AI泡沫要破？比特币归零？对冲基金遭重创？听起来市场底部不远了

This week, I think we have a record number of slides, so I’m going to move fast. For subscribers, I’ll put all the slides on the site, so you can look through them if I go too fast. There’s a lot of history. I think we’re at an important point because we’ve seen one of the largest deleveraging events of my career. Believe it or not, that’s how I see it, especially relative to the market. Let’s go through what’s happening.

Here is the Goldman Sachs VIP basket, the crowded long side relative to the S&P 500, shown as a 75-day rate of change. I chose 75 days because it captures the largest drawdown, and it also coincides with the record performance in the first half of the year. This is really about the period since June 30. Hedge funds had a huge head start to the year, which matters because deleveraging is very different when funds are down than when they’re playing with house money and trying to protect it. I believe that’s where we are right now.

Based on everything I’m going to show, I think the deleveraging is either done or in the seventh or eighth inning. That’s a big statement, considering we closed at the lows of the week. But this has been about hedge fund deleveraging, and, as I’ve said repeatedly, AI is inevitable. What’s going on in the market doesn’t change that; this will continue. This is not a bubble. Subscribers saw that this week in the videos I did. I also did a video explaining why the word “bubble” is distracting and doesn’t mean much unless you’re talking about going through a recession and being in one for a long period. I don’t think that will happen. I also covered how important NVIDIA has been as a signal.

1. Hedge Fund Deleveraging

First, we’ve had an intense sell-off. Over 75 days, it’s been as bad as anything this century other than the global financial crisis. Hedge fund leverage reached record levels after ChatGPT, and this was the peak. Now we’ve had a full-blown giveback. This isn’t hedge fund performance; it’s the VIP basket, and it assumes Goldman Sachs is updating it. It’s not a total performance measure, but it is a big retracement: about half the overall move since ChatGPT.

I also wrote a paper this week about time. You’re going to hear a lot about time; it’s going to become an even bigger theme. AI is going to destroy leverage because of time. You can no longer live on arbitraging when things should change. If your earnings beat estimates every quarter for the next 8 quarters, your stock can still go down. That’s the world we’re living in. If you depend on time arbitrage, meaning that things stay together, you have to understand that multiples can expand and compress. They’re just doing it much faster in this environment. I talk about the hidden leverage of the AI cycle in the paper.

The world is changing, and I’m trying to teach people, give them confidence, and empower them to use AI. Anthony Pompliano gave a talk this week, as he usually does. I talked about how parents can help their children adopt AI, given how fast things are changing. I’m going to go through videos from some of the best thinkers in the world on AI and what they think about it. I’ve always enjoyed the markets. Now I get to spend more of my time looking at them, looking for signals, doing podcasts, and talking about my views. Call Mark Whaling if you want to learn more. I’m going to go through videos from some of the best thinkers in the world on AI so you can stay on top of their views. I also talked about how parents can help their children adopt AI, and how you can help family and friends understand it.

The first 10% of this presentation covers what the week before last was about, the week ending October 16. Jefferies’ weekly net flows were lower than during the DeepSeek panic, another way to see the unwind in net leverage. Goldman Sachs reported the second-largest net selling of global AI stocks in the past year. Net buying since 2025 had been fully reversed. That’s why I’m getting much more comfortable with where we are. Overall hedge fund net positioning, according to Goldman Sachs, has dropped to levels we haven’t seen outside of Liberation Day in the past 5 years. Put volume exploded on October 16.

During that week, the deleveraging was deep, and it wasn’t limited to long-short equity, which is what most of the data I’ve shown so far covers. Goldman Sachs broke it down by stock pickers, quants, and giant firms, meaning multistrategy firms. Those 3 strategies lost more than 80 basis points in a single day. Since the COVID crash, that has happened only 3 times, and they were all this year: February 4, July 28, around the situational-awareness lows, and October 15. We had another one this week, and it was actually bigger. I’ll highlight that.

Global equity selling hit minus 3 sigma, driven by long liquidation, according to Morgan Stanley. UBS noted that it mirrored July’s situational awareness, particularly in that it was mainly outside the concentrated positions. Goldman Sachs told clients that stock pickers lost 2.7% for the week. Asia-focused funds lost 6.8% in the first part of the week. Global hedge funds were hammered by the AI sell-off, with Asia leading the losses.

That Goldman Sachs data is on a weekly basis. It goes back to what I showed on a 75-day basis: two weeks ago, the week was down 12%. The week ending October 16 was bad, and this was another bad week. As we moved into this week, here was one of the days: the horrible October 22, a complete bloodletting for hedge funds. On Thursday, almost everything was green. It shows that concentration is what’s driving this.

This week started with pure panic. Monday was going to be Black Monday. Over the weekend, tokenization finally started reaching the front pages. Western Digital and these other names were hammered violently overnight Sunday. Some closed below those levels, some above, but it was a leading indicator. More importantly, you don’t get to sleep on the weekends anymore.

I’m going to keep saying that the AI bubble didn’t pop; hedge fund leverage did. You can’t run a levered portfolio where you have to make money within a month or a quarter, or stay within a drawdown limit. That won’t work in a world where AI agents are involved. I’ve also shown the destruction that will happen: the hacks, the bad news, the geopolitical risk. Every kind of thing will keep popping up. That’s how this year is going to go. We are never going back to a time when, in a world of AI, the market moves and isn’t worried for 6 months.

That’s why I’ve given presentations on the time mismatch. You have to think this way when exponential innovation is moving this fast. I’m eventually going to go through the severe hacks happening across the globe. I’m also going to talk about Chamath and his “Bitcoin is going to zero” silliness. The reason it’s silly is that all the banks are being hacked right now.

2. The Quant Quake Returns

On the equity side, this reminds me of August 2007. I went through LTCM last week. In many ways, the AI trade is another version of what happened then.

Now, the quant unwind had no concentration risk. This is a concentration risk, but I want to bring you through it. For those who are wondering whether this is the end or the beginning of a bubble, I want to go through this and highlight what happened. I’ll show you what’s going to happen based on August 2007.

This paper, which is on AQR’s site, was written for the 10-year anniversary of the quant crisis, or the quant quake. It basically goes through 2 things. First, it was a factor unwind: large losses, and then, once everything was taken off, it almost made a full recovery. Unlike the situational awareness episode, where we never really got a big bounce coming out of it, I do think that, given the amount of deleveraging I’ve seen and what I’ve read from the prime brokers—I didn’t include the JPMorgan one—we’ve had a complete positioning reset.

A complete positioning reset. Are some positions still on? Sure. But at some point, hedge funds aren’t the entire volume. They get to a point where they’ve taken off the risk. Then we’ve got November and December, where once you get to a new month, you get a new clock. I hate to say that, but I was in the hedge fund world for a long time. There’s a big difference between having a month where you’re up 5% for the year but down 5% for the month, and then suddenly being up 5% for the year but flat for the month. You start trading a little differently, even though you just had a 5% drawdown, assuming you didn’t get fired.

Again, this reminds me of that. You guys should read more about it, but this was from MSCI, the same thing. This was a very big one, but they called it the “quant liquidity crunch,” and that’s what I believe is happening right now. Because of time and the number of hedge funds that can’t handle these types of drawdowns when they’re levered, volatility in the fund is explosive. That means the prime brokers are questioning, asking, and constantly there. The clients are asking, too. You end up in a situation where you’re reducing risk.

Again, liquidity is the problem because everyone comes to the market. When you come to the market at the same time to get out of positions, and it’s long-short, trust me: when you have to get out of part of the market, the AI trade, there’s no problem with liquidity. But when you have to buy homebuilders back, consumer staples, and all this other stuff, that’s why you get that screen I showed you that had so much green on it.

The 2007 quant quake was a liquidity time mismatch. Again, it wasn’t simply a shortage of buyers; it was a collision between years of accumulated, crowded alpha exposure and the sudden need to liquidate in days. What I remember about this is the Sharpe ratio leading into it—the size of the gains, and how long it took to build the positions—when it tried to unwind all at once. If you take 18 months, and every hedge fund gradually, slowly gets into that, momentum drives it. It’s a self-reinforcing play that everyone pushes up, like the AI trade. But then, if you try to get out of it within 75 days, everyone tries to get out. There isn’t enough liquidity to handle that. It took 18 months to build it up and 75 days to take it down. That’s the example I’m using.

If you guys want to see this, here are the largest volume contributors. This is the top 50. What I wanted to go through is the top 50 versus the bottom 50 performers for October 22. This is the dollar volume of the top 5. On this day, October 22, these daily returns are for the biggest volume among the top 50. This is their return. The largest volume contributors in the bottom 50, and the total sum of both.

Basically, to get out of your longs, which drove this performance down, there was plenty of liquidity: $187 billion traded that day. The 50 didn’t have the same liquidity. When you do this over time, you don’t have as big an impact if you’re putting this trade on over the course of 18 months. That is the problem. And that is why today is the widest day for S&P 500 breadth versus price since 1990. It’s the second widest since 1990. DeepSeek was the only other day that saw that breadth. There were 324 stocks up on Thursday, with the S&P down 1.36%.

I just want to show you the performance for these names that day. These are the 20 best performers in the S&P. This is why, when you look—Moderna was down big. It’s up 553% through September. Micron was up 2… When you go through these numbers and start realizing what you’re dealing with, you’re dealing with names that have made a lot of money for people, and now they’re getting out. Again, this makes sense. But there are a lot of reasons these names are up, based on their earnings growth at this point.

“Hedge funds pile into shorts as Goldman flags stock squeeze risk.” Hedge funds have loaded up on stock shorts this month, pushing bearish bets near a 5-year high. The reason this is happening with stock shorts is that there’s no way to use the S&P 500 as a hedge. So everyone starts creating baskets or doing something with stocks, because that’s the only way to hedge the AI risk. They should have used my index, which Morgan Stanley can trade. You can trade options on it because it covers pretty much everything.

The problem is that when index hedges like the Nasdaq and the S&P don’t work, and they don’t actually hedge what you want, you’re left having to use shorts. What happens is gross exposure actually goes up because you’re trying to stop the pain from those names going down. Instead of selling them, you’re trying to find a hedge.

Again, these are all for Thursday, because remember, Monday, Tuesday, and Wednesday, the market was up. Today’s sell-off deepened into the close. More than two-thirds of hedge funds finished down on the day. By this point, hedge funds’ net exposure is the least, so the S&P was down, but two-thirds were down. Global fundamental long-short funds were down over 1%. You can see where we are. Tech exposure and squeeze in crowded shorts are now down 3.8%, still up 8.8%. This is critical, guys. They are defending the month, but for the year they still have a chip stack. This is very important. This is very different from other years, particularly 2022 and other periods when you start getting these.

As you go through it, global systematic long-short managers were down 1.3% on the day, their third-worst day since 2023, but still up 19% year to date. Multi-strats were down as well. On this day, all lost more than 1% in a single day. I showed you that this hadn’t happened before. It’s the first time that all 3 were down at least 1% in a day.

“Hedge funds exiting trades fueled by…” This is the importance of this spreading into multi-strats, and this will widen. At some point, if it’s just in long-short equity, which was really what had happened in July with situational awareness, then you start spreading into credit. Obviously, we saw that with convertibles and CoreWeave last week, which I showed, but it’s really been spreading into French bonds.

That story was from the second. This story is from this week: significant losses on leveraged fixed-income strategies, basically carry trades involving French bonds. If you didn’t get to see it, these kinds of charts are never good for relative-value strategies, especially when you’re trading at a level like this. This is the spread differential between French 10-year bond yields and German yields. The spread is now 173 basis points, so French bond yields were 173 basis points higher than German bond yields as of Thursday’s close. We were at the top end of the range, except during the EU crisis back then. That created a problem.

On a weekly basis through then, it was the worst week in the last 16 years of data. The other one was the end of September into October 2. So you’ve got this unwinding. This led to European banks breaking below the 200-day moving average, which hasn’t happened much since COVID or since ChatGPT launched. There have been a few days in between here, so we’ll see if that’s enough. If this spreads further into the banks, then you’re going to have more waves of unwinds coming from different places. European banks relative to Euro Stoxx had their worst week since SVB, Silicon Valley Bank. You can see how the contagion is spreading into things in terms of the unwinds.

This was the S&P. I should have updated this. This was the percentage for the month through Thursday’s close, which was up 1.38%. Actually, that was through Wednesday. We’ll get through the rest of that as we go through it, and I’ll give you the new numbers.

To start the week, Microsoft pulls out of data center. IPO disaster exposed. Again, this was a computing specialist; a lease came through with its IPO, and it broke down. Firmus collapsed. The IPO collapsed. This is in Australia. Again, this is a neocloud. Then you had all of these conversations in the FT. Coming into the week, between the massive move we saw on Sunday in the stock tokenization side and all of the bad news, overall, except for Thursday, it was an okay week. Much, much better than last week, which again says to me that we’re through most of the deleveraging, even though Thursday was a brutal day.

A lot of that had to do with what was going on in France. One clear signal for me, similar to the situational awareness on Thursday, came Wednesday night, when we got Meta’s reports. The stock was up initially, then down, but there wasn’t a lot of movement. When I came back and looked at the S&P futures at 7:30, they were down 30 handles. That told me a hedge fund somewhere in the world needed to get out of things. This is the thing with hedge funds: it’s a deleveraging scenario.

I want to separate this out, because I know a lot of you are not hedge fund aficionados. Hedge fund deleveraging can happen, but economic deleveraging is what matters for the market. For the AI trade, the only thing that matters is economic deleveraging. Hedge funds, as I’ve always said—and because I was part of one—are just leeches on trends. They jump on the trends, the earnings trends and estimate revisions, and ride them. Historically, they’ve had a huge advantage because they’ve had the data and the ability to trade quicker. That’s all gone away, and now they’re competing with AI agents.

Whether or not you’ve adapted to that, hedge funds have a massive disadvantage compared to what they had. The bigger places still have a tremendous advantage in risk management, but that’s why we’ve seen a consolidation of funds. I was at a multistrat that was forced to close down. We were trying to compete with them, but we didn’t have the budget, among other reasons, to keep the talent and keep up. At this point, hedge fund deleveraging has more to do with what’s happening with AI, and it’s changing everything.

I’m going to take you through where economic deleveraging fits in. What I want you to see is a typical hedge fund deleveraging: earnings are still good, profit margins are still good, hiring is still good, and asset prices have a fast decline. That’s what we’ve seen in the AI trade. When you get into economic deleveraging, earnings start going down. Profit margins lead, hiring comes next, and asset prices usually go down after profit margins. Profit margins tend to lead, and I’m going to show that.

Here’s 2007. For the quant quake I showed, I got a recession signal in September of ’07. My model involved 5 components. I’m going to take you through at least 4 of them, and I’ll talk about the other one. One is jobless claims. Clearly, we haven’t had any rise in jobless claims. We didn’t have any rise in jobless claims back then either. Jobless claims are stronger right now than they were a year ago. Back then, they were starting to move higher.

This is the 4-week jobless claims. What I was using before was continuing claims. Continuing claims are different from weekly claims because after you file a weekly claim, you move on to continuing claims if you’re still not getting a job. Continuing claims were giving me a recession signal. Right now, we’re not getting any recession signal in claims.

Junk spreads actually bottomed and made their lows before the quant quake. The quant quake started here, and this is when it really got bad, so it actually coincided. If you look at spreads now, junk spreads look a lot like this. We made all-time tights, but we’re starting to widen. I could say right now that junk spreads are starting to show the risk of an economic deleveraging. The S&P is not.

The S&P was one of the signals. Year over year is what I cared about. We still have a massive positive year over year. A year ago today, we were here; we’re now up here. That’s nowhere near a signal, but it can change, obviously, with a big fall like we had in 1987. Then you change everything. As of right now, nothing there.

I wanted to show you that on this particular chart, we closed above the 20-week moving average. Since COVID, you can see we’ve very seldom gone below the 20-week, which is about the same as the 100-day moving average. I wanted to show just how strong this has been and what’s been happening since there are no more recessions. Recessions would happen, and one way to measure it was to look at how many times, or what percentage of the time, we’d be structurally in a bull market.

This shows the percentage of weeks during these time periods that we closed above the 20-week moving average. Since ChatGPT came out, since profit margins basically started to go parabolic, we’ve been above it 85% of the weeks, far higher than any period by far. Over the course of the entire period, we’re closer to 72% to 73%. In these periods, it’s been getting better as exponential innovation has impacted our lives. We had the dot-com bubble, then the Great Financial Crisis as part of this. Since then, it’s been getting better, but it’s really had a step-up function here. This is the best leading indicator I can give you for when a recession is approaching or when an economic deleveraging, like an AI bubble, is showing up.

Year-over-year corporate profits, guys, plain and simple. If you have profit margins growing, it’s like what I said about hedge funds: you have a huge window. Corporate profits are up 20% right now. Before the recession, this was one of the sell signals. This was one of my recession signals. It happened in March and June of ’07. Profit margins and corporate profits started to go down. You get profit margins going down, and then you get the recession. This bar here is when the recession started.

The dot-com bubble started in March 2000, and profits went down. In 1990, profits went negative before the recession. They went negative before the recession in 1980, too. That was a weird period because you had Volcker fighting inflation, but profits still went negative. This one didn’t break down until later. The main point is that profits are still going straight up, guys. We have nothing here.

3. AI Demand Keeps Growing

On the earnings front, SpaceX reported probably the most important earnings report I’ve seen in the last—I want to say, except for NVIDIA. This just confirmed it. If you didn’t get a chance to read the quotes and go through it, it is unbelievable what Elon said. Compute demand is substantially greater than supply. “We could rent out our compute several times over if we said yes to everyone.” This is critical because he’s also talking about being able to build an incremental 10 gigawatts. The revenue on that is somewhere between $30 billion and $50 billion. He’s saying that in year 1, based on what he’s already gotten from Anthropic and what they’re willing to pay now.

Let’s assume it’s lower than that. It’s $20 billion. You’re still talking about $200 billion of revenue if he’s able to do it. “I’m fairly confident of being able to bring an incremental 10 gigawatts of compute online in 2027.” He very seldom lives up to what he says in this. Let’s assume it’s 5 gigawatts. You’re still talking about $100 billion. More importantly, he is literally CoreWeave on steroids in terms of seeing where the demand side is. “I think in 3 to 5 years, there’s probably a demand for 1,000 times more,” just measured in bandwidth, “than there is currently.” Again, he’s talking about the agents and how fast they’re going. Capital expenditure is expanding, not slowing.

JPMorgan doubles its SpaceX price target. The Street goes from $240 to $500 and now forecasts a quarter of a trillion dollars in 2027 revenue. I mean, guys. JPMorgan also put out a memory report basically questioning how this is even happening. Again, I’m going to say this over and over: regardless of where Micron is today, if we didn’t learn anything from the first quarter—and you guys should remember this—things move fast. There is no doubt in my mind that by the time we get to the end of February, Micron could be at $1,500. Absolutely no doubt in my mind.

For everyone who’s bearish and now beating their chest because the semis are having what is a normal correction for them, continuing from what happened back in June, they’re still up more than 30% year to date. Be very careful with the amount of deleveraging that’s happening and how little has actually gone on.

Cyber, Super Micro Computer, and Texas Instruments preannounced. I think this is all well and good, but the main point is that humanoid robots are being brought into the equation. Humanoid robots, drug discovery next year, health care—all of these things are going to be part of next year’s bull market. Again, it’s a bull market with many, many corrections. Jon Gray is a huge part of AI. When you think about a bubble, you think about building capacity and then hoping demand comes. With AI, you’re building capacity and hoping you can keep up with demand. “We can’t keep up with customer demand. Demand is almost parabolic.” I completely agree, and this is seen around the globe. This is not just a U.S. thing.

The interesting thing is to show the hedge fund unwinds. Blackstone, which was trading near 52-week lows even though they’re part of the AI trade, was up 8% on Thursday, which just shows—they reported earnings. They opened down significantly on the day. I want to say they were trading down 4% premarket, then finished up 8% on the day, their second-biggest day in the last 3 years. Here so far, out of my 100 names in the portfolio—and the reason it’s important to pay attention to my portfolio—here is what they’ve said in the earnings quarter so far over the last 2 months.

During the deleveraging, you guys can go through all of it. I’m waiting to see a company that says, “You know what? Demand’s slowing. We actually think it’s going down.” Instead, the numbers are massive. Even with companies preannouncing, they’re taking their numbers up significantly. When you go through every single number, they’re staggering, and the guidance is still staggering. The chips are still staggering.

I wanted to show this because, again, you hear about how GPU prices are supposed to be here. This is where they are based on the years. This is each year, with the residual value right now. These are supposed to be worthless already. They’re supposed to be down here, but they’re up here. You can find anything.

These are the same guys that put out the token data, which is still so funny because they tell people it’s not what they think, but people still go over it. Andreas put this out as a leading indicator of what happens. Exports from Taiwan and Korea, which are a major part of everything related to technology, are still going higher in the latest data.

The earnings for the S&P, for MSCI U.S. forward EPS. Duet was released, so we’re getting another agent. You’ve got Meta Muse, you’ve got Dots, you’ve got Grokbot, you’ve got Instinct. They’re all coming, and Jensen highlights that personalized agents will push token usage through the roof.

For the week, the S&P was down 47 basis points. With all of the panic and all of the news about deleveraging, we’ve had the last 2 weeks of insane amounts. The July period looks extremely similar. Here it is for the month: down 36 basis points, very similar to July. Again, not much has gone on.

The S&P is up since hedge funds peaked. That is part of the economic deleveraging. There isn’t anything happening. The Qs were down only 76 basis points for the week and down 4.78% for the month. Again, that’s not outside of the norm. They had a 16% decline. You probably don’t even remember this, because when it’s going on, the media makes it sound like the world is ending.

IWM, small caps, were up 2% this week, part of the unwind and the deleveraging. My thematic portfolio was down 3.8%. Here is the SMH drawdown from the peak, guys. I want to reemphasize: we’re down 22% from the peak. That’s how much of a bear market it is. And guess what? This happens almost every single year, especially when you’re up as much as they are year to date.

This is where my portfolio finished up: 20% of the names. We’re at the same point we were here, which was right around the bounce. I think we’re in this part of the cycle, unless this turns into earnings starting to miss, which I don’t think is going to happen. On the 50-day side, we’re much more oversold, which is important because that means we’ve gotten more people out. On the 200-day, absolutely: we’re down to 38%. We were still up in the 70s.

What has not changed is the 200-day slope: 85% of the names are still pointed upward. We finally saw a little bit of volatility. We’re nowhere near the peak here, but we’re still higher. Bitcoin was up 50 basis points for the week. For those of you looking for an asset that just continues to ignore what’s going on, which is obviously part of the AI trade from the agentic side, it’s still hanging in.

4. AI Credit Risk Comes Into View

The first thing to listen to is the All-In Podcast. They did a great job this week going through the CoreWeave crash. I’m not going to spend a lot of time on this one, but I think you guys should listen to it. They did a good job of explaining the difference, especially Ben Horowitz, who kind of started the neocloud business back in the late ’90s. He talked about that, so I think he’s a very important person to listen to—far more important than the perma-bears on X that you may listen to and get involved with.

He talks about how they don’t invest in it, but it’s not because he doesn’t believe in their business. He just realized the mistake they made getting into the business, which is basically a real estate business. I showed in the subscriber video this week: go look at the commercial real estate market and how it’s paid off for people who got bearish at 70 cents on the dollar. They go through the different ways of looking at it. I think David Sacks and Ben Horowitz are the people to listen to on this.

They weren’t arguing that CoreWeave has no financial risk. They were arguing that credit market stress does not automatically translate into corporate bankruptcy. I think it was very important to understand what to look at. I would go listen to it. I’m not going to spend a lot of time on it for the subscribers. You guys can get the slides, and I’ll also give you the details from each of these. You can either read the summary or go to it and download it yourself. That’s the whole point of giving you guys this stuff.

David Sacks—this is the one thing I will read you guys: “Whenever you’re underwriting to multitrillion-dollar outcomes in the future based on very rapid rates of growth, exponential rates of growth, the market’s going to be super sensitive to any data points we have in the present.” This is why the AI bubble is going to be here. I’ve said this before, and I said it in the subscriber video: we’re barely into the credit side. Barely.

I hate to say that to people because they seem to believe we’re at the massive stage of it. When you see $10 trillion of capital needed and then start associating that with “we’ve taken out $10 trillion,” no. We’ve barely started this, because the reason free cash flow is negative is that they used the free cash flow. Google raised equity. So, yes, are they coming to the market for debt? Yes. But we’re in the very early stages.

Is capital becoming challenging? There’s no doubt that when you come to the market in a short amount of time for massive amounts of capital, you need to build the risk premiums in, just like with liquidity for hedge funds. Plain and simple. That’s effectively what it is.

5. AI Reaches Healthcare

Eric Schmidt—not the best-looking photo, but I like Eric. He was interviewed on Norges Bank with Nicolai, and it was good. Eric is a good person to go listen to. I think he did a good job. AI is bigger than every previous technology revolution. He rejects the bubble narrative and goes through the reasons why. He talks about all the different things, but he gets into biology, and I completely agree.

If you haven’t thought about this, with Eli Lilly making new all-time highs and me writing about it and talking about it in this video, I think you need to go through it. He talks about the fact that you don’t want to be viewed in history as not participating and not being a major leader in something that will be looked at for thousands of years. It is a brilliant statement. This gets into whether you agree with how these guys are spending their money. The mega-hyperscalers have the money to spend. They are not going to finish in second or third place and not be part of history. It is not going to happen. That is the trend.

So everyone looking for, “Oh, they’re going to cancel their capex spending”—you’ve got to go listen to Eric Schmidt. It’s just something that’s not going to happen. “The Rise of Intelligence Is the Biggest Revolution in History: Why AI Is Not a Bubble.” There’s real growth there. It’s not fake. It’s real money. He’s talking specifically about Anthropic. Anthropic plus OpenAI are going from about $28 billion or $29 billion a year ago to $175 billion, no matter how you get there. Maybe it’s $165 billion. You’re talking about massive dollars, as I’ve shown you.

For people going through the nuance of trying to say, “Well, this is $150 billion, this is $170 billion, it’s slowing down,” you’re wasting my time and everyone else’s time with your ridiculous views on something that a year ago you said wouldn’t even be $50 billion at this point. So now, when it’s not $175 billion, we’re going to look and say, “Well, look at this. It’s slowed down,” when every quarter a company reports that doesn’t even make nearly as much money as they do. Of course, there are going to be quarters that go down or don’t grow as fast. But that doesn’t mean it’s stopping.

You’re talking about the fastest-growing companies in history. More importantly, Meta’s taking some of that. Grok’s taking some of that. Gemini’s taking some of that. All of them are taking it, and they all count, even if those guys aren’t getting all of it. Business will be flatter, smarter, and more intense. You have to listen to this. The faster the learning loops, the smarter it gets. I have friends who think that in the next 15 years we can cure all disease. Every single thing in here is important to listen to.

I just want to bring up that specialized superintelligence is coming. Many people believe that in the next few years, we’re going to build superintelligence in key categories. This aligns with Eli Lilly’s TuneLab, which I’ll get into. I want to make sure you guys remember the TuneLab thing.

The Big Technology Podcast—Alex Kantrowitz always does a great job—had Sundar Pichai on. I’m not going to take you through the whole thing. He basically says the frontier models will be needed. They will always be needed, and they will always be the best models, but they won’t be for everything. He explains in a very intelligent way how much of the work will be done by open source. This is important, but you will always need the frontier models.

As he explains, there are going to be lanes for them. We’re starting to get into a world where some of them are open and some of them are closed. You should go listen to Sundar to understand why. Technology will adapt to humans rather than the reverse. He goes through a lot of things that I think are important, but he also says he sees major gains ahead in health care, materials discovery, and bending cost curves.

There’s Eli Lilly. It was down on Friday. I did this through Thursday, but it made new all-time highs this week. As a reminder, I wrote this paper in June, where I said it was becoming the most important AI company in the world: Eli Lilly TuneLab, the Manhattan Project for AI-driven discovery. It’s one model trained on data from 135 biotech companies: their wins, their failures, all of their work, anonymously. This is how you get major breakthroughs and collaboration. This is how the specialized world will accelerate things.

It’s not simply an app to help with discovery. It’s building a specialized bio model. All of this is going to lead to massive changes. I went through the size of health care and its importance from a barbell perspective in the subscriber video. What’s happening in pharma, life sciences, and biotech is beginning a massive bull market. We had one very similar to it in 2000 during the dot-com bubble. We had one during the Great Financial Crisis. Health care does better during periods of lower volatility.

But remember, health care is now the biggest spending category in this country. It’s going to rerate as more of a growth area, and you’ve started to see that. You saw the move with Moderna. Forget how much it fell and that it closed around 200. Look at the move we’ve had over the course of this and how far it’s gone. You’re going to see these moonshot developments all the time. If you need something to compare TuneLab to, Harvey Legal AI is very similar, just different ways of going through it. This is the whole importance of it. Harvey raised half a trillion dollars at a $15.5 billion valuation, meaning TuneLab has an enormous amount of value.

6. Cybersecurity Pushes Crypto Forward

On the negative side, and something to keep track of—particularly when I get into Chamath’s garbage about Bitcoin going to zero because quantum will destroy it—AI agents are in your bank accounts, guys. If you didn’t see the story this week, they’re in there. Seventy percent of websites that use PHP had a break-in. I heard about Reverse Engineer Anything on an a16z podcast today, and I’ve seen it on X a bunch of times. This is another reason to realize that nothing in the world is safe from competition.

If you think hacking is something to worry about, quantum hacking in six, seven, maybe 20 years, as Chamath said, look at what’s happening now. The hacking of companies through reverse engineering, and the hacking of banks and websites, is all happening right in front of us with agent swarms. That’s today’s business because it’s all based on software. This is re-engineering hardware. Very scary.

I did my crypto video this week and focused a lot on getting from here to here. This is happening, guys. We are in the tokenization world. You have to start getting ready for it. Tokenization is absolutely going to disrupt hedge funds, just like what we’re seeing so far. Deleveraging is over, which means the concentration and risk around AI, in my opinion, have reached a more neutral stance.

For people looking for an entry point, the only way I see this being a bad entry point over the next three months is if we go through an economic recession. I don’t think that’s going to happen. We’ve cleansed the system with the losses and forced liquidations. That doesn’t mean we’re going to start rallying ahead of the midterms or ahead of whatever else you want to think about. But that’s why I say that, at this point, from a swing basis, you have to start thinking differently.

I covered an important part of tokenization in this week’s video. Here is the 126-day volatility of the AI thematic portfolio, now trading close to 40. Here’s Bitcoin, that volatile asset people are worried about, trading at 30 over the same period. The tokenized index itself is down 10% for the month, but still up 50% since August. This is the one I want you to realize: year to date, it’s up 35%. It’s now up more than the AI trade. That’s right. Bitcoin is down 7%. The index is there.

I’m working very hard on this. I have another meeting next week about getting it to be tradable. I’m going to put money into it. I’m trying to use a group that’s going to help me with it. Here’s where we stand. Again, this is bull market stuff down here. It’s oversold on a short-term basis. Here’s the chart of it. This is the tokenized index, of which Bitcoin is one of the 46 names. You’d think that crypto—the 46 names, 42 tokens for public companies—would outperform the AI trade now.

a16z is very important. For those of you who don’t believe in crypto, this is what a16z did. They announced they’re now setting up a fund with Chris Dixon, who has always been their crypto guy and, I believe, the third partner behind Ben Horowitz and Marc Andreessen, one of the original ones. They’re now calling it a16z Frontier.

“When we started investing in crypto, it was a world of its own. But as blockchains have become more widely used, the lines have blurred, and the infrastructure has gradually receded into the background. Wall Street and fintech companies now regularly launch products built on blockchains. Does that make them crypto companies? The label matters less now. It makes more sense to define companies by what they do rather than by the technologies they use.”

If you still don’t believe in looking at crypto, I’ll continue to say it: you’ll be doing it nine months from now, because this is the beginning of a large, sustainable bull market in tokenization, agentic commerce, and stablecoins. Do your homework. Start with this: Frontier Ideas, with Dixon, Ben Horowitz, and Marc Andreessen all in one podcast talking about AI and crypto. I was the one who brought AI and crypto to you guys beginning two years ago. Here we are now, and it’s merging for everyone.

“The lines have blurred, which we find ourselves asking. I’ve been writing about this for a while. Is Stripe a crypto company? Whereas years ago it was sort of in its own world, with its own companies and investors, the lines have blurred, and it’s become more part of the tech landscape. Crypto is no longer a separate industry. It is becoming infrastructure.”

I’ve talked about the crypto guardrails. I’ve talked about the ghost rails. We’re now getting what it was made for, which was the agentic side. For people who don’t agree or don’t believe AI is connected directly to crypto, and that eventually people will realize it and come to the same deep philosophical viewpoint I’ve come to: Bitcoin is the purest AI trade, once you get into what Chamath is wrong about, in my opinion.

Where we differ is that, to me, people should be worried not about quantum hacking Bitcoin, but about AI hacking capitalism. That’s what’s happening, and that means all companies—all incumbents—are being disrupted. Marc Andreessen talks about that extensively in this interview. AI is compressing decades of technological development into years, fundamentally changing what small teams can accomplish. This is my tokenized world: tokenized small businesses competing with incumbents. At the same time, crypto is becoming part of mainstream financial infrastructure, and autonomous AI agents could reshape the internet’s economic model. This is it, guys.

All I’m going to say is this is stupid. It’s not to say that he is stupid. It’s to say that the statement, whether it was meant to get a lot of clicks or for whatever reason, or whether he has some PTSD from getting out of it too early, has no basis. For people who reached out saying Chamath is whatever he is: does he get lots of listeners and tons of followers? Sure. Adam Back has a PhD in computer science. He invented Hashcash. He is cited in the original Bitcoin white paper and has 30 years of experience in cryptography and computer security.

Chamath has no established professional specialization in quantum cryptography or Bitcoin protocol development. He’s a podcaster who worked at Facebook and has a degree in electrical engineering. No offense to Chamath, but when you make a statement and I disagree, I’ll tell you who I listen to on this. I would never listen to a podcast where he talked about Bitcoin. I listened to Adam Back before I put money into this. So I’m talking my book. Obviously, I believe in it because I talk about it. But if you’re worried, follow Adam Back.

The quantum threat is potentially 20 to 40 years away. This is what he’s saying. Existing quantum computers cannot break Bitcoin. Bitcoin can upgrade to quantum-resistant cryptography. He’s saying the Bitcoin upgrade process may be too slow, and that’s a fair statement. But to say it’s going to go to zero because the upgrade won’t happen—that’s not a fact, it’s a story. Prepare carefully and avoid rushing.

Here’s the most important thing: why Back believes Bitcoin can survive. Hardware versus theory. Quantum algorithms exist, but sufficiently powerful fault-tolerant machines do not. This is where you get into software versus hardware, and that’s the whole point he’s making. The quantum threat to Bitcoin is primarily a hardware problem. We’ve learned, if for no other reason, that hardware takes time. It’s based on physics.

Can we get there? Yes. The software is happening today. AI lowers the cost of cyberattacks, identity theft, and exploiting vulnerabilities. I’m going to say there will be a cybersecurity crisis. Where he says it’s going to zero, I’m saying this story becomes the most important story for the market over the next 3 years.

We’re going to have severe, massive cyberattacks that make front-page news—more than Hugging Face, more than the 2 stories I showed you from Axios this week. In my opinion, this will drive more people toward cryptography and force the banks to focus on getting everything onto the blockchain. If anything, I believe agent swarms and hacking in the software world are going to speed things up toward safety while we wait for a hardware issue.

Chris Perkins, someone I’ve gotten to know over the years and respect tremendously, wrote this paper: “Quantum Won’t Kill Crypto, but It Could Accelerate Wall Street’s Adoption of Blockchain.” It’s very similar, though different. He’s saying they’re going to move to the blockchain because of fears about quantum. I think it’s going to happen now.

Again, unvetted AI agents are accessing U.S. bank accounts. Go read the story. How many passwords do you guys have? If you use crypto and have one identity—an anonymous identity on the blockchain—it’s very difficult. All of a sudden, you don’t have to deal with all this garbage. I think everything should move to the blockchain, and then the banks, the governments, and everyone else who’s now getting involved in Bitcoin and advising their clients to do it.

So if Chamath is saying it’s going to zero, while Jamie Dimon and everyone else are saying, “We’re advising you to put money into Bitcoin,” and JPMorgan just said they think Bitcoin will be the best asset over the next decade, sorry, I’m not going to listen to Chamath.

Japan declares a cybersecurity emergency. That’s already going on. Again, PHP was hacked. Why did Spark scramble? Stocks are getting hammered. Tokenization: this is my tokenization moment in terms of the next stage for stocks. As a reminder, we have the quarterly refunding announcement. I mentioned it before. This could be a catalyst to really get crypto going again.

For those people looking to listen to someone on why it’s a bull market in crypto, Tom Lee does a good job on Token 2049 of going through a bunch of different things I’ve talked about. He doesn’t really talk about the agentic side, though he does talk about agentic commerce. He also talks about the generational wealth transfer. It’s definitely worth listening to, particularly when you get past “Bitcoin’s going to zero” and think, well, if it is going to zero, it’s not going to happen for another 6 years. By then, can we solve something? He’s saying the community won’t get around it. What if the community owns JPMorgan in the United States, Japan, South Korea, and everywhere else? I think they might have a chance to do it.

Tokenized AI stocks slid over the weekend ahead of the reopen. I’m just showing this because this is what’s going on. Switzerland approves its first Bitcoin ETF. JPMorgan predicts Bitcoin will outperform U.S. stocks, gold, and real estate over the next 10 years. BlackRock’s Bitcoin ETF, IBIT, pulled in $664 million. To get a sense of it, we’re at all-time highs in the outstanding shares of Bitcoin. This happened during a bear market. What do you think will happen to this line in a bull market?

As I talked about, the silent IPO—I wrote that in here. I believe we’re at a point where this was a buy-the-rumor, sell-the-news event. It should have been a sell-the-news event when Donald Trump became president. Instead, we’ve consolidated for 2 years. Once we get back above this level, which in Bitcoin terms is basically going to be $100,000, I think this line is going to go up a significant amount, particularly now that people are exhausted by the volatility in AI and you’re going to have hackings and everything else going on consistently.

JPMorgan now has 3 tokenized products live on Ethereum. Tokenized stocks are expanding across DeFi. I mean, this is just happening, guys. Banks and institutions are on-chain. The old guard isn’t fighting tokenization; it’s slowly but surely moving in. These are all the companies that are on-chain now.

Yuval Rooz—I spoke to him. Canton, very, very interesting guy, and he’s done an amazing job getting Canton to this point. You should go read about it again. The more I meet and talk to these guys, the more I see that they’re all incredibly smart and could easily be doing big things in other markets, but they believe in this. Canton is connected to DTCC, so this is a very important thing for tokenization. Speaking to him, I can tell you things are moving ahead. Tokenization is accelerating. There are now over 360 on-chain tokenized equity products, up 200% year to date. This is the kind of stuff you want to invest in, guys.

Stripe’s agentic commerce is revving up. John Collison said in an interview with Mark Zuckerberg and Alexander Wang that agentic commerce went vertical last week. Brian Huang, who I spoke to and met with last week and who’s with Glider, is introducing automated token portfolios. This is what I’m hoping we can do with mine. It’s a very interesting way to approach this. Very cool, as someone who’s been on Wall Street a long time.

That’s it for this week, guys. I tried to keep it close to an hour. I think I did that. I went fast. I did 3 videos this week before this one. This is the fourth video I’ve done this week. When times get tough, I stay in front of the computer and try to bring you things that at least calm you down, give you some of the facts, and help you organize things. At the same time, I think this is going to be the life we’re in for a long time. I’m trying to give you the places to focus: agentic compute, for sure. Healthcare needs to be part of your portfolio as well, on the pharma side, the biotech side, and life sciences. I showed a combination of a portfolio that I think people should be involved in. You can watch the video for the entire presentation; the slides are also up there. If nothing else, I’ll see you guys next week. We’ll see if there’s another tokenized wipeout over the weekend.