Gavin Baker on Anthropic's S-1, GPU Prices, and SpaceX
- Baker reads Anthropic’s S-1 as an upside surprise with substantial enterprise headroom. He points to $10 billion burned, plus encouraging profitability, cash generation and unit economics. About half the Fortune 500 has adopted Anthropic, yet it represents just 0.1% of their expenses—suggesting diffusion is “maybe earlier than I thought we were.”
- Anthropic’s disclosed customer mix challenges the concentrated-spending bear case. Baker cites only 250 accounts spending over $10 million, while a host highlights 11,000–12,000 spending over $100,000. Baker sees a “very diversified spending base,” while allowing that Ramp’s concentration data might still describe AI spending overall.
- Open-weight infrastructure companies are outgrowing OpenAI and Anthropic, according to Baker. Fireworks, Baseten, Together, Modal and Cognition underpin his argument that demand is broadening beyond the major labs. Atreides led Fireworks’ latest round, where growth meant “the multiple was effectively going down” daily. Growth beyond Cursor is particularly notable as SpaceX insources Cursor’s compute.
- GPU pricing signals tightness through 2027 despite the announced supply pipeline. Baker reports daily rental-price increases during October and expensive forward capacity: portfolio companies seeking another 4,000 Blackwells for September 2027 face sharply higher quotes than for their May–June 2026 clusters. “We’re not in backwardation anymore.”
- Compute returns increasingly depend on execution and retaining pricing upside. Baker praises CoreWeave’s operations but says substantial capacity was contracted far below today’s market. A host argues that more spot exposure and prepayments can produce paybacks “well under a year.” Running clusters resembles “a Formula 1 team”; Baker says a new Jensen product with preassembled racks could shorten deployment and reduce working capital tied up in idle GPUs.
- Training demand creates a potential inference-supply risk within Baker’s bullish framework. A bearish argument he considers compelling is that Anthropic dedicates 74% of its compute to training. That suggests room for more inference revenue, but if training needs decline, capacity could enter the inference market.
- IPOs and repriced cloud contracts could reduce the build-out’s dependence on external financing. SpaceX and Anthropic becoming public and fully unlocked could restore mutual funds’ capacity for private investments. Higher cloud pricing could also lift operating cash flow enough to fund more of 2027’s infrastructure internally; Baker does not know whether external financing has peaked, but says, “I think we’re getting close.”
1. Anthropic’s S-1 pairs enormous scale with low enterprise penetration
- Baker calls the S-1 “pretty clearly an upside surprise”: $10 billion burned, with encouraging profitability, cash generation and unit economics. About half the Fortune 500 has adopted Anthropic, yet it represents 0.1% of their expenses. Enterprise diffusion may be “earlier than I thought we were.” Baker also cites roughly $67 billion annualized if the reported third quarter is annualized.
- A host argues that concentrated AI spending is natural among early obsessives, citing Ramp data suggesting 1% of companies account for 90% of spending. Baker disagrees about Anthropic’s customer mix: only 250 accounts spend over $10 million, while a host cites 11,000–12,000 spending over $100,000. Baker calls the S-1 more authoritative and says it shows a “very diversified spending base,” while allowing that Ramp might still describe AI overall.
- Baker’s broader demand signal is open-weight infrastructure: Fireworks, Baseten, Together, Modal and Cognition are growing faster than either OpenAI or Anthropic. He says that was true during OpenAI’s summer acceleration and remains true amid Anthropic’s October reacceleration. Baker says Anthropic lost share in summer and is now gaining it. He doubts it would leave the October ARR blank in the S-1 without expecting a strong number, while hoping the market moves beyond monthly ARR figures.
- A host says private-market investors on the West Coast already know these companies are at scale and growing, while the news is less familiar on the East Coast. Atreides led Fireworks’ latest round, and Baker says growth made pricing difficult: “The multiple was effectively going down” every day. He finds inference-cloud growth beyond Cursor especially striking because SpaceX has been insourcing Cursor’s compute. His assessment is that the ecosystem has accelerated and diversified.
2. Grokbot wins Baker through ease of use and information routing
- A host asks whether Cursor’s product and distribution can help Grokbot despite xAI models not stunning on a particular benchmark. Baker answers as a user: Grokbot is “the most impactful AI product” for him and his firm, and he has about 17 Grokbots. He had kludged together a setup with Claude Code and says that may have been user error.
- Baker describes investing as “making sure the right information gets to the right person at the right time,” and says Grokbot is very good at that. A host adds the Sunday-night research example: an agent can spend a million tokens investigating a market-size question without asking an employee to work for six hours.
- Personal agents have not displaced his assistant Susan. “I guess I have executive assistant privilege,” Baker says; he has downloaded Instinct and played with Muse, but says Grokbot has been transformative for his professional use cases.
3. Compute scarcity extends into 2027, but operators capture different economics
- Baker says October rental prices for essentially every kind of GPU have risen daily. Silicon Data, in which Atreides is an investor and which Baker calls the most authoritative source, says compute is now in contango: forward capacity is expensive too. Portfolio companies with 4,000 Blackwells for May or June 2026 are shocked by the cost of adding another 4,000 in September 2027. Despite supply coming online, Baker says it is hard for them to find compute well into 2027: “we’re not in backwardation anymore.”
- Baker takes Sarah Friar’s comments literally: long-term compute commitments signed before May 2026 are “way in the money.” A contract buying compute for perhaps $10 or $15 a watt could be resold at a large premium; Friar described a “3, 4, 5X premium.”
- Baker says CoreWeave runs GPUs well but locked in substantial compute under long-term contracts now far below market. A host explains that CoreWeave sold at fixed prices and argues that more spot exposure and fewer long-term offtake agreements could improve the economics. The host says selling into the spot market could mean paybacks well under a year and large prepayments.
- The host compares running a GPU cluster to running a Formula 1 team, requiring experienced operators and coordinated execution. Baker agrees that the key is bringing on GPUs, energizing them and running them well. He names CoreWeave, Crusoe, Nebius, SpaceX and the big hyperscalers as companies with a track record of execution, while saying the future remains uncertain.
- Baker points to a new Jensen product: seven racks, preassembled, that you “just literally plug in and it works.” He says some operators otherwise take three or six months to energize GPUs; at this scale, six months could leave hundreds of billions of dollars in assets generating no revenue. Shortening deployment would take working capital out of the market.
4. Responsible optimism includes a training-driven supply risk
- Baker says he treats X as an exercise in “cultivated equanimity,” searching criticism for useful bearish arguments. One he finds compelling is that Anthropic dedicates 74% of its compute to training. That could indicate room for more inference revenue, but if training needs decline, substantial capacity could enter the inference market. “I’m not on X to be right”; he says he is trying to learn and get to the right answer.
- A host raises the timing concern in biology: infrastructure costs arrive now, while drug revenues may take years. Baker says “we’ll see how long it takes,” and points to Montana’s new-drug program and the possibility of using existing compounds for new diseases. He says AI has helped identify a drug that has existed for 50 or 60 years and can effectively cure a rare disease that devastates 10,000 children a year. A friend did the work, but Baker says it is the friend’s story to tell.
- Baker says AI’s public-relations effort is “-5 out of 100” and the industry needs to tell more stories. His public case includes lives saved, benefits for blue-collar Americans and small towns, reindustrialization, and increased electrical generation: “You cannot reindustrialize without power.” He recounts a skeptic demanding data centers in America’s wealthiest county, then points to Loudoun County, which Baker says is the wealthiest and has the highest per-capita number of data centers.
5. Commodity research is broader than Atreides’ portfolio exposure
- Asked how far he follows energy infrastructure and commodities, Baker says he pays close attention. He ran a diversified fund for nine years and previously invested in energy and natural resources. He has not made those areas a large part of the portfolio, but says investors need to cast a wide research net and focus capital where the firm has a competitive advantage.
- Baker says gold has long had a place in portfolio construction. He finds it amusing how strongly some gold bulls react to asteroid mining, calling it a challenge to their worldview; a host compares the possibility of technological disruption to lab-grown diamonds.
- His concrete private-market exposure is a company with a chemical process for refining rare earths, rather than acquiring mines. “The problem with rare earths isn’t that they’re rare”; refining them is dirty. Baker also credits DARPA’s efforts and says investors never know where a weak but important signal will come from.
6. IPOs and cloud cash flow could loosen the capital constraint
- Baker argues that SpaceX and Anthropic IPOs can unlock another pool of private capital. American mutual funds face limits on illiquid holdings; when those companies are public and fully unlocked, funds regain room to buy private assets. Baker names Capital Research, T. Rowe Price, MSAM, GSAM and Franklin Templeton as existing participants.
- Sovereign funding is already substantial, particularly in the GCC and UAE. Baker singles out MGX and Mubadala, calling the UAE “extremely front-footed” on AI; Atreides and Valor co-led Crusoe with Mubadala.
- Baker expects aggregate hyperscaler and neocloud revenue growth to keep accelerating as underpriced contracts reset and more compute gets spot exposure. Revenue is dropping through at high margins, lifting operating cash flow and potentially financing more of the 2027 build-out than people expect. His hedge matters: “I don’t know if we’ve seen the peak of external financing,” but “I think we’re getting close.”
- A host suggests public ownership could moderate polarized AI sentiment as direct holders grow from perhaps 50,000 people to millions. Baker favors direct ownership over a foundation whose board or bureaucrats allocate the money. He says every child in America will have a SpaceX share and encourages more companies and wealthy people, including Michael Dell and Brad Kirshner, to support broad participation: “Give everybody a stake in capitalism.”
Full transcript
Our guest is Gavin Baker from Atreides Management, the managing partner—
You may have heard of him before—
Oh, he’s been on the show before.
…the Baconator.
We’re going to bring him into the TBPN Ultradome and catch up with him. How are you doing, Gavin? Good to see you.
I’m awesome, man. Happiness is a choice. I’m only ever fantastic, awesome, superb, phenomenal.
I completely agree.
Why be anything else?
I love that happiness is a choice, but it also feels like this is a week to be extra happy. The Anthropic S-1 went out, and it seemed like a beat. If you were worried the economics weren’t secretly working behind the scenes and there was going to be some bombshell in there, the haters have been kind of quiet. Ed Zitron hasn’t posted. It seems like things are on track. What’s been your interpretation of this week?
1. Anthropic’s S-1 Beats Expectations
I do think the Anthropic S-1 was pretty clearly an upside surprise.
Yeah.
The fact that they’ve only burned $10 billion, the profitability, the cash generation, the unit economics—they’re all very encouraging. I put it on X, but I just cannot believe that half of the Fortune 500 has adopted Anthropic and it’s 0.1% of their expenses. It’s just a wild stat.
Yeah. The funny thing is that—
Did the internet just go out?
I read that as—
Yeah, we can still hear you.
No, yeah, we can hear you. I read that 50%—
No.
What?
Yeah, yeah.
Oh, just like, okay, that’s incredibly impressive for a company in just a few years to get to half the Fortune 500 or the S&P 500 as customers. But I was kind of thinking, what are the other half of the S&P 500 doing? I felt like every Fortune 500 or S&P 500 company would be using everything across multiple products. But there are a lot of companies that just haven’t gotten through the adoption hiccups of rolling out AI broadly. Maybe some are all in on Google or OpenAI, but by and large, I think some are still figuring out how AI can be deployed in their organizations. Is that your read on it—that we’re still early on the diffusion story in enterprise?
Yeah, very early. If anything, maybe earlier than I thought we were. It’s just astonishing. Whatever the number is—$67 billion annualized, if you annualize the reported third quarter, wherever the ARR is right now—to be that big with such low penetration into the Fortune 500 is pretty extraordinary.
Yeah, totally.
How have you processed some of the criticism of the AI trade right now—that a very small number of companies are driving the majority of the spend? Historically, that same dynamic probably played out with social media ads, where a very small number of companies were driving a massive amount of the spend. You’re going to see power laws in a lot of places.
My read is that there are companies based in San Francisco that are 1,000 times more obsessed with AI than the long tail of public company employees. Of course they’re going to spend dramatically more. It’s possible that long tail never actually cares enough, and they fade to relative obscurity.
You have the disruptors saying, “Yeah, this technology is real. It works. It’s not perfect yet. It’s not a drop-in coworker that you immediately give $1 and it gives you $10 back.” But it’s very capable, and if you obsess over it… That’s been my overall read. I don’t think it’s necessarily a problem. Of course, you’d want a more diversified revenue base if you could, but it feels very natural that all the spend is coming from a very small group of people who realize something fundamental has changed and are trying to take every advantage of it while the rest of the world is asleep.
Mm.
I might respectfully disagree a little. In the S-1, there are only 250 accounts spending more than $10 million. Somebody in The New York Times said maybe 10 days ago that 80% of their revenue came from eight customers. I was actually shocked by how diversified their customer base was.
Hmm.
And if you are a—
The number to me was something like 11,000 or 12,000 spending over $100,000.
Yeah.
That feels diversified.
Oh, yeah.
But I’m talking about the Ramp data that says 1% of companies are spending 90% of the dollars.
The Ramp data is what it is, but we have much more authoritative data now—
Yeah.
Yeah, the S-1.
—from the Anthropic S-1.
Yeah.
Yeah.
And it shows a very diversified spending base for Anthropic.
Yeah.
Maybe the Ramp data is correct for AI overall. But relative to the idea that Meta was spending $10 billion and there were multiple billion-dollar-plus accounts driving it all, the Anthropic S-1 shows a much more diversified customer base.
For AI broadly, I do think something has happened since June. First, we saw OpenAI really accelerate in the summer. As OpenAI accelerated, so did open-weight models. There are so many companies—Fireworks, Baseten, Together, Modal, Cognition—that are at scale and really open-weight first, and they’re all growing faster than either OpenAI or Anthropic. That was true when OpenAI was really accelerating in the summer, and it’s still true in October, when Anthropic has had a pretty big reacceleration.
I’m skeptical they would have put a blank space for the October ARR in the S-1 if they didn’t know it was going to be pretty good. I do think it’ll be great for the market to get away from these monthly ARR figures. But to me, the story—
Yeah, it’s interesting. Seeing the Modals, Basetens, Fireworks, and Togethers all out in the market—you can probably… I feel like this is common knowledge. They’ve all been out in the market, so private market investors have been getting these updates. A lot of these companies have been starting fundraises and growing dramatically from when they got their first term sheet to when they’re actually pulling in capital.
That data is broadly known on the West Coast, I’d say, but the East Coast just isn’t aware of it at all. Some are picking it up, but it’s not common knowledge. Over here, people are like, “Yeah, these companies were already at scale,” and then they’re growing double digits just from first term sheet to close, or first close.
My firm, Atreides, led the last Fireworks round, and it was actually hard to price because they were growing so fast.
We had a handshake, I thought.
Yeah.
You’re like, “Well, we’ve increased revenue by 20%.”
Yesterday’s price is not today’s price.
Yeah, you’re like—
Yeah.
“Please, let’s just get the deal done.”
Yeah.
Lock it in.
Literally, it was like the valuation was going—the multiple was effectively going down every day.
Every day. Yeah, that’s crazy.
If anything, it feels like that ecosystem has accelerated and diversified, because earlier this year Cursor was a real big driver for all those inference clouds. The ex-Cursor growth, I think, is what’s really extraordinary, just because SpaceX has been insourcing the Cursor compute.
Yes. On Cursor and SpaceX, with the post-merger integration, it feels like Grokbot has stepped forward in terms of capability. Is that the key to growth there? We haven’t seen the models from xAI stun on a particular benchmark, but the Cursor team has always been fantastic on product and distribution. In the prior era, Grok was putting up pretty remarkable results on the benchmarks, but then there were all the political considerations, and the product was banned in some companies. It feels like Grokbot might be in a particularly different slot because they have that go-to-market motion through Cursor, with the Cursor design team behind it.
Is that what’s driving growth there? What are you seeing on the Grokbot side that’s showing acceleration in that product category?
2. Grokbot Changes How He Works
As a user, Grokbot is the most impactful AI product for me and our firm that we’ve had yet. I kludged together all this stuff with Claude Code, and it really… I’m sure it was user error on my part.
Yeah.
I’m 50 years old.
Yeah.
I’m doing my best. But Grokbot makes it so easy. I think I’m up to 17 Grokbots.
Whoa.
And every day I think of a new one to add.
That’s great.
Yeah.
Yeah, you’re getting a layer of somebody—the senior manager, the senior Grokbot. How are you personally dividing AI agents across your personal and professional life? A lot of people have a family office with a personal assistant who manages things like getting gifts for family members or planning birthday parties. Separately, they’ll have an executive assistant who plans business travel and their work calendar. It feels like we might be entering a world where those two merge and you have one agent that does everything. But we might also end up in a world where people use Grokbot or Claude Code at work, and Instinct or Muse for personal use, while OpenAI has products in both. There’s a whole array, and it’s a decision for everyone. How have you been grappling with it?
I guess you’d say I have executive assistant privilege, because my assistant Susan is exceptional.
Okay.
And she, yeah—
Not going anywhere.
I’ve not felt the need to… I’ve downloaded Instinct and played around with Muse.
Sure.
I could see they’d be super amazing if I didn’t have someone like Susan in my professional life, but I do. For the more professional use cases, Grokbot has just been transformative for me.
Yeah.
So much of investing is making sure the right information gets to the right person at the right time. Grokbot is so good at that.
Yep. There’s also something satisfying about AI agents: it can be 10:00 p.m. on a Sunday, and you can kick off a real research effort. Yes, you could probably call someone on your team and say, “I need you to…” But that would cross a professional line. You’d be a bit of a hassle as a boss, and people might not want to work there because they’d think, “He calls me every time he takes a shower. He has a shower thought, and then I have to work for six hours.” Instead, the AI is like, “Yeah, boss, I’m right here. I’d love to look that up for you and spend a million tokens figuring out the market size of something or other.” But, yeah.
My policy is that anybody at Atreides is always welcome to call me anytime on the weekend. I try not to call people on the weekends.
So—
Rewinding this week, it’s Anthropic S1 week, so there’s a lot of focus there. Last week, the general attention was on how much supply is coming online next year, and how many gigawatts SpaceX is going to bring online. You’ve generally maintained a sort of bullishness that some other people on the timeline have lost here and there. How are you processing how the industry can absorb all of that supply if we hit the higher end of some of the ranges being discussed? Is it just Jevons paradox again and again, until everyone finally internalizes it? People like to say it, but it’s hard to fully internalize.
I think that’s right. It’s easy to say; it’s hard to internalize and believe. October has been a strange month because the price to rent essentially any kind of GPU has gone up every day. The spot market for compute is getting tighter, and it’s not just the spot market. I love reading these primers on X from commodity and macro investors about these super-complicated concepts called backwardation and contango, and what they might mean for the forward curve and future price of compute.
Yeah.
Contango, backwardation. Silicon Data—which my firm is an investor in, and which I do think is the most authoritative data source—has a founder, Carmen Li, who stood up at GTC and said the compute market is now in contango. The forward market is not just tight in the spot market. If you want a cluster a year from now, you have to pay a lot of money.
Yeah.
This is maybe another discontinuity between public and private markets. The number of companies in our portfolio saying, “Okay, we got a cluster of 4,000 Blackwells in May or June of ’26, and we want to bring on another cluster of 4,000 in September of 2027,” is striking. They’re shocked by how much more they have to pay in the future. Everybody is aware of the supply that’s coming on, but even with all of that supply, it’s hard for the companies in our portfolio to find compute well into ’27. So, for all the commodity experts focused on backwardation and what that implied, I’ve got news for you: we’re not in backwardation anymore.
Mm-hmm.
As far as how much capacity is going to come online, I think one healthy thing about what’s happened in these financing markets is that I hope it concentrates capacity and supply among competent operators who can finance it through prepayments.
Hmm.
You can finance these GPU purchases two ways. By the way, CoreWeave is great. CoreWeave runs GPUs well. They bring them on; they’re very, very good. They just sadly locked in a vast amount of compute in long-term contracts that are now way under market.
Oh, interesting.
They’re doing 100%—
So they sold their compute for fixed prices, and that’s why they’re getting hammered right now. It’s not that they don’t have compute or aren’t bringing it on. They have the SemiAnalysis Diamond tier. Everyone knows the clusters work. They’re good and reliable. They just sold for too cheap, and they should have held on or had variable pricing.
Do you think Elon studied that and said, “We’re not going to make that same mistake,” and that’s why they’re adopting this more short-term contract model?
I don’t think Elon was probably studying CoreWeave or Nebius, or really anyone else. He’s a first-principles thinker. He was probably reasoning from first principles. But the conclusion he came to is that prices are going up. If we’re willing to sell into the spot market, we can have paybacks well under a year and get big prepayments. I think that’s a better financing model for the market we’re in. What I also like about it—and CoreWeave is very aware of this—is that they’re trying to get more spot exposure and do fewer of these big, long-term offtake agreements.
Standing up a big GPU cluster—not just bringing on the power and standing up the cluster—is like running a Formula 1 team. A bunch of people with no relevant experience acquire some land and power, then get a zero-down-payment offtake agreement from someone. A lot of that is being cleaned up. I think supply is being allocated through this forcing function to people who can actually bring on a cluster, energize the GPUs, and run it well. Running a Formula 1 team is really hard. You have to pack everything up, put it on a plane, go to a new city, unpack it, set it up, and set the car up. You need all the people to change the tires, change the rear wing if you get in a collision, do whatever needs to be done during the race, work out the strategy, and communicate with the driver.
Yeah.
You need all of that.
It’s really, really hard. Some of these people don’t have the relevant experience. CoreWeave, Crusoe, Nebius, SpaceX, the big hyperscalers—
These are companies that can bring on GPUs, energize them, and run them.
I’m sure.
Seven racks, preassembled. You just literally plug it in and it works.
Yeah.
Right now, when you get the GPUs, maybe if you’re SpaceX and you’re really exceptional at execution in the physical world, you can get them energized quickly.
Yeah.
But for a lot of people, it takes 3 or 6 months to get them energized and really working. If we can get that time down through this new product from Jensen, I think that’s healthy, because it effectively takes a lot of working capital out of the market.
Mm-hmm.
As these numbers get bigger, if it takes 6 months to energize them, that could be hundreds of billions of dollars of assets that aren’t generating revenue. I think that’s another product innovation that will help with this. We’ll see who energizes GPUs in ’27. I’m confident the companies I named—
Yeah.
—I think they’re going to. The future is always uncertain, but these are companies with a proven track record of execution.
Yeah.
We’ll see about others.
I have a follow-up on that. One click higher from the neo-clouds or the neo-labs: there’s been speculation that neo-labs might pivot into neo-clouds. Is that dynamic going to look like they sign a multiyear offtake agreement with a neo-cloud, and then they want to sell into the spot market? Is that what’s going to happen?
Yes. What Sarah Friar said—and Sarah Friar is an exceptional CFO—was, “Everybody made fun of us for signing all these compute commitments, and we could sell them all at a 3, 4, 5X premium.”
Today, yeah.
She was being literal. If you signed a long-term compute contract at any point before May 2026, you are way in the money. You’re buying compute for maybe $10 or $15 a watt, and you could sell it at a massive premium to that today.
Yep. Sure.
The point Carmen made about the forward curve being in contango is really important. It’s not just that the spot market is tight today. Even with all this capacity coming online, you have to pay a truly vast amount to bring on a cluster in mid-to-late ’27. I think public markets may not be appreciating that, even though all of this is public. Sarah Friar said that.
Yeah.
All these inference clouds have put out press releases. You hear a bunch of people on X talk about how hard it is to find compute into ’27. Carmen Li stands up and says the market in ’27 is crazy strong. It’s all out there, but for whatever reason, it’s not legible to the public markets.
Sure. Jordan?
What’s your outlook on biology as an area the labs are investing in? It feels very exciting, with tremendous potential, but the overall concern is around timelines. There are very real infrastructure costs coming online, and the opportunities in biology seem very real, but it may take years to get drugs to market and generate revenue that would then finance them.
We’ll see how long it takes. Montana has an interesting new program for new drugs, which I think is helpful. A really big unlock is that we have a lot of compounds, and people have used AI to figure out that you can use a drug we’ve had for 50 or 60 years for a very rare disease that devastates 10,000 kids a year, and effectively cure that disease.
Yeah.
A friend of mine did this. It’s his story to tell, so I’m not going to tell it, but AI has already saved a lot of lives. I do think the AI industry is at -5 out of 100 from a PR perspective.
100%.
They need—
We’ve got to be more like pharma.
Yeah.
The popular kids.
Yeah, exactly.
They need to start telling these stories.
Yeah.
I think we have a really high-quality group. I picked most of them. You don’t have to be bullish. I love skeptics and bears.
Yeah.
You just have to be polite, nice, and constructive. You can be bearish while being polite, nice, and constructive. We have loads of skeptics and optimists. I think it’s going to be great. I hope every CEO does a hit on your show after they’re on stage, or before.
Yeah.
I’m pretty excited about it.
Yeah. It might be the longest—
—it might be the longest TBPN episode ever.
I think so.
We’ll see.
Yeah.
3. Handling Attention and Criticism
How do you deal with all the attention, positive and negative? Does that bring—
You mean for me?
Yeah, for you. I just feel like you talked about your mindset right at the beginning of the show: it’s a choice to be happy. But when the spotlight shines directly on you once every week, there’s a lot of love, with a little bit of hate sprinkled in there.
No, there are a lot of detractors, let’s say. We can call it what it is. Antonio, who I’m doing the conference with, says being on X is an exercise in cultivated equanimity.
Hmm.
I don’t take any of it personally.
Hmm.
I’m not on X to be right. I’m a responsible optimist, but I’m trying to be open-minded to new negative facts and constantly search them out. I’m really on X to learn. With all of the hate, I try to sift through it for, “Oh, wow, that is actually a very good, compelling bearish argument.”
There was an interesting post from, I forget, a PE analyst or LBO analyst. It said, “Hey, the fact that Anthropic is dedicating 74% of their compute to training”—you could read it one way. On the one hand, it’s bullish because they could be generating a lot more revenue. On the other hand, if all this compute is being dedicated toward training, and for any reason somebody decides you need less compute for training, that’s a lot of supply that could hit the inference market. That’s a really interesting bearish perspective.
I’m human, but I’m here to learn. I want to get to the right answer. I’m sharing my views. If somebody shares a helpful negative perspective, I try to thank them. I need to go thank that person, the LBO analyst, I think it was. I didn’t have time. It’s an exercise in cultivated equanimity. You can’t take it personally.
Yeah.
I am trying to do what little I can. As Gavin Baker, I don’t have millions or tens of millions of followers. But I am trying to help paint a more optimistic vision for AI, doing what I can to help the American public understand that you actually want a data center in your town.
One of my favorite exchanges was somebody saying, “Well, I’ll wait until they start building data centers in the wealthiest county in America.” Loudoun County is the wealthiest county in America. They have the highest per-capita number of data centers.
Yep.
So just understanding that—
Virginia.
—so just—
US East.
Yeah, just under—
We know it on Amazon.
Yeah, US East availability.
We love it. Yep.
So I understand why some people are worried. There’s been a lot of fear. We’ve all seen scary movies. But a lot of awesome things are happening today because of AI for ordinary Americans, for blue-collar Americans, and for these small towns. We’re reindustrializing America. We’re massively increasing our electrical generation capacity, which—you cannot reindustrialize without power. That’s job number one.
A lot of good things are happening, and I’m trying to do what little I can to tell that side of the story and be pretty balanced. Anthropic, I called it like I saw it in the summer. They were losing market share. I’m calling it like I see it in October: they are reaccelerating and gaining share.
The S-1 was an upside surprise to me. I’m trying to call balls and strikes because my job is to just get it right.
Yeah. Jordy mentioned—
Yeah, and it’s funny because people will get so flustered by seeing different posts on X. But imagine how boring X would be if no one ever took a stance. It would just be like, “Well, it might be this. It might be that. And it might be this. But have you considered that?”
Oh, yeah.
Yeah.
Yeah, yeah.
No, no, no.
Every post has to be, “It’s over,” or, “We’re back.” “This is the best thing ever,” or the worst. Then we can have a debate. At least you took a stance.
Yes.
Yeah, it’s great.
Yes.
Jordy mentioned getting sharp on bio and starting to think about opportunities there. How deep have you gone in the AI stack? I saw a funny post that was like, you could spend all your time thinking about OpenAI, Anthropic, and Google, and then meet all the neoclouds and meet Lynn and Tuan and figure out what’s going on in the private markets. Or you could just buy copper and sit there and hold it, because it’s almost a perfect proxy or something like that.
How far have you gone down into energy infrastructure and commodities? How sharp are you on those? How relevant or interesting are those markets to you in this world, since they do seem adjacent?
I pay close attention to all of those.
Yeah.
I ran a diversified fund in a prior life for nine years.
Yeah.
I was an energy investor.
Yeah.
Natural resources, too.
Yeah.
For many years, I would do marketing meetings with the gold analyst at my prior employer, Fidelity. I always knew about the merits of gold from a portfolio construction perspective and would always have it in the fund.
By the way, just because it triggers people, I want to say it: I find it so amusing how upset gold bulls get at the idea of asteroid mining. No one has ever been—I’ve never gotten more hate than from the gold crowd about the idea of asteroid mining and the asteroid cycle.
Increasing supply.
I mean, it—
That’s so funny.
I mean, talk about a challenge to your worldview.
Wait, are they mad about that new Alchemy startup?
Yeah, but it’s one of those things you have to imagine: you need to imagine technological disruption. Look at the diamond industry.
Yeah.
Right?
Totally.
Yes, absolutely.
If you understood that technological disruption was possible, you probably built labs 20 years ago.
You—
Right? Ten years ago.
Maybe you would’ve done something different with the diamonds in your family. But as far as energy, industrial infrastructure, and natural resources, we have spent a lot of time on those. We have not made them a big part of the portfolio. We’ve done some things in private markets to try to help with rare earths.
Mm-hmm.
This is not through acquiring mines.
Yeah.
It’s more that the problem with rare earths isn’t that they’re rare. It’s that refining them is a very dirty process. We’ve invested in a company that has a chemical process for refining rare earths. I actually think DARPA has done some pretty smart stuff.
They kind of have a lot of different efforts around this. I try to pay as close attention to all of it as I can because, as an investor, you never know where a weak but important signal is going to come from. You want to cast as wide a net as possible, but then also focus a portfolio on the areas where I feel like we have the biggest competitive advantage as a firm.
4. Who Finances the AI Buildout
How are you thinking about the evolution of the marginal capital decider potentially changing in 2027? If you go back a couple of years, you have seed-stage VCs making bets on these AI labs, and then eventually more scaled-up venture firms becoming important capital partners. Some have taken board seats, but they are not deciding the race now. It’s big hedge funds and pre-IPO rounds. We’re coming up to a world where the Fidelities of the world, insurance agencies and funds, and sovereign wealth funds are going to be increasingly involved in financing the build-out. What does the next year look like in terms of how we get the big dollars into the AI build-out?
I think you named all the big dollar sources. Something that I think is pretty underappreciated is American mutual funds. You’re limited in the percentage you can hold in a liquid asset, so SpaceX and Anthropic going public is actually going to free up a huge amount of buying power in private markets once those are fully unlocked, because they have to be public and unlocked.
I think you will see mutual funds. They’ve always been involved. My former colleagues at Capital Research, T. Rowe Price, MSAM, GSAM, and Franklin Templeton—I’m sorry if I forgot anyone. I love you all. They’ve been pretty involved. The sovereigns have been very involved, particularly the ones in the GCC and the UAE.
Yeah.
Like MGX—
Mubadala.
Mubadala, all the sovereigns there. Atreides and Valor co-led Crusoe with Mubadala. I think the UAE in particular has been extremely front-footed, with all their sovereigns, with respect to AI.
And now Apollo’s taking a stake in Mubadala, which sort of completes the cycle.
Yes. Yeah. But that’s going to have to evolve. I do think we’re probably getting to a point where these clouds’ underpriced contracts—signed well below market values—reprice higher as they bring on more compute with more spot, like we discussed with CoreWeave. I think you’re going to see that there was a pretty big acceleration in the clouds last quarter.
Forget relative to expectations or this or that, but I think you’re going to see a pretty continuous acceleration in the combined cloud revenue growth of the big hyperscalers and the neoclouds. That number is just going to continue to accelerate. It’s dropping down at very high margins, so operating cash flow is following. I think you’re going to be able to finance more of the 2027 build-out out of cash flow than people think.
Hmm.
I don’t know if we’ve seen the peak of external financing, but I think we’re getting close.
Hmm. Yeah. That’s—
And that’ll be really helpful.
Totally. Yeah. No one can complain about that. That’s investing in the future.
AI skeptics will be—
In a bad place.
—in a horrible place.
So I need to check on them.
We’re going to have to check on them.
Sure. Sure.
You’re going to have to go around.
And look, maybe I’m wrong, but I will change my mind when the facts change. I’m just going to try and keep calling balls and strikes, and then try to be an advocate to help ordinary Americans understand that this technology can and should be awesome for them.
Yeah. Yeah.
I’m very interested to see how Anthropic and OpenAI becoming public companies impacts overall sentiment, specifically on X, because right now such a small number of people actually hold shares in these companies.
Oh, sure.
Maybe there are 50,000 people who have some incentive to be relatively optimistic. You should be able to see both the bull case and the bear case for any company. But when that changes and millions of people globally can suddenly have exposure, how does sentiment change? I think people are going to be much less extreme once the middle-of-the-road investor has some exposure.
Oh, sure.
It’s not totally binary, where a very small number of people have exposure. And it’s direct exposure, because if you happen to own shares—
Yeah, Google.
—through some retirement fund or through a hyperscaler, it’s not the same.
Maybe some of them will do what Elon did. Every child in America is going to have a share of SpaceX.
Sure.
That’s awesome. I think that’s way better for America than setting up some foundation where a board or a bunch of bureaucrats decide how to allocate the money. Just give people the money.
Yeah.
Just let people make individual decisions. I would encourage more companies and more wealthy people in America to do what Elon and Michael Dell and, of course, my friend Brad Kirshner have done. That’s awesome for America and great for capitalism.
Yeah. It feels so important that the infrastructure got built before the big liquidity wave, before SpaceX goes public, so that it’s there, it’s been tested, and it’s been implemented and rolled out. Anyone who wants to participate can take advantage of it without needing to say, “It’d be nice if the government had a program that would let me give shares directly to the American populace.” It’s there. It works. It’s up to date. It’s in the App Store. You can go download it. I love it.
Just give people the money.
Yeah. Simple.
Give everybody a stake in capitalism.
Yeah. I completely agree. That’s great. Thank you so much, guys.
Thank you.