Comfy Sunday Spaces v37: OAI/Anthropic Revenues, Neocloud volatility and Trading Philosophy

It’s been a pretty exciting week in markets. We try to schedule Sunday Spaces around periods of volatility, or sometimes periods where there’s a lack of volatility and we think something interesting might happen. This was obviously a week with a tremendous amount of volatility, specifically in the AI trade.

I think let’s start with a lot of the headlines that came out in the last week, week and a half. OpenAI ARR was revised from the whisper number of $70 billion down to its self-claimed $50 billion. Then there were the leaked numbers from the Anthropic investor meeting, which I guess will eventually make it into its S-1 if it files one and actually tries to IPO. And then a lot of the inflation numbers came out: CPI and PPI were crazy hot.

There was a lot of data in the last week, and markets didn’t care at all until they cared a lot. Maybe let’s go into that.

It’s very clear that quite a bit of the market is reliant on 2 private companies: Anthropic and OpenAI. Whether people realize it or not, quite a bit of GDP growth is downstream of, or has a direct overlap with, some of the spending these 2 companies are doing. Specifically, that’s spending on data centers and revenue generated by other hyperscalers and companies around the world paying for the fantastic services these 2 companies provide.

It’s unclear, and there’s quite a bit of discussion around whether these businesses are profitable. Anthropic—which is… Well, hold on. Before we jump into it, it’s important to say that it’s slightly frustrating that so much of the U.S. economy is based on these 2 private companies. They’re not small; they’re some of the biggest companies in the entire world by market cap, not necessarily by revenue.

Anthropic is slated to go public at presumably a $1.5 trillion market cap, which would put it among the top 10 companies in the world, almost—just below Broadcom. So much of U.S. spending and growth, and pretty much the only industry that’s growing, has been AI and AI-related spending. It comprises so much of the index at this point, and it’s basically the trade that has dominated everything.

1. The Case for Going Public

Before we get into the actual figures and what we’re looking at, I think it’s important to say that what’s frustrating about trying to speculate in markets right now is that there’s a tremendous amount of asymmetric information. We have no idea what the actual revenues of OpenAI or Anthropic are. In fact, some investors in OpenAI and Anthropic don’t even have information rights, so they can’t get the companies’ revenues, net profitability, total spending, or even their balance sheets in real time.

You have protections as a private company in the United States, and that’s how the law works. But we can still comment on whether that’s fair. The beautiful thing about public markets is that there’s a standard of reporting. If you misrepresent your company’s financials, lie, or do something illegal, you go to jail. That still happens with private companies, obviously, but there’s quite a bit of leeway.

We’re not saying Anthropic or OpenAI are lying about their revenues. But it’s quite frustrating that people who have access to that information may be making very informed decisions in public markets. If someone knew that OpenAI or Anthropic’s revenues were slowing down, presumably that would affect their future spending, and they might short companies that are direct or indirect beneficiaries of the rapid, ravenous spending by these 2 massive frontier labs.

Again, this is less about markets, sentiment, or trading, and more about frustrations I’ve had personally. Kyle, I don’t know how you feel about it, but it does sort of blur the line. I know we say “crime is legal,” and joke about these things. I personally don’t joke about them. It’s quite different from public markets, where you’re not allowed to have asymmetric information on a company.

You can do research and spend quite a bit of time trying to become a very informed market participant. But it feels quite unfair at times that certain market participants have access to information about these 2 massive private companies, especially when we’re trying to navigate these fairly complex and sometimes theoretical markets. I just wanted to open with some of that. I know I had an interesting week on Twitter, as many of you have seen.

At a certain point, I think it’s important to consider this, especially if these companies are seeking outside funding. There’s a big difference if it’s a family business that’s been compounding over generations and generations. I suppose that can be grandfathered in and remain private. But a new frontier technology company being allowed to stay private and become one of the largest companies in the world, affecting so much of the market that everybody is exposed to, feels like something that might be looked at in hindsight.

People might say, “If a company has tens of billions of dollars in revenue and has raised hundreds of billions of dollars of capital, maybe it’s sort of a public good. Outside investors should have the opportunity to invest, and this company should be forced to go public.” There are forcing functions that made companies like Facebook go public back in 2009, or whenever it went public, based on the number of investors. People at Facebook essentially launched these SPVs to get around the rule because Facebook really didn’t want to go public. But they ended up being forced to go public by the SEC.

I’m a free markets guy, but I do think it’s important that the public is able to share in quite a bit of the upside and speculate if they choose to. Additionally, maybe public markets should be a bit more efficient, where reporting can happen once a year. That way, companies aren’t as short-term oriented. Companies could choose to report every quarter or once a year, with a little more flexibility, and still be liquid, tradable instruments. I think that would probably be a reasonable change.

Markets have changed considerably since these rules and regulations were created. Some of the SEC rules we follow were made almost a hundred years ago, because of speculation in the 1920s. Again, I just wanted to put that out there. It’s not really relevant to what we’re thinking about markets; it’s more about my philosophy on markets in general.

It’s quite frustrating that someone like Situational Awareness, or people with exposure to Anthropic or the ability to get that information, could have asymmetric information and make really informed decisions in markets. It does feel quite unfair. Anyways, Kyle, what are your thoughts?

I’m completely unbothered by any of that. I just don’t feel any need to—I don’t feel any reason why a company should be forced to go public. I think they have the autonomy to do what they want to do.

I do, however, think these companies are unique because, for better or for worse, they’ve distilled all of the public information on the web. It’s kind of weird how there’s a lot of uncopyrighted material. If you’ve ever contributed anything to the public domain, written an academic paper, made a blog post, or uploaded a YouTube video, you’re training data.

These companies have essentially taken contributions that random people made to the web or to the collective consciousness, distilled them, and used them for profit. You’re seeing a lot of that sentiment being addressed, I think, because Anthropic just promised—this news came out hours ago—to donate up to 20% of its shares to the Trump kids’ accounts, or something. What are they called? The Trump Accounts. Basically, they’re creating these vehicles to give the economic upside of their company—not the voting rights—to regular Americans.

2. AI’s Unaddressed Externalities

There’s something being done to address that, but it does feel like there are a lot of externalities. There’s a lot of fear that they’re conducting things that are getting quite dangerous. There were a lot of cryptography fears in the last week, where they’re doing frontier math that is bordering on being dangerous for people who enjoy the benefits of cryptography for communication.

There’s a lot of hacking cases coming out, too. A few more companies, in addition to Hugging Face, claim to have been victims of swarms of AI agents hacking them. The fastest-growing repo in GitHub history is, I think, called REA, “Reverse Engineer Anything.” The idea is that you can take a compiled binary, throw a decent amount of AI at it using this special harness, and really bust the thing up and start breaking programs like Photoshop and a lot of other stuff.

There are real people being harmed by this now. There are externalities that I don’t think are well addressed. Being a private company, especially in a regime where the world so badly needs you to work for economic reasons, puts you in this weird spot of being simultaneously a little bit dangerous but also very untouchable. That’s what I’m uncomfortable about.

But whether or not a large company has to… If Anthropic and OpenAI made furniture or something, and they became trillion-dollar companies just making furniture, I don’t think anybody would give a shit. They’d say, “Let them be. Those guys want to be private, let them be.” But because of the way their companies are set up—they borrowed from everybody and also directly impact everybody—it feels odd for them to be private.

Yeah, and that’s why I make the distinction around whether they’re raising outside capital. As a furniture store, you’d presumably be fairly cash-flow positive, and maybe you’d issue some debt or something. But a lot of the LPs in these VC firms are public pension funds, especially at this scale, so people do have some exposure to them.

Just based on raising outside capital and then reaching a certain size, which has downstream effects on quite a few public companies, maybe… I don’t know. But yeah, I agree. I think free markets ultimately win, but there needs to be reasonable regulation. Again, you could make the argument around healthcare and a bunch of this other stuff.

Kyle, I also want to give you a shout-out. Kyle did a podcast with Tykie, who’s a good friend of ours, a friend of the Spaces, a friend of the pod, a friend of the family, and you crushed it. It was one of the best podcasts I’ve ever listened to. Obviously, I might be a little biased.

Thank you.

But it was truly fantastic. Podcasts in person do make a big difference. A lot of the commentary was quite interesting. Obviously, people thought you were me, thought we were not two different people, which was kind of funny—flattering, maybe, that we have this split personality.

I highly recommend anybody listening to the Spaces go and listen to Kyle’s podcast with Tykie. He goes over quite a bit, and far more eloquently than I’ve ever been able to. He’s been able to—

Yeah, some good RamCap lore in there.

Yeah. A lot of good trades from Rollbit and Hype and a lot of this other stuff. But okay, let’s get into OpenAI and Anthropic revenues and why they matter.

Yeah.

These are two massive companies.

Yeah.

These are two companies that are driving a substantial amount of return in public markets. It came out recently—and again, these are not audited financials, and these are all rumors, so you have to verify on your own—that OpenAI was on track for $70 billion in ARR.

Yep.

ARR is different from revenue. It’s a bit of a confusing metric. It’s annualized run rate. What does that mean? You could take the last day of revenue and project an ARR. If you made $1 million that day, your ARR would be $365 million over the course of a year. If you take the last week, then multiply it by 52, and so on and so forth.

It’s not disingenuous, but it’s non-GAAP accounting. It’s kind of saying, “We make this amount of money, and if we continue to make money at this rate, this is what we would make for the rest of the year,” or something like that. Typically, it’s year or year to date.

It came out recently that Anthropic—or, excuse me, OpenAI—walked its numbers down from $70 billion to $50 billion. This happened, I believe, on Monday or Tuesday of last week, and the market didn’t really care. The market kind of shrugged it off, and the S&P closed kind of green on the day.

Then the market started selling off. This also coincided with Anthropic doing investor days last week, where they went because they’re trying to go public. We haven’t seen their S-1, the filing document that will tell us all about their financials before they go public, so investors can make an informed decision about whether they want to potentially purchase the IPO.

They went to their investor days, and numbers always leak. The numbers came out that in Q3 they made something like $17 billion—I think it was $16.8 billion. Anthropic, on the other hand, had been projecting—and this was based on maybe July or August figures, but certainly early-to-mid Q3—that they were on pace for $100 billion of ARR.

If you do the math and say, “Okay, well, $17 billion times three is about $50 billion a year in revenue,” that doesn’t really equate to $100 billion in ARR. Most likely, what they’re saying is that by Q4, by the end of the year, they will make something like $25 billion in the quarter, and then they’ll be able to project a $100 billion ARR. That seems fairly reasonable.

A 50% increase is exactly what they did quarter over quarter from Q2 to Q3, but that was during an era of token maxing. Since then, Anthropic tokens—and generally the cost of using some of these programs—have simultaneously gone up, in terms of the pricing wars for some tiers. Using the API and some of the heavy-use options has actually gone up in cost for the consumer, while the actual price per token has gone down.

There are two schools of thought here. The first is that these companies are the best businesses on Earth, and that’s the bull argument. The only reason they’re so cost-prohibitive and not massively profitable is that they have to do pre-training for their models. That’s a totally reasonable argument.

If you remove pre-training, inference is only 37% of actual data center usage at this point. The rest of it is pre-training or something akin to that. If they were able to use 100% for inference, or 90% for inference, this would be a massive gross-margin business.

The bear argument would most likely be that these businesses don’t really have a moat. Once they stop pre-training, why would people use the frontier intelligence, which you may not even need for the majority of tasks? People would go to the free tier, and it would become a game-theoretical race to the bottom in economics, where people use these free open-source models or free subsidized alternatives as companies fight for market share.

People flamed me for this comparison as a false equivalence, but when you use search, Google has a monopoly, so that’s not the best example to think about this. The best example for me is economy class on commercial airlines. When you’re in economy, do you really care if you’re flying Delta or Southwest or Spirit or anything? At the end of the day, there are three people in your row, and you can get the middle, window, or aisle seat. The plane is going from destination A to destination B, and ideally it leaves on time.

The end result—the end product, the end state—is probably the only thing the consumer cares about. The duration of the flight or the actual way the product and outcome get delivered will probably be pretty uninteresting to consumers. That could make you believe people will use these open-source models.

I don’t profess to be an AI expert. In fact, sometimes, especially when talking to other investors based in Silicon Valley, I think they’re extremely AI-tilted. I recognize that AI is a fantastic technology and is definitely changing the world, but I believe in the laws of financial gravity. I want to remain fairly unbiased and not be totally one-shot by the technology, so that I ignore some of the other financials.

So anyway, this was a bit of a rant, but this is what’s happened for people who may have missed it. One reason a lot of equity markets, specifically AI names, have been selling off considerably is that people are saying, “Hey, we were projecting all of the spend based on these incredible growth rates that these companies have.” These are some of the craziest businesses in the entire world, even if their growth rates are a little lower than people were projecting.

But this does affect the projected growth baked into a lot of these other companies. Specifically, neoclouds were among the stocks most affected in this recent sell-off. Names like Nebius, CoreWeave, and IREN were down 30%, maybe even 50%, over the past 2 weeks, which is quite unusual for public markets. This demonstrates that things are downstream from these 2 companies, and it’s very important to pay attention to sentiment, revenue projections, and the eventual, ideally, IPOs of these 2 companies. They’re going to have massive effects on the general equity markets and, specifically, equity names. And crypto is fairly correlated with equities.

Anyway, that was a bit of background, a bit of a primer. We chose to de-risk quite a bit earlier this week. We’ll get into more of what we’re looking at and some general trading philosophy, since I got quite a few interesting comments on Twitter this week. But yeah, Kyle, tell me a little bit about whether you agree or disagree with that summary of what’s been going on and what you’re looking at specifically.

3. How AI Spending Moves Markets

Yeah, I mean, look, I think what’s important to understand is the structure of the AI trade. You have 2 companies, OpenAI and Anthropic, that sit at the bottom. They say, “Look, I make a cool thing, but I require a lot of compute, and then I sell the thing.” You can take their revenue and the revenue of anybody else like them who’s maybe a smaller player, add it up, and that’s how much money AI makes—the actual utility that escapes this web of AI companies.

What sits on top of that is a bunch of stuff plugged into other stuff that’s all kind of self-serving. I buy a product from Nvidia, and maybe that product goes into a data center built by CoreWeave. Maybe that uses networking from Broadcom and touches some other thing. We have racks of servers, some with CPUs powered by AMD and others powered by Intel, and they all use memory. The GPUs use memory, too.

There’s a list of 30 or 40 or 50 names that are all downstream of OpenAI and Anthropic needing stuff. The entire amount of spend there is amplified, because one guy’s revenue gets spent at another company, and that company then has revenue it spends on another company. You have this massive amplification factor. Today, what looks like an annualized run rate is about $250 billion of people actually spending money on AI who are not themselves part of the AI complex. That’s the actual value being generated for the rest of society. Then that turns into trillions of dollars of this guy making money and spending it on that guy, who makes money and spends it on someone else.

So I think everything is very levered and sensitive to the real amount of revenue generated by the whole complex. We had misses, or at least misses relative to the rumored numbers we thought we were going to reach. There are a few reasons for this. One is that a lot of people correctly forecast that the numbers were way higher than they ought to have been, because there was a kind of willful ignorance around blocking distillation by China.

I think companies like Anthropic and OpenAI knowingly let China distill their models because it was good business. They could report really powerful growth and use it to raise these rounds. Both of them took advantage of those bumps in metrics to raise many tens of billions of dollars in capital. But that means their growth kind of flatlines when they turn off distillation and start banning those users. Now it looks like these are plateauing, stale companies, and stale companies that aren’t growing very much aren’t worth very much. That makes everybody downstream who depends on those spending commitments very nervous.

So there’s a wobble in the market. I wouldn’t say the market even really sold off that much, but there’s a wobble that says, “Hey, maybe things aren’t quite as good as they ought to have been.” Then people look, double- and triple-check what they own, and ask, “How convinced am I that this business is solid? If AI were to draw down, would this business be the first one sacrificed, or is it one of the more durable names in the space?”

You saw a lot of credit default swaps on CoreWeave, Oracle, and some of these businesses spike. I think the market is sifting out which companies would be the first to go if we have to sacrifice any. It’s interesting. I don’t know that it means anything more durable than that, but there’s definitely a lot of anxiety around what happens if Anthropic and OpenAI are only 70% as good as we thought they were.

Yeah, and that would still be a fantastic outcome if the other companies didn’t trade at pretty spectacular valuations. You saw quite a bit of hedging and de-risking over the past week in sector-specific names, not really in the global market generally. We’re still up 15% year to date. A lot of people were thinking, “Well, if things are slower, who is left holding the bag?” The market seems to think it’ll be CoreWeave and some of these neoclouds.

It’s very complicated. Kyle and I don’t really trade these names because you need deep, sector-specific knowledge of very complicated financing structures and contracts. If a data center is blocked politically, or if they can’t get it delivered on time, or if there are issues partway through, who is responsible? It’s pretty complicated, but there are really sharp pools of capital that trade these markets, and it’s important to pay attention to them. Obviously—

Yeah, so—

Yeah. No, go.

Yeah, so the way we’re thinking about this, or the way we’ve been thinking about this, is that the last time we did a Sunday Spaces, rates were really spiking.

Yeah.

The cost of money was going up a lot, and quickly, too. Rates actually got pretty high, at about 5.3% on the 10-year Treasury, and there was a rate hike, so money was kind of soaring. But nobody really cared. By nobody, I mean markets broadly speaking didn’t really think this had a meaningful impact on the price of equities.

Even when we had a really hot CPI number and a really hot PPI last week, rates actually went down. Money got cheaper. But at the same time, AI sold off a lot. It’s pretty clear, when you look underneath the market at the microstructure of how debt is being priced for a lot of AI companies and how money is being priced in terms of the cost of U.S. debt like Treasuries, that the thing really driving the market right now is how much money is going to be borrowed to spend on AI stuff.

Actual U.S. government things don’t matter that much. Inflation really doesn’t matter at all; nobody really cares. The main question is: Will there be $500 billion of borrowing next year for AI stuff, or will it be closer to $1 trillion? If the AI trade is strong, companies aren’t defaulting, and things are looking pretty good, they’re going to borrow a lot, and that makes money very expensive.

When AI stuff sells off and there’s a belief that things might cool for a bit, then we don’t have to—there won’t be that demand to borrow for AI-related spending. Money actually becomes quite available again. We’re seeing this ping-pong: if AI is cooked, at least for a little while, money is abundant and there’s a lot to go around. U.S. borrowing gets cheaper, and things get a lot nicer on that front. Businesses outside of AI now have capital available to borrow at reasonable rates.

But if AI is still the epicenter of everything, then money is going to be very expensive. For example, this next week, if AI rebounds and we learn that a fund blew up—that a lot of these signals in the stock market aren’t people’s genuine belief about whether the trade is solid, but rather a forced liquidation or something—then we should see a lot of AI stocks trade up. We should also see money get a lot more expensive and Treasuries probably return to their highs in terms of yield.

On the other hand, if AI continues to unwind, we should probably see money continue to get cheaper. In fact, we should probably see a lot of signals that go to the two-year Treasury suggesting that we don’t have any more rate hikes, right? If AI isn’t borrowing or competing for capital, I think those market signals largely disappear.

It’s just something to keep in mind, because what’s going on is very unintuitive. I think a lot of people and market participants are conditioned to look toward inflation and suggest that it’s the governing factor in what money costs. The reality is that we’ve never seen borrowing or competition for capital at the scale AI is bringing right now. It’s really something, kind of awesome.

And it’s not just the borrowing. It’s also the fact that quite a few of these companies are issuing equity. When they issue shares and inflate their share count, they’re subtly saying, “Hey, we think it’s prudent to shore up our balance sheet because we have large capital expenditures coming up. We believe it’s better to sell some of our shares to willing buyers than to borrow more.”

When you think about your counterparty in a trade, and the people running the company decide it’s a good idea to sell some shares—which is effectively what they’re doing—you may want to reconsider your position. That doesn’t mean you should immediately sell your entire position. I think people look at trading in a very binary way: because they feel slightly bearish, they immediately dust their entire portfolio; or because they’re bullish on a particular asset, they buy that day rather than being patient.

It’s important to look at equity markets, which have historically been very deflationary markets, and realize that quite a bit of equity is being issued because these hyperscalers and the largest companies in the world feel it’s existential for them to win the AI race. They might very well be correct, but that doesn’t mean we won’t see some short-term volatility as the market tries to figure out the value AI brings and the eventual economics of that business.

Right now, AI still doesn’t comprise a very large percentage of these hyperscalers’ revenues, other than in fairly recursive cloud contracts. It’s not like Meta’s Muse is suddenly making it billions of dollars. Microsoft doesn’t have a particularly large line item either. I don’t think they even really disclose one.

It’s important to balance your views and say, “If I believe AI has a chance of retracing, how should I position myself in the short term?” At the same time, you should always be able to articulate the bull case: “If I think AI is the greatest business in the entire world, how should I position my portfolio?” This is how you make money in trading, not by impaling yourself on a stake because you have a particular viewpoint for a period of time.

But okay, Kyle, what else do you want to cover regarding data centers, OpenAI, and Anthropic? I think we’re really just waiting to see Anthropic’s S-1. That will clear up a tremendous amount of this cloud over the market. We may see that it’s the greatest company on earth, and markets are going to absolutely rip, right? And bears like me are going to get blown out.

If we see that the growth rate has really stalled—and I don’t believe they have to disclose Q4 numbers in their S-1; I think their first reporting requirement will be sometime in January or February, when they report whatever Q4’s revenues are—there’ll be a period where the market may be really optimistic or pessimistic. It’ll be interesting, because there were some rumors of Anthropic coming out at $1.5 trillion instead of the $2 trillion they wanted. To be clear, I think they want to raise something like $200 billion. SpaceX raised—what was it?—$175 billion. Yeah.

Yeah, it was partly debt and equity. They did a little over $100 billion in equity, and then a big slug of debt at the same time.

Yeah, correct. I’m including the debt. Anthropic wants to raise something like $200 billion.

I don’t know. The S-1 is going to say a lot. Things feel odd, right? If you look at the Polymarket numbers for when Anthropic is going to IPO, they’ve been getting pushed back and back and back. The rumor or agenda started floating around about the timeline for when they wanted to IPO, and basically every update since then has pushed things back.

As of early September, the market had it at around 94% odds that they would IPO this year. Currently, that hovers around 70%, with only a 50% chance—a coin flip—that it’ll happen by the end of November. That used to be well north of 70%, I think. Clearly, the timeline has been slipping.

What we’re hearing from investors is that they’re not keen to share a lot of details with the buy side coming out to work on the IPO, which is odd. That’s not extremely typical. There was a rumor—I think a charitable one—that they wanted to wait for the Q3 numbers because they were going to be so good that the market would value them much more highly once those numbers were available.

But we have a rumor about what the numbers are right now, and I think they’re kind of disappointing. This suggests slowing growth, not increasing or reaccelerating growth. So it’s probably not the case that they waited specifically because there was a really good batch of numbers they wanted finalized. They need capital, and it’s not clear what the holdup is or why things are taking so long.

Now there are a lot of safety concerns. There was the initial safety scare, where everybody came out and said, “We need to pause the frontier.”

Yep.

Trump invited everybody over for dinner, and it was kumbaya. They all said, “Hey, you guys, pinky promise that you’ll have an auditor.” And they said, “Okay, cool. We’ll have a self-assigned auditor.” Things marched forward.

But now the crypto stuff has come out, and the reverse-engineering stuff is out. There have been a couple of new cyberattacks. Anthropic just floated the fact that it did some testing and saw self-replication of thin AI models hopping from IoT devices to IoT devices.

I think a lot of people feel that maybe the safety concerns aren’t over, and that this is something they need to work on. The fact is, these guys need a lot of capital, and it doesn’t appear that the private markets have much more capacity to fund this stuff, so they’ve got to go public. It’s not obvious what the deal is.

Then the $1.5 trillion number comes out, which isn’t a sign of strength, right? If you were slaying, you’d be asking for $3 trillion or $4 trillion.

I think it’s a little bit confusing right now. There’s always a chance the S-1 is phenomenal, but I look at it this way: as Flood alluded to, we don’t know shit, man. I’m not an insider at these companies. I don’t have a team of people sorting through every bit of minutia to build a detailed model of whether these companies are going to do much. I consume a lot of public data, like everybody else.

The view, or the spin, that I have is this: I believe the cadence of these people is that they genuinely think they’re building God. If that’s the case, the payoff is that you should do everything you can to achieve it as quickly as possible. You could frame it in a number of ways: it’s an AGI thing, or it’s a “we have to beat China at all costs” thing. But if there’s a dollar to borrow, you borrow it and use it to accelerate and go faster.

The limiting factor that tells you you’re spending as much as you can is when the cost of money gets really high. That’s what we saw in the past couple of weeks. That was the catalyst for us de-risking: the cost of money went up, and the AI trade was going to work and continue to go parabolic as long as money could be borrowed to fuel the spending.

Company earnings look phenomenal when everybody can borrow a lot of money to pay each other, right? As soon as borrowing gets prohibitively expensive, you should expect things to decelerate. I think that’s what we’ve seen in the past couple of weeks. Rates have come in and put a ceiling on what spending can look like.

Correct.

The next thing is, as we’ve seen in a lot of markets, we saw this with SPACs in 2021, and we’ve seen it in crypto in many cycles. There’s a certain amount of capital that wants to buy risk assets.

There’s a supply-and-demand dynamic between the money available to buy risk assets and the risk assets available to be sold into that capital. We’re seeing SpaceX unlock tremendous amounts of equity every couple of weeks, many tens of billions of dollars. We’re seeing Anthropic probably go to the market and ask for $200 billion in one slug up front, and then unlock another trillion over the next year or so. We’re going to see OpenAI probably do the same next year, taking $100 billion or $200 billion out of the market and then starting a TWAP, selling another trillion over the year after that.

We’re going to have something like $5 trillion of equity coming to the market in the next year, looking for people to buy it. It’s really hard to be constructive on risk assets when you have this really, really, really heavy sell order of risk coming onto the market. I think it’s particularly dangerous for capital sitting in the AI trade right now.

One of the lessons I take is from Circle. Circle IPOs; they’re the stablecoin issuer behind USDC, and their stock goes from a low price to a huge price in about 3 weeks. It goes vertical to a number that’s almost complete nonsense. What really happened is that you had a certain amount of capital in this world saying, “I want to bet on a pure-play stablecoin thing. I don’t want to own Coinbase equity, where the business has some exposure. I don’t want to own this other kind of equity, where the business has this exposure. PayPal technically has a stablecoin plan, and there’s a revenue line item they get from it. I really want to buy the pure idea.”

Nobody in that moment is really concerned about valuations, asking, “Am I getting a good price?” They’re saying, “Look, I want 1% of my portfolio to be in this theme, and I want it in this pure, clean, unencumbered thing.” It creates a lot of really, really goofy imbalances.

I suspect there are trillions of dollars of frustrated capital in the world right now saying, “I want the pure AI play. I don’t want the guy who makes the memory that goes in the chips, that goes into the data center, that gets used for AI. I don’t want the guy who builds the things or supplies the power or does whatever. I want to buy the AI company, the guy who sells AI as a product,” and that will be Anthropic.

I suspect there’s going to be a crazy rotation of capital out of these sort of “cuck trades,” where it’s AI-adjacent or AI beta, into the purest form of that. So I fully expect Anthropic to do really, really well. I just think it comes at the cost of people selling a lot of correlated things to buy it.

That’s a fantastic point. I think a lot of people make their trading really complicated, paying attention to 10 different things, when it’s important to have a clear thesis. It can be fine to have a complicated reason to get into a trade because these are complex systems with a lot of overlapping and interlinked variables. For example, rates affect people’s ability to finance. People financing large infrastructure deals can affect rates, since it’s the same type of borrowing pool.

But I think it’s extremely important for people to recognize and think about invalidation for their thesis. Whether it’s a level or some sort of fact changing, it’s very important to have something to point to when you can say you’re wrong or your thesis is violated. Otherwise, you may not see you’re wrong until it’s too late, and then price has moved materially away from you.

We can get into trading philosophy later, but I think you’re right. It’s important that everybody pay attention to the AI trade. Whether you’re bullish or bearish on AI, you should be ingesting sources on both sides and coming to your own conclusion. Then come up with a variable, a reality, or a state change that would make you wrong, and you can take another look at your portfolio.

I like to keep things very simple. When I see rates going up, I have PTSD from 2022, when I was complacent about rates going up and my portfolio was eviscerated. When I see rates squeezing substantially and the Fed hiking, I’ve been conditioned by previous experiences to learn that I should probably de-risk quite a bit. That’s a very simple reason why we de-risk.

We were also in substantial profits on a lot of the Bitcoin we had bought. Not everybody’s portfolio is exactly the same. I recognize there are quite a few new people following me and Kyle this week, some from Kyle’s brilliant podcast, some from the potential virality and reach of some of my tweets, some of which have a hint of satire in them.

That’s why we like to do these Sunday Spaces: we can be very clear and articulate about what we’re paying attention to. This isn’t a space where we say everybody has to short equities, long Bitcoin, or buy Hyperliquid. We try to provide a very balanced view, and we share exactly what we’re looking at to form our own conclusions. Two people can look at the same set of data and form two very different opinions, and that’s why we have a market, right? I think people forget that.

People love to make fun of the opposing side, but you should be thanking them for filling your bet, right? For you to be able to buy an asset at X price, someone has to sell it to you. For you to be able to short an asset at some astronomical valuation, someone has to be willing to buy.

I think it’s very important that people try to stay generally balanced. I noticed that when I spend quite a bit of time on Twitter, if you like bullish stuff specifically about any asset, the algorithm sort of overfits and starts feeding you every piece of bullish news it can, presumably to try to keep you on the app. When you’re bearish on something, it’ll suddenly do the same in the other direction. I remember I liked a CoreWeave bear tweet because I wanted to go back and revisit it later, and suddenly my entire feed was bearish CoreWeave posts.

It’s very important to attempt to remain balanced. This is also a reminder to myself, since it’s easy to get wrapped up in a position or have an asset become part of your personality, and that can be very dangerous. That’s happened to me in the past in a not-so-successful way. Thankfully, it’s happened in very successful ways as well. But it’s important to remain balanced. I can’t stress that enough.

Kyle, how—

So what?

Yeah, go ahead.

No, I’m asking you a question.

Sure.

4. Connecting AI to Crypto

Okay. Money has gotten pretty expensive. Depending on the market’s view on AI, it might get cheaper or it might continue to get more expensive. We had this sort of wobble in AI in the past week. We’ll see if that continues or if that was an aberration.

How does this all come back to crypto? How do you think about what goes on in equity land and AI land reaching over and affecting crypto?

Yeah. I was actually talking to a close friend about this today. Shout out to—I won’t name him, but he’s a very, very large investor. He was sort of like, “What do you think about the market here? You seem pretty convinced, or you seem pretty bearish. What are you thinking? What are you looking at?” And I relayed the thesis I just laid out for you guys, mainly relating to rates and the partial deceleration of AI. Those may go hand in hand.

I said, “Look, specific to crypto assets, I own Hyperliquid, I own some Bitcoin, but in smaller quantities than I have in the past—substantially smaller quantities. I’m about 70% cash.” For me, it’s important to have context here: I’m typically the kind of person who runs 80% net long when I have no thesis, just because I’m so bullish on Bitcoin and crypto generally. That’s how I’ve made the majority of my wealth: being exposed to crypto assets that have repriced materially since I entered the space. A lot came from trading gains as well, but a lot came from long-term spot appreciation. That’s been one of the fantastic things about the space.

I don’t have a particularly strong view on crypto at this moment. That might be a sort of sellout, cop-out answer, but it’s also important as a trader to recognize that and check swing, and say, “I’m going to wait for the next pitch.” If markets go a little bit higher, that’s fine, because I bought Bitcoin quite a bit lower and sold it for almost a $2,000 gain per coin, right? If markets go lower, I’m going to be happy, but I’ll try not to celebrate too much or root for them to go down every single day because I’m sitting in cash. It’s important to remain as nuanced as possible.

On crypto specifically, I’m going to keep it very simple again. If I’m worried about equities, as much as I would love for Bitcoin to outperform or trade really strong, and for this to be the time that things are different, more often than not it doesn’t turn out that way. In fact, in all of my nearly decade and a half in crypto, I have never seen Bitcoin trade exceptionally strong while equities sold off a material amount—material being 10%, 15%, or 20%.

I have never seen Bitcoin trade flat, much less go up. In fact, Bitcoin is often down even more than equities because it is a liquid instrument. It trades 24/7, and it is pretty heavily correlated with equities, whether we like to admit that or not.

By the way, KOSPI just opened down 2%. Just noting.

Just noting.

Just noting for all the people who are really excited about Nasdaq futures being up 70 basis points. KOSPI is very important to this AI trade. But anyway, I’m keeping it simple. I’m pretty bearish equities, and that has made it prudent to de-risk crypto. That’s it.

If my thesis changed around equities, I would most likely rapidly re-add my crypto exposure. If my thesis around crypto changed because I thought, for example, that they were going to give very preferential loan terms to AI companies, which would be highly inflationary, then obviously that would change my perspective on crypto. But I like to keep it very simple: if I get very bullish equities, then I will re-add to my crypto positions.

You’re not going to catch every move. You’re going to be wrong quite a bit in investing. I can always tell when people on Twitter are not very profitable because they love to go back and dig up any type of call from the past. They like to point out, “Oh, you tweeted something bearish close to the bottom.” It’s like, yes, it’s very easy to be bearish at the bottom. In fact, the majority of people are. That’s what creates a bottom.

You’re going to be wrong quite a bit. There was another great podcast that came out recently by Basics TNZ[?], also a friend of the show, and he mentioned that his win rate was something like 30%. Basics TNZ has made a tremendous amount of money trading over the past few years. We’re very close friends. I know this for a fact. He uses my same tax guy.

It just shows that, with a 30% win rate, having substantial P&L makes it interesting: you can be wrong so often, but in the moments when you feel an extreme level of conviction, you size appropriately, and that’s where you generate the majority of your returns. I like to think that I’m right more than 30% of the time, and I would say that my win rate is materially higher than that. But a good win rate in trading can be in the 50s.

People have this fetish for calling every move, and they think that people like Soros and Druckenmiller are infallible people who never lose money. Soros had mid-double-digit drawdowns plenty of times in his career, and he’s one of the greatest traders of all time, potentially the greatest trader of all time. We can’t all be perfect quantitative trading firms that never have a losing day.

Even when you look at presumably the smartest people in finance today, the people who are making the majority of capital, someone like Jane Street, they also had a massive losing month recently in July. Presumably, they’ll have a probably not so great month this month if equities continue going down and they maintain similar positions.

Anyway, this is just to say—and this is why I included trading philosophy in the tagline—that it’s important to recognize you’re going to be wrong a lot, and that’s okay. It’s all about how you respond to being wrong and whether you’re willing and have the emotional maturity to admit that you’re wrong. If you don’t, you are going to get liquidated and lose all your money. Maybe not all your money, but you are going to lose far more than you would have if you allowed yourself to remain fluid.

Anyway, another… I’m ranting this episode.

Yeah, I think being right or wrong is a timeframe thing. Suppose your thesis is that you have a bunch of factors: you think money being expensive is a factor, and you think competition for risk capital from a lot of IPOs coming out is a factor. You’ve got a lot of factors, but the timeline in which they materialize is six months.

That means that in six months, hopefully overall, you’re basically trending in the right direction. But you can get your ass kicked for many weeks. There can be some really nasty reversals where you’re dead wrong. In any sizable drawdown in the market, there are some really sick 3% or 4% up days, and up weeks that are plus 10%.

I think it’s important to also say: look, my alpha or my thesis, if right, is something that should be proven right over a period of months. What happens on any given day is kind of random.

Yeah. When we are not trading day to day, week to week, or hour to hour, that’s a very different trading style from the way that we like to allocate capital. Some people have been enormously profitable trading very short timeframes. I just personally find it kind of exhausting. I find it endlessly entertaining to speculate, but typically it takes a considerable amount of time for something to resolve itself.

Take Hyperliquid, for example. You bought the airdrop at, let’s say, a single-digit price. It went up to $30 quite quickly. If you had just taken that trade, you would have been great, and you would never have thought about that asset again. You would have done fantastic. To really hit the maximum win and have the opportunity to sell it close to $100, you had to hold through many peaks and valleys.

There were quite a few places where conviction was certainly tested: during Jelly Jelly, during the 2024 sell-off into the elections, or I guess the early 2025 pre-inauguration retrace as crypto came down a bit. There were quite a few moments when Lighter was significant competition. People were very worried about that. If you were trading very short timeframes, you would have sold at a pretty inopportune moment quite a few times.

So it’s important to recognize what timescale you’re playing on. Are you trading very short term or longer term? What would cause that to change? Is there something in the short term that could affect your long-term thesis? Is there something long term that may affect or bias your short-term trading?

I recently tweeted that I think it’s pretty important for people to read Buffett’s letters. I think Buffett is a very clear thinker who distills things very simply. Berkshire never really borrowed money or leveraged. Buffett always had a tremendous amount of cash, which gave him quite a bit of leverage and opportunity at times when people were strapped for cash. That is sort of how we are thinking about markets, specifically me.

Again, just to be clear, Kyle and I run our capital together, but we manage separate books. So Kyle and I may have opposing positions at times. We may disagree at times.

Mm.

So just because I’m tweeting that… People like to send one wallet that we have with some positions in it, and they go, “If you’re so bearish, why are you long Bitcoin?” I’ve explained it a few times, and I just don’t want every other tweet to be about explaining the machinations of our trading style. That may not be me. That may be someone else at the firm who is trading.

It’s important that you surround yourself with people who are able to think critically and disagree with you. Anyway, okay, Kyle, what are some other things—

The other shout-out I’d give—

Yeah.

While you’re giving the Buffett stuff, the other person I think is worth reading, and who has very concise letters every month, is Howard Marks.

Yeah.

Marks has been very good. If you’re looking for the weekly alpha of whether the market is up or down this week, he’s not your guy. But I think he’s really good to read if you’re thinking about whether this is a year to take a lot of risk or shed a lot of risk. He has some really good points of view, but it’s not stuff that anybody should think is going to play out immediately.

I think he’s very good at distilling the big picture and saying, “Look, equities are sick and AI is dope,” but then he’ll hit you with, “Yeah, but stock market wealth is a higher fraction of household income—or household net worth—than at any time in history.”

And a bunch of other factors, and you sort of look at it and say, “Wow, there are a lot of hundred-year trends that are all sort of crescendoing.” Which is interesting. So I think his writing also allows me to take a step back. I’ll add that to the Buffett pile.

Yeah, exactly. You want to learn from people who have demonstrated that they have tremendous amounts of alpha over very long periods of time and have seen similar state changes around the internet. They’ve lived through the internet boom and bust, and they’ve lived through the social media era, and they were probably wrong.

They probably got a lot of stuff wrong, but they were still able to maintain a very interesting way of thinking and a clear, concise rationale for why they owned the assets they owned or why they had the portfolio allocation they had. So, yeah, I think it’s important to go back, learn, and reflect on previous market cycles. That’s all you have: history. History… What’s the term? It doesn’t repeat, but it rhymes. Things aren’t going to be directly translational.

Listen, because Black Monday 1987 is the same day, I think that’s hilarious. I think it’s really funny. I’m definitely bearish on asset prices, but did I think there was going to be a 20% correction on Monday? No. I also use quite a bit of humor on Twitter, which some new people may not understand.

But, yeah, it’s important to listen to people if you feel like you don’t know what we’re talking about when it comes to rates, financing, or debt. Listen to people like Gundlach and some of these other managers who specifically trade fixed income. Read some of their writings and try to understand what all of this stuff means, because whether you want to be aware of it or not, it affects almost every aspect of the economy. The U.S. is the largest economy—

Yeah.

—in the world, and the rate of financing—

To be honest—

—in the U.S. economy affects everything.

To be honest, the market sometimes finds itself in regimes where a certain kind of investor class is in the driver’s seat, and everybody else is left to look at what they’re doing. Sometimes it’s commodity traders who run things. Sometimes it’s forex traders. Sometimes it’s equity guys. But right now, it’s definitely the fixed income people, right? They’re the ones causing a check on AI—

Yep.

—by suggesting that a lot of AI debt is at high risk of default. They’re also the ones determining whether the U.S. government gets to borrow at favorable rates, or whether other governments do, for that matter. So I think that if you wanted to listen to anybody’s sage advice right now to see where things are going, don’t listen to the equity commentators. They’re not the ones setting the marginal price of anything right now. Really listen to the fixed income guys.

Yeah. They’re experiencing far more volatility than equity traders, compared to whatever the typical market state is for fixed income. Spreads are widening. There are very interesting restructuring and refinancing deals coming for private equity and private credit. And, obviously, Oracle CDS, CoreWeave CDS, and the bonds around those 2 names specifically—which are 2 of the largest companies doing AI capex buildout other than the hyperscalers—have been going nuts.

It’s important to understand what that means, if anything just for the sake of becoming more informed. It doesn’t mean that everything you research means you suddenly have to take a trade on. We have no positions in Oracle, CoreWeave, or any of those other companies because we don’t feel like we have alpha. It’s very complicated.

But it’s very important to say, “If AI affects me, and I’m long AI companies, or I have this particular thesis, it’s important to try to learn as much as possible about the entire industry.” That’s where people ask, “How do I make money trading?” You make money trading by understanding things and having the ability to forecast and model certain realities better than other people.

5. Preparing for a Market Crash

So, another thing I want to talk about is, Kyle: in the event of a market crash—and again—

Yeah.

—while we may be bearish, we’re not forecasting some sort of seismic crash. I’d be very surprised if markets traded, like, 10% down next week. I think that’s exceptionally unlikely. Our thesis is on quite a bit of a longer timescale, and we’re not super massively aggressively short. In fact, we probably have low double-digit percentage shorts on, and we just have quite a bit of cash, which is very bearish for us.

But, Kyle, in the event that we have something like 10/10 that occurred last year—

Yep.

—what’s important to do prior to going into that day? Obviously, most people won’t have the foresight to know a day like that will happen, so if you’re lucky enough to be short, great. If you have a decent amount of cash because you think probabilistically, and you’re like, “Hey, I should have cash at any time so I can make moves if a day like 10/10 ever occurs again,” what’s important to do? What’s a way that—you trade best on these kinds of days—what’s a way—

Yeah.

—that you prepare for a day like this, regardless of knowing whether it’s going to happen or not?

Yeah. Okay. When you come into a day like that, the first thing you want to do is make sure you don’t get your ass kicked that day. What does that mean? It means you don’t want your capital tied up in things that are illiquid, where you can’t realize any cash from them.

In a crypto sense, if you’re behind some sort of staking unlock where your capital is always 7 or 30 days away from you, tough shit. You’re not going to be able to participate in a productive way. You want to make sure you have very instant access to your money. That’s how you’re going to be able to capitalize.

Beyond that, I think it’s really an asset selection game. You have to say, “Look, what do I want to buy here?” There are going to be a lot of things that go down by an amount that’s clearly wrong. They’re clearly going to reach their terminal price through some sort of forced deleveraging. Then you’d expect that when the deleveraging occurs, they snap back to something more reasonable, albeit at a price substantially lower than where they started.

You have 2 strategies. You can bottom fish for these things and say, “Look, good shot that something like SOXL, a leveraged ETF product, ends up way overshooting,” and then buy something like that. Or you can say, “Look, I have durable views on a couple of assets in this space. I have a few things I want to buy here that I intend on selling for long-term capital gains in a year or 2 or 3.” Then you just buy them at a good price.

You have to make sure you run that playbook for yourself ahead of time, before anything like this ever happens, because you’re not going to feel calm or in control. You’re not going to feel like you have a lot of time to sit down and contemplate these things when it’s happening. I think it’s a good drill to rehearse the what-ifs. Just say, “Look, what if we had a lot of stuff? What if the Anthropic S-1 comes out and it’s horrific, and investors panic, and the debt guys sell stuff down, and there’s this domino effect where the neoclouds look like they’re going to default, OpenAI and Anthropic aren’t going to make their spending commitments, Oracle’s in a tough spot, Nvidia backstopped a lot of these guys, and there’s this contagion?” In that moment, how would you respond?

Just running that playbook as a rehearsal will bring you a lot of calm, because as markets go down, you say, “Look, I’ve prepared for this. I’m ready for this.” If it doesn’t happen—and it almost certainly won’t happen, at least not in any sort of dramatic capacity—then no harm, no foul.

I think that, in terms of preparation, sometimes people—and ideally, they’ve learned—but some of my friends, people we know socially who trade crypto semi-professionally, were liquidated because they had bids sitting. A lot of people were in the bad habit of saying, “Oh, flash crashes happen all the time on some of these more illiquid venues.” This was when people were farming more perp DEXs. People would say, “I’ll set a bid 10% lower on some altcoin, or 20% lower.” Then those altcoins went down 60% or 70%. If you were conservatively 2 times levered, you got liquidated.

During periods like that, when we see 2-sigma massive events, all the weaknesses in your trading will be hammered and punished brutally. Do you have a concentration of assets on one particular venue? Well, guess what? What if that venue goes down? Not becomes insolvent, not FTX or anything, but what if that venue just goes down and you miss the most important hour of trading because these situations move very quickly?

What if you have assets on one particular venue, but the chains are all congested and there’s basically no amount of gas you could pay that would actually move your assets in a reasonable period of time? What if there’s an asset you want to buy that becomes extremely mispriced on a venue where you don’t have capital? That’s frustrating.

But more than that, what if you have a substantial amount of your net worth sitting on an exchange—which you shouldn’t, but if you do—and it’s cross-margined, and you never imagined an asset could go down 90%, but it does for literally just 1 second and you get liquidated?

These are important things to think about, and this is where you, as a trader, make money: other people won’t be thinking carefully about potential tail risks or things that could happen during periods of extreme volatility and low liquidity, but you will, if you have the foresight to prepare.

Again, when I tweeted, “Get your shopping list ready,” this is something Kyle and I have been talking about with the intern. When markets go down, you already want to have certain levels and certain assets that you’d be interested in buying. That doesn’t mean you need to have a limit order sitting there that gets filled and you just let it happen. You can always adjust. But it’s very important not to have analysis paralysis during market drawdowns, where you’re stuck watching the screen and thinking, “Oh my God, this asset’s down so much. I’ve just lost so much money.” You get shell-shocked, like a bomb went off near you.

It’s important to have a list and look at the market and say, “Well, if HYPE drops to $40, I’m just going to start clicking in X dollars every minute and hope it drops even more so I can buy as cheap as possible,” because I think anything under that level is value. These days don’t happen very often, but the difference between having a day where you make double-digit percentage returns and one where you lose double-digit percentage returns, in my opinion, all comes down to preparation.

We have this joke in the office that Kyle came up with: markets are pop quizzes. Every now and then the market will say, “Hey, pop quiz. Did you actually read the textbook?” If you did, it’s really easy, you should pass, and you’re going to pick up quite a bit of free money from all the other people who were caught woefully unprepared that day.

Again, this comes down to whether you’re a serious investor. If you’re not a serious investor and you look at markets for fun, mess around, and treat it as a hobby, that’s great. You shouldn’t be trading significant size, and you should be relatively unbothered whether indices go up or down because presumably you have something productive to do, like a job, Kyle. You focus the majority of your time on that, and you’re very interested in markets as a hobby. That’s great. If you’re a professional investor and you make these mistakes, then you’re not a professional investor. You have significant leaks in your game where you’re bleeding expected value. So, yeah.

Bro, bro.

You focus the majority of your time on a job that you’re very interested in, and markets are sort of a hobby for you. And that’s great. If you’re a professional investor and you make these mistakes, then you’re not a professional investor. You have significant leaks in your game where you’re bleeding expected value. So, yeah.

I would say the other thing, too, is that this is kind of a symptom of markets going down. If you started trading equities in the past three years, we haven’t had a real gut check in equities yet. Maybe the last real gut check we had was the Trump tariff scare about a year and a half ago.

Suppose there’s meaningful pressure on AI. It probably won’t be some sort of flash crash anytime soon, but maybe it’ll be this grind where every month is underwhelming. The growth isn’t really there. Things are bleeding down; it feels bad. There are new share unlocks from these big IPOs seemingly every week, and at some point things start getting really cheap.

When a market doesn’t go up and starts going down, it changes investor psychology. You’re a guy who owns a stock that’s been down 18 out of 20 weeks, and you start asking yourself, “Why the fuck do I own this?” Right?

Yeah.

Like, what is this?

Many such discussions around holding crypto in ’25. Yeah.

Yeah. And it’s like, “Why the hell do I have this again?” This is a common crypto thing, man. Ask some guy with a mega ETH bag right now, who’s probably thinking, “Why the fuck do I own this again? What is Plasma doing for me?” Any coin or asset that just goes down every day eventually makes you think, “Why the hell do I own it?”

Markets bottom when you can convince yourself, “I own it because of this. I can make a valuation case for it.” There’s a saying that markets oscillate between cash flows and stories. It bottoms when you can justify it and say, “I own this because it makes a lot of money, and I can reasonably expect to get paid back in a meaningful amount, or in a reasonable amount of time without much risk.” It goes up as high as the story will allow: I can tell a really good story about what it might do one day.

We’re probably pretty deep in story territory for most stocks. Not some chip names that are trading at a P/E of three or four.

Yeah.

But for a lot of aspects of the AI trade, especially the labs themselves, we’re in story land. If things slide down and we have six bad months or something like that, you want to be able to tell yourself, “I own something that doesn’t require a story for its valuation. I own something because it really is worth this much.” It’s a very different way of looking at your portfolio.

With crypto, you see things like Wormhole or StarkWare go up, and on a whim we invent a story: “Here’s what it might do. Here’s why I think another guy might buy this.” Then as soon as it stops going up, you’re like, “Why do I own this shit again?” And it falls until you hit a level where you can justify the valuation.

Crypto tends to do this pretty quickly, but I think this process takes months, if not years, to play out in equities. It’s something to think about. You might have had a lot of success in the last couple of years buying stories and owning stories that have a low probability. Then, as the story spreads and seems more and more likely, it gets priced in. But markets, when they do go down, tend to punish stories.

I remember in 2021 we had stories about how Zoom was going to be a company where we’d all be remote, in-office work and in-office schooling would be dead, and this was how we’d all communicate. The entire business world was going to exist on this product. We had Peloton, where going to a gym was toxic and everybody was going to use this equipment at home. And that was it. The stories work until they don’t.

I think AI has a lot of stories. Flood had a tweet about companies that have ballooned their valuations from $1 billion to $100 billion in a few months of fundraising.

Literally $50 million to $10 billion in a matter—

Yeah.

—of months.

In a matter of months, yeah.

Again, maybe that’s the logical thing to do in this market, where there’s so much M&A and so much interest around AI. If you’re a VC, it’s positive expected value to allocate. But sometimes I see things and pause. That would be an immense amount of value created in a few months by a relatively small team.

And it’s—

$9.95 billion of value in a few months is unbelievable. It’s one of the craziest value creation stories of all time, and it makes you pause.

Yeah, you look at it and say the storytelling around this got really good. There was really good storytelling that took place here. But if this thing got slammed for 100 days in a row, I think people would rightfully ask themselves, “Why the hell do I own this? Is the story really that good? Or would I rather own something I have a little more certainty over that allows me to sleep better at night, because I’m reasonably sure I’m going to get paid back on it?”

It’s something to think about as you consider what an all-weather portfolio looks like. Would you be willing to sacrifice some upside by owning things that have more durable multiples, more durable stories?

Yeah, and the VCs would say, “I have a 1x preference.” So when this company gets acquired, my downside is 1x. Important asterisk to that: it has to be acquired.

6. Questions from the Audience

Okay, let’s open up to some Q&A. Kyle, do you have any closing thoughts or anything you feel like we didn’t cover? I’m pretty much tapped out.

Yeah, I think we sort of said our piece. Our framework for looking at equity markets, in particular over six months or something like that, is that money is expensive, which probably caps AI spending to some extent, and that’s bad for a lot of names. The competition for risk capital is pretty high. These IPOs—we’ve never seen anything like it, right? Crypto has been exposed to it.

We’ve run it in our little sandbox of financial experimentation. It usually ends in tears. But equity markets have literally never seen a wave of new IPOs like this. I think the results will be obvious in hindsight, and so it’s making us very cautious.

Crypto, I think, is still its own thing on the side. Debasement is very interesting, and certainly the more AI gets nuked, the more likely it is that extreme measures are taken to bail it out or rectify it, or present it as some kind of race against China.

Keep it going.

But, yeah, to keep it going. And that will be extremely constructive for Bitcoin and debasement trades. So I think crypto is not something I lose sleep over. I don’t worry about it the way I worry about a lot of equity market stuff right now.

I have a closing thought. Don’t let anybody bully you into taking a trade because you have an ego or some sort of reputation to protect. We don’t monetize our Twitters. We do this for the love of the game. We do this because we love the crypto community. Also, it was $250,000 of premium. It’s important for you to only take the trades that you want to trade. That’s how you remain profitable as a trader.

Just because some people on Twitter want to see two people make a bet, trade your own book, focus on yourself, and try not to let other people tilt you too much. This is probably me speaking about myself as well. You don’t make money by rage trading or getting very sticky towards a particular thesis. You get paid for being fluid, responding to state changes, and being early and recognizing that change before other people do.

Okay, so let’s open up for Q&A. Hopefully this was insightful for people and maybe makes it a little clearer how we’re thinking about markets, or how we think about markets. We try not to be dogmatic or super binary about things. If you guys have questions, please queue up and we’ll approve you. We’ll probably take 10 or so. I don’t think we’re going to stay too long, but we really appreciate you guys tuning in.

Yeah.

And I know we do some trolling on Twitter, but for a lot of the new people following, let’s relax. I got a death threat. Obviously, I think the person’s kidding, but let’s chill out a little bit. Let’s keep it silly. Let’s keep it fun. If you have a lot of money on the line, that’s your fucking problem, not mine. So don’t come yelling at me.

All right. We’re going to go with “I Like Cardano.” That’s a crazy username. I Like Cardano. This Riley guy actually requested to speak. Spaces is always a fucking disaster, honestly, but it’s really the only way to do Q&A. Okay, let me see.

No, it’s terrible. It’s the worst part of the app, for sure.

Yeah, it’s terrible. We’re really limping around. All right, Riley, you’re up first. What’s your question?

Yo. Recently, I’ve been keeping a close eye on Bitcoin, and I’ve been thinking a lot, hypothesizing. In the ’70s, when we raised rates into a sort of global oil crisis, we experienced almost a 10x, or more than a 10x, I think, over the following years on gold.

Unfortunately, for the sake of trading this type of thing, it also coincided with us going off the gold standard. But I’m curious: do you think there are any parallels we can draw between the ’70s and now, or do you think that’s totally worthless and we should try to focus on the current situation as it is, and not try to draw historical parallels? Or do you think there are better historical parallels to draw than the ’70s?

I think that in this particular case, the historical parallels will likely fail you. I listen to a lot of people’s frameworks for what will happen or what is likely to happen as rates go high and the cost of money goes up. I think Jeffrey Gundlach probably has the framework I subscribe to the most right now. It’s the following: normally, as rates go up, the value you get for just holding US dollars and staking them with the government is pretty good. That crowds out things like gold or other things that look like money substitutes, because those don’t pay a yield.

What happens this time around is that the debt is so freaking high that you start wondering, “Is this sustainable?” The debt outlays each year are a huge fraction of GDP, and the forecast is that, as we run these deficits, it’s going to balloon. A spiral, I think, is something that is feared to exist in a couple of years, right? The forecasters of the spiral might be 20 or 30 years out, and it keeps getting revised four years closer every year.

So what’s likely to happen is that, as rates go up, instead of saying, “Hey, the US government has got a great yield farm,” they’re going to say, “Whoa, these guys are cooked.” There are serious things that are rotten here in terms of how much tax income is spent on servicing the debt. So that’s probably dollar down, and that’s probably really good for gold and Bitcoin and hard assets like property, just as a debasement thing. I think the historical correlations are actually quite likely to fail us here.

Totally agree. No notes on that. The Volcker hiking cycle is very, very difficult. Just because oil had a crisis, I don’t know that it’s directly translational, but, yeah, no notes on my end. That was perfect.

No, I appreciate that 100%. It gave me a little bit to think about. I might join back with another question next Sunday, if you guys run this again, but I really appreciate your time, and thank you so much.

Yeah, please do.

Thank you for the time.

Well, thanks for tuning in. All right, we’ve got Cooker Flips. It takes a while. I’m going to approve some people, and if you guys could please calmly wait. All right, Cooker, you’re up.

Hey, guys.

Yep.

I don’t know if you guys talked about it, because I joined the Space late, but I know you talked about the valuation of all these AI companies and tech companies. I was wondering if you had any thoughts on the private markets for some of the prediction market companies, like Polymarket and Kalshi, their valuations, and how much money they’re raising. All the talk in private markets is usually about AI and tech companies.

I’ll take this.

Yeah, it’s a good question.

Yeah.

Yeah, yeah. Take this. Of course.

I’ll take this. They’re very different companies. I think Kalshi is reasonable. I think prediction markets are generally a good idea, but implementation can be quite tough, and it’s going to be a very competitive market long term. I’ve publicly said that I’m very bearish on Polymarket. I think at $20 billion it’s crazy. Even at $10 billion it’s pretty crazy.

I’m not sure what multiple these companies should get, but it should probably be something similar to Nasdaq, where I think it trades at about 8x. A 10x or 5x multiple for exchanges is pretty reasonable. So the fact that some of these are being priced at 20x, 40x, or 100x because they had a Sydney Sweeney ad just feels kind of crazy to me. The sector will probably grow, but I’d probably be broadly neutral, if not a little bearish, on most of the current companies, especially at the valuations they’re at.

Yeah.

Can I get Kyle’s thoughts on that, and also on what you said about Anthropic people wanting to go to the non-related trades, just the kingpin companies, essentially? Do you feel like that’s applicable here, or do you think, like Flood, that something like Polymarket is way overvalued?

I don’t know that it’s way overvalued. It’s tough, because these guys have this huge regulatory overhang where it’s not obvious that their business is even legal.

Good point.

I think it’s hard to handicap what the hell it’s worth when, as far as things are leaning, these guys might just not have a business in two years. That being said, if you look at how Robinhood traded when they had a lot of prediction market revenue, the market seemed to like it. It was seen as growth, and it was seen as different and unique.

But I think that, from my point of view, prediction markets are largely a substitute for other things, like typical exchange activity. You see them wrapped into things like DraftKings now, where they’re kind of a sports betting substitute in a lot of cases. So I don’t think they represent a completely novel, new industry. I think the multiples or the comps I would look at are probably closer to MGM or a lot of these more traditional gambling plays.

If that's the yardstick, then that would suggest a much lower valuation than they get in private markets today, which I think is what I would be willing to pay for them.

Yeah. What does DraftKings trade at? They probably shouldn't trade that much higher than DraftKings, depending on their growth rate.

Thank you.

All right. Fill Your Bag, great name, is up next. I'm going to approve a few more people, and then we'll wrap up here. Phil, what's your question? The floor is yours.

My guy.

All right, we got—

How you doing?

We got full throttle here. Phil, you lost—

All right.

You've lost your connection. You've been frozen. All right, go.

Excuse me?

Go for it. Full throttle. What's your question? You’ve got the floor.

Yeah. Bear with me, okay? I'm from Southern California. I think you've alluded to your whereabouts previously through these Spaces and whatnot. I'm just curious: I think you said something about Pacific Palisades or something like that. Have you ever lived there? Brox Miller [?] Is that kind of off base? I know it's a little weird. I'm just curious, bro.

Holy fuck, that was the worst question of all time, bro.

Okay, we're going to—

You're talking to Flood right now and you ask him where he's from. You're Flood, bro. Where do you live, bro? Be honest.

We're based on the East Coast, in New York. Jazz, what's your question?

Are you long or are you short right now?

We're pretty risk-off.

I'm with HPD, too. He has a question after me.

We're pretty risk-off. If you listen to the Space, we went pretty in-depth as to what we're thinking, but I appreciate the question.

W Flood, bro, you're an inspiring trader, to be honest, bro.

I appreciate it. We'll go with the last question. Vaughn, you're last. We appreciate any and all questions, by the way.

Best for last. What's up, guys?

Perfect. What's up?

Just a quick question. I don't think I caught the beginning; I caught the middle. How do you see the midterms shaping up if the Democrats end up sweeping everything and then not passing the Clarity Act? What do you think that does to broader crypto?

I think it's way worse for crypto than it is for the rest of the market because crypto needs at least a little something to get done. To put it into perspective, the midterms don't matter that much because the worst that happens is that both the House and the Senate flip Democrat, and then you have a lame-duck president, which is usually actually pretty good. Markets like it when nothing happens, and that's usually seen as pretty good.

The problem is that I think crypto needs more than that. Crypto needs some things to go right and some things to get codified into law, just in case the Democrats win in 2028 as well. Then there could be a bit of a swing back the other way, where they become very restrictive on crypto. A lot of people are sitting in the middle right now, saying, “My business is okay right now. Nobody's being hostile towards it, but it's not obvious that that'll continue for two more years.”

As we were just talking about, prediction markets in particular are in the crosshairs. Their valuations look a lot different if the Democrats win in 2028, and there aren't yet things inked into law to protect them if that happens. I think the midterms are a trial run for whether public sentiment is swinging Democrat, or whether this is just a goofy two years and things will go back Republican in the next few years. My intuition is that things are swinging Democrat, so I think the Democrats probably win the midterms, and I think they'll be favored for a while to win in 2028 as well.

Okay, cool. One more little toss-in. I caught the valuations part. Do you think bigger desks and private equity are driving valuations so high because they're just like the GameStop, quote-unquote, “dumb money,” where retail floods in and they couldn't tell you what a valuation is or means? Do you think a lot of desks and bigger private equity firms are acting with that trader psychology, pushing valuations crazy high, riding the story for what it's worth, and then rotating the money out once the story dies?

It's funny because, depending on where you look in the market, valuation is either extremely high or extremely cheap. If you're one of these memory producers, like SK Hynix or Micron, you're making 90% net margins on some crazy revenue number, where you could buy back every single share in the market at today's price in about three years. A lot of people look at this and say, “Well, actually, the valuations are very cheap.”

The problem is that it's also well recognized that all this money flowing around is a weird circular thing, where a lot of companies have unsustainable earnings. A lot of companies are overearning right now, so their earnings look really, really, really good. The market is telling you that Nvidia isn't going to have earnings like this for that long, and the memory companies aren't going to have earnings like this for that long.

On the flip side, if you had to value Google based on how much money AI is actually making them relative to what they're spending, or ask the same question about Microsoft, Amazon, or Meta, they would be disgustingly expensive companies. Thankfully for them, they have other revenue that smooths things out and makes it look less bad. But if you're a producer of AI and sell AI services to an end consumer, you look like one of the most incredibly expensive companies ever on that endeavor. If you're a guy who just builds chips, you look very cheap.

I think the truth lies somewhere in the middle. I think the market is right to price the chip companies cheaply, and AI revenues will go up. These companies will figure out more ways to be productive with it, and generally speaking, it takes people a long time to figure out how to use these things. But is every human on Earth going to spend $3,000 a year on AI? If they don't, a lot of the math isn't mathing. It's tough, but the question about valuation definitely depends on where you're looking. You can find a lot of people who are right to say things are either omega cheap or giga expensive.

Yeah, that's what I was thinking, too. All right. Appreciate that.

All right. Thanks, guys. Thanks for the great Spaces. Thanks for the great questions. We appreciate everyone that listens in. Let's try and be positive this week regardless of what happens. Like I said, I think a lot of new people tuned in this week, which is great. I wish we did this every week, but we don't, just because I think the show would be too boring sometimes, or we'd sound repetitive. So we try and do it around exciting time periods. But, yeah, we love you all. Well, most of you. And we really appreciate you guys tuning in. If you guys enjoyed this, leave us a like. I know this was a pretty long one, but it was a pretty exciting time period in the markets.

Yep. Have a great night, everyone.

All right. Bye, guys.