last 6 months. Just look at these 6 months, guys, and take it back over time. This is a very weak period. And if you want to see really how weak it is, this is the weakest 6-month rate of change in aggregate weekly payrolls going all the way back to 2012, when you throw out the COVID time period.

I don't care how you measure this. Maybe it'll change, but right now the data is a fact. The number of people looking for work and the number of people counted as unemployed both fell. We're losing people in the labor force.

This is 10-year yields, the 20-month moving average, which is at 4.29%, and we're currently up here. Here's the labor force participation rate. As more people leave the labor force, the rate should be coming down. That's it.

We'll see what happens. But again, we've got the rise of the million-dollar companies happening now. So, I think the future of the job situation will remain weaker because AI agents are coming. It's not that they're going to collapse, but I think people are making a huge mistake by not understanding that these little tiny companies that have no employees will be replacing bigger companies with maybe 50 employees or 100 employees.

When that company closes, those people have to recycle in. I just think the quits rate and all of that stuff are showing what's happening much better than what your economists are telling you. At some point here, I do think we're going to see weakness, either in inflation or inside the job market, driven by AI, which surprises people.

So, here we are. Bitcoin is still in a bear market until we get above there. Anything, from my perspective, is a trading side. So, I think this is going to hold, and I think we're going to break out, but as of right now, we haven't.

But there is some good news. I want you to go read this because I do believe that this is the most important time because of the AI agenda kicking in. But what I really like is, as the AI agenda and my Crypto 40 NAI Index, the ecosystem is doing well.

We just saw the Clarity Act basically, over the course of the last 2 weeks, go from about 33% to 40%. Back here, let's say 3 weeks ago, we were getting close to 50%. We collapsed to a 17% chance that it gets signed into law this year.

We also had Strategy—Michael Saylor—sell more. And we also had a Coldcard exploit. If you guys didn't see this, there was even a pretty big story on the cold-storage problem with 1 company and people's wallets. Yet, with all that going on, Bitcoin was okay this week.

That's it for me this week, guys. Keep enjoying your summer. Reach out to Mark Whalen, reach out to me, and we'll keep going through it. Let's make some money starting now for the rest of the year. See you.