[BidClub_]
iConnections · TomorrowX Summit · · 24 min

Brad Gerstner: Why Anthropic's IPO Is 10x Oversubscribed | TomorrowX Summit

Brad GerstnerRon Biscardi

EquitiesVC/PEAI & SoftwareInvesting
YouTube ↗
TL;DR
  • Gerstner pitches Anthropic’s 10x-oversubscribed IPO as access to “the largest TAM in the history of technology.” His $1 trillion revenue scenario requires about 4% of global knowledge work. He expects his fund’s biggest-ever order despite a semiconductor index he says has fallen 60% and election-driven fears of data-center restrictions.
  • Gerstner now sees a path to model profitability that he previously could not establish. He says he didn’t know how to answer Bill Gurley’s compute-cost objection. Now he cites Anthropic approaching a $100 billion revenue run rate with 5,000 employees, versus Google’s 105,000 at that scale. He says compute costs are falling 90% annually while greater capability creates pricing upside.
  • He argues customer-serving compute needs are smaller than many fear. In his first-nine-months-of-2026 figures, serving customers used 26% of Anthropic’s 2GW compute budget, about 500MW; he says another $100 billion of forecast revenue next year needs only 500MW. Training uses most capacity because the labs believe they are in “the foothills of the singularity of superintelligence.”
  • Biscardi’s IPO pushback is that strong demand has not protected investors from losses. He says the year’s top ten IPOs are, on average, about 20% below their issue prices. Gerstner expects volatility, advises sizing positions accordingly, and says he would hedge part of his hedge fund’s Anthropic exposure if it overshoots; he may leave other portfolios untouched.
  • Gerstner expects most AI value to accrue to the top one, two or three players. His model-maximalist view, which he says he has held for two and a half years, runs against the consensus he describes: that 20 models and open source will commoditize the model layer. His portfolio rule follows: “Do not own things that get worse as intelligence gets better.”
  • Gerstner wants Nvidia’s board to establish the facts before deciding Huang’s future. He says he knows only what has been reported: personal transgressions appear to have occurred and been acknowledged, while other claims remain speculative. He warns against political pressure and a rushed decision. “Jensen Huang is not above the law”; if Huang is cleared, Gerstner hopes critics apologize.
  • Invest America’s purpose, in Gerstner’s telling, is to spread ownership as AI raises fears about jobs. With a $1,000 birth contribution matched and $10 added weekly, he projects $50,000 at 18, $175,000 at 30 and $1 million at 55. He casts universal capital accounts as the AI era’s counterpart to Social Security, aiming for “more independence.”
Digest · the substance, structured for research

1. Anthropic’s compute economics anchor the IPO case

  • Gerstner expects the 10x-oversubscribed offering to draw his fund’s biggest-ever order despite a backdrop he describes as a 60% semiconductor-index decline and fears that data centers may be shut down. He calls the addressable market “every single consumer and every single business” and says $1 trillion of revenue requires about 4% of global knowledge work.
  • Bill Gurley had argued that compute spending would prevent profits. Gerstner says he did not know the answer. Now Anthropic is approaching a $100 billion revenue run rate with 5,000 employees, versus Google’s 105,000 at that scale. Gerstner says compute costs are falling 90% a year while greater capability gives the company room to charge more. He also points to revenue per unit of compute, or per watt, and the amount of compute needed to serve customers.
  • In his figures for the first nine months of 2026, serving customers used 26% of Anthropic’s 2GW compute budget, about 500MW. He argues another $100 billion of forecast revenue next year needs only 500MW. Most capacity goes to training; he says the labs want to reach “the foothills of the singularity of superintelligence.”

2. IPO demand does not eliminate drawdowns

  • Biscardi says the year’s top ten IPOs are, on average, about 20% below their issuance prices. Gerstner expects Anthropic to be volatile: he says new companies face unknowns, supply coming to market, and a contest between bulls and bears. He recalls that after Google’s 2004 IPO, his position was sometimes down 10–15% in a day in 2005. His advice: “Size your position accordingly”; investors without the stomach for volatility can wait.
  • The stated timetable is pricing “tomorrow night,” then trading Thursday around noon or 1 p.m. Gerstner says he would hedge part of his hedge fund’s exposure if Anthropic overshoots, while he may leave his private, personal and long-only holdings untouched.

3. Better intelligence must improve the portfolio

  • Biscardi asks how venture investors can choose decade-long bets when the market changes monthly. Gerstner says to identify where a supercycle will create or destroy economic value, then focus: “If it’s not a fast yes, it’s a quick no.” He recalls that in 1999 his search-and-commerce framework helped narrow the field. In 2012, when many investors thought cloud software was less secure, more expensive and unable to scale, his firm took the opposite view; that fund returned over 50x and included Snowflake, MongoDB, Twilio and Okta.
  • Gerstner says his AI view has been “orthogonal to the market” for two and a half years. He describes the consensus, including Jensen Huang and Bill Gurley, as expecting 20 models, open source and a commoditized model layer. Gerstner calls himself a “model maximalist”: he expects scale and network effects to concentrate most value in the top one, two or three players, and says AI may be a more natural monopoly than earlier technology waves.
  • He wants investments to benefit from improvements in general intelligence. He gives Harvey and Legora as examples of vertical applications that could be threatened if better intelligence makes their “harness” unnecessary. His rule: “Do not own things that get worse as intelligence gets better.”

4. Founder loyalty does not settle Huang’s case

  • Discussing the letter he and Gavin sent to the board, Gerstner says he favors founder leadership but does not know the facts. He says personal transgressions appear to have occurred and been acknowledged, while other reporting is speculative. He worries political pressure could push the board to act before it gathers the facts.
  • “Jensen Huang is not above the law,” Gerstner says. He wants Huang to answer questions and be held to account. The board should explain a decision to remove him if that is where the facts lead; if it concludes he can return, “so be it.” Gerstner hopes critics who rushed to judgment apologize if Huang is cleared.

5. Capital accounts are Gerstner’s answer to AI anxiety

  • Gerstner says fewer than half of Americans under 40 believe in capitalism and 60–70% feel left out or left behind. He argues that wider ownership could help address that disaffection and preserve the promise of capitalism: “more independence,” not more dependency.
  • His proposal gives every newborn $1,000 invested in the S&P 500, with contributions open to states, philanthropists, parents and companies. With the birth amount matched and $10 added weekly, he projects $50,000 at 18, $175,000 at 30 and $1 million at 55.
  • He says 10,000 accounts are created daily for newborns. He credits Michael and Susan Dell with helping make the initiative possible, and says Elon offered a share of SpaceX for each of 70 million accounts, valuing the pledge at $11.5 billion and calling it the first of many.
  • Gerstner casts the Invest America Act of 2025 as the AI era’s counterpart to Social Security in 1935. He connects both to fears of technological unemployment and says capital accounts give the country options for managing the AI transition. He also urges people to contribute locally, including small amounts to friends’ accounts.

Verification Notes

  • The transcript’s account figures are unclear in scope: Gerstner says, “We have 70 million accounts that have been opened in New York City alone, 2 million accounts funded with $450 million.” The digest preserves the separate 70 million figure in his SpaceX pledge claim but does not resolve the account-count statement.
Full transcript
Brad Gerstner

Let’s go. I’m doing great. It’s good to be here.

Ron Biscardi

Where are we? Who plays here? I heard it’s a pretty good basketball team.

Brad Gerstner

Oh, great. Great to be here.

1. Invest America goes bipartisan

Ron Biscardi

You too. Let’s go. Good. We have 20 minutes. Chop, chop. All right, let’s get started. We’re going to circle back to your Invest America and Trump Account initiative at the end, because we obviously have a big IPO happening this week. Wasn’t that pretty amazing, though? You and I talked about it a long time ago, and obviously people are caught up in the politics of it. Across the country now, all the presidential candidates on the Democratic side have endorsed it. Former Presidents Clinton and Obama have endorsed it. People are really starting to get behind it.

2. A trillion-dollar IPO

Brad Gerstner

I love that we have something that can be a bipartisan movement. A little-known fact is that when I had this idea in 2020, one of the first people I called was Antonio Gracias. I said, “We need to do something to save capitalism.” He said, “I’ll send you a million bucks. Where do I wire it?” He was one of the first people to really get behind it. So thanks to—

Ron Biscardi

He is. Yeah. Yes. All right, talk about an IPO. I heard there’s an IPO this week.

Brad Gerstner

There is. It’s the second IPO of the year where the valuation is going to be over $1 trillion.

Ron Biscardi

Tell us what’s going on with this company. I’m a kid from the dot-com era. I remember when we were going through the IPO craze in the ’90s, but those companies didn’t have revenue, profits, or customers. This is a completely different story.

Brad Gerstner

I don’t know if they can give me control over some slides backstage. I love to show slides. You kind of took me to the end of a few slides, but let’s go to some slides on Anthropic, because this one is pretty interesting. It shows the trajectory of Anthropic and OpenAI relative to Google and Meta. Obviously, something very different is happening with these 2 companies. If you look at the amount of TAM we need to penetrate to get to $1 trillion of revenue, it’s about 4% of global knowledge work.

3. Why the Anthropic IPO is 10x oversubscribed

Let me get to the specific slides on Anthropic. Why is the Anthropic IPO 10x oversubscribed? The semiconductor index is down 60% over the course of the last few months. We just had an election that’s scaring everybody that we’re going to shut down data centers, and we had Jensen Huang pushed out as CEO of Nvidia last week. If you had told me that was the backdrop going into the Anthropic IPO, I would have said, “Maybe they pulled the IPO.”

Notwithstanding that, it’ll be our biggest order, and I’m sure it’ll be the biggest order in the history of a lot of other investors’ funds. So why is that? The TAM is amazing. It’s the largest TAM in the history of technology: every single consumer and every single business on the planet. It’s diffusing faster than any technology in history, or any technology in the history of Silicon Valley.

4. This will never be profitable

But there are a few things that make this particularly compelling. My podcast partner, Bill Gurley, and I have wrestled with this for the last 3 years. Bill said, “Brad, this will never be profitable. They have to spend too much on compute.” I didn’t know the answer. That’s just the truth. But here we are, on the verge of a $100 billion revenue run rate. The company got to a $100 billion revenue run rate with 5,000 people. When Google got to a $100 billion revenue run rate, they had 105,000 people. So if you take out 100,000 people and all the associated costs, it turns out you can spend a lot on compute.

Then they drove massive compute efficiency. The cost of compute is coming down by 90% a year, while the amount they charge for compute has a lot of upside because it’s becoming more capable. So that’s the first thing. The second thing is revenue dollars per unit of compute, or per watt. The third thing is the amount of compute they had to use to serve us as customers.

5. Where Anthropic's compute really goes

On the far right, this is the first 9 months of 2026. Everybody thinks, “Oh my God, they’ve got to spend so much money to serve Claude and run these agents.” Only 26% of their compute budget this year went to generating $100 billion in revenue. That’s 500 megawatts. They had 2 gigawatts of compute this year, and only 25% of it went to serving all this up. The vast majority goes to training, which says to me, “Wow, next year the consensus forecast is another $100 billion of revenue. They only need 500 megawatts.”

When you hear all this concern about data centers, they just don’t need that much compute to hit their numbers for next year. Of course, they want a lot more. Why do they want a lot more? Because they know we’re in the foothills of the singularity of superintelligence. They want to be the first ones to get there. It takes a lot of training to get there, but this shows why it’s such a good business model.

6. Can one IPO revive the IPO market?

Ron Biscardi

Awesome. The IPO is scheduled, I believe, for tomorrow. I guess it’ll price tomorrow night and allocate tomorrow night. It’ll start trading Thursday around 12 or 1 p.m. This has been the best year for IPOs in a decade. We went through this horrible drought, and SpaceX kind of breaks everything open. But as we stand here today, I believe the top 10 IPOs are, on average, down about 20% from their issuance price. Is this going to be the thing that rejuvenates the IPO market, or do you think this is just going to be the last one in that chain?

Brad Gerstner

First, I imagine a lot of folks in here have seen the Twitter slide that’s gone around, showing that almost all IPOs, within 12 months of going public, have volatility that takes them from their peak. I don’t know what the peak was for SpaceX, maybe $200 or thereabouts, but from the peak to the trough in the 12 months after the peak, it’s down 50%. Why is that? There’s a lot of volatility and a lot of unknown unknowns about new companies. There’s a wrestling match between the bulls and the bears. Obviously, you have a lot of supply coming to market.

7. Volatility is my friend

I was involved in the Google IPO in 2004. In 2005, my partner Paul Reader would come into the office when it was my largest position in the portfolio at the time, and it would be down 10% to 15% a day. He would be panicking: “What are we going to do?” The reality is that in the early phases of an IPO, things are going to be volatile. Size your position accordingly. If you don’t have the stomach for the volatility, don’t buy it. Wait.

I think Anthropic is going to be volatile. No doubt about it. As an investor, we do early-stage venture capital, late-stage growth, and public markets. In the public markets, we have a long-only fund, which is kind of buy and hold, and then we have a hedge fund. My mandate in the hedge fund is to make more on the way up and lose less on the way down. If Anthropic overshoots, I’m going to hedge part of that exposure, because that’s my job. That’s my promise to my investors. I may not touch it in my private funds, my personal funds, or the long-only fund. The job is to manage that.

8. Has AI changed venture?

I expect there will be volatility. As an investor, volatility is my friend, because I put in the 10,000 hours to develop the conviction other people won’t have. Ninety-eight percent of investors are trend followers. When things are going down, they sell, and when things are going up, they buy. That’s a hard way to make money. Warren Buffett is not a trend follower. First principles: do your own work. Develop your own conviction. You ought to be doing the opposite of the trend. It’s just hard to do.

Ron Biscardi

Yeah. Can we talk a little bit about the venture market? One of the things I find interesting when I think about venture is that it’s a 10-year game. You’re trying to pick things that will become amazing businesses way out into the future. This market is changing monthly. Especially at the application layer, how do you pick anything with confidence when 6 months or a year later, the whole world looks completely different? Do you think AI has really changed the venture game?

Brad Gerstner

Yes and no. I’ll give you why I believe that. In venture investing generally, you have to develop a framework. You have to look at the supercycle and say, “Where is economic value going to be created, and where is economic value going to be disrupted or destroyed?”

I started in 1999 helping to start General Catalyst, and we sat down and said, “Where is all the money going to go? We have this incredible new force. It unlocks massive knowledge and convenience for humanity. Who are going to be the tollkeepers of knowledge and convenience? All of search.” So we got really excited about search: vertical search and horizontal search. Then you had to figure out who you thought the winners in search would be. We said commerce. So who are going to be the winners in commerce?

Once you had that North Star—search and commerce—the other 98% of bullshit that walked into your office were quick no’s. Our business philosophy is driven by the book Essentialism. You all should check it out. If it’s not a fast yes, it’s a quick no. You just can’t focus on everything. You can’t boil the ocean. So in every supercycle, I think you can develop that mental framework. For us, the—

9. The 2012 cloud bet

In 2012, the world was really nervous about the cloud. They said it was less secure, more expensive, and didn’t scale. We had the exact opposite perspective in that early fund we did in 2012. It ended up being over a 50x venture fund. We had Snowflake, MongoDB, Twilio, and Okta.

Why did we do those deals? We weren’t that well known in 2012. Anybody could have done them. But 98% of people had the view that the cloud was worse than on-prem software. Those differentiated views that you develop with conviction, you execute with conviction. That’s where all the money is made.

10. Why Brad is a model maximalist

In AI, I’ve had a view for two and a half years that I think is orthogonal to the market. The consensus view, largely held by Jensen, with Gurley in this camp too, is that there are going to be 20 models, they’re all going to whack up the market, there’s going to be a lot of open source, and the model layer is going to be largely commoditized. That value would accrue in other parts of the stack.

I had a view that the first three and a half years would all be about AI infrastructure, but then it would be about intelligence. I was a model maximalist. I was like, “I don’t see that happening. This is a scale game, a network effects game, and I think it’s going to play out just as the internet did, just as mobile did, just as social did, and just as hyperscalers did, where most economic value is going to accrue to the top one, two, or three players.”

I might argue that this will naturally be more concentrated, a more natural monopoly than the other ones because of the nature of the market. If I look at the evidence on the field, it leads me to have higher conviction today than I’ve ever had before that we’re going to have most of the economic value captured by a couple of players.

11. How big is the prize?

In venture, I don’t think, “Oh, 10 years from now.” If one of my guys comes in and says, “Don’t worry about it, Brad. It’s going to be shit for the next four years, but it’ll be good in 10,” I’m not investing in that. It just doesn’t work that way. When they walk in, they know I have a couple of questions. If everything you say is true, how big is the prize? What is the size of the prize? If everything you say is true, and if they execute and achieve all the goals, if it’s not a really big prize, it’s not a venture-backed bet.

12. Own what gets better as AI gets better

Number two is, how is it consistent with the framework that we have? Everything that we own must be consistent with this framework. I want to wake up every morning and have everything we own positively levered, positively correlated to improvements in general intelligence. If I wake up in the morning and read that Anthropic has a great new model, or OpenAI has a great new model, I better feel great about that investment.

If I wake up in the morning and feel shitty about it, I’m like, “Oh my God, man, I own Harvey or Legora, and this intelligence just got so much better. Maybe they can take out the harness and we don’t need vertical applications after all.” That’s bad. Don’t own things that get worse as intelligence gets better. Own the shit that gets better as intelligence gets better. It’s the easiest bet we can make at this moment in time.

13. Jensen Huang and the Nvidia board

Ron Biscardi

That makes sense. I feel like we can’t finish this without talking about your friend Jensen.

Brad Gerstner

Yeah.

Ron Biscardi

You and Gavin wrote a letter to the board that went out, and I believe you got a response. Do you want to chat about the response?

Brad Gerstner

It was an acknowledgment that we sent a letter. I’ve been doing this for 26 years. I’m not an agitator when it comes to boards. We’re a deeply founder-friendly firm. We believe founders have a unique ability to lead their companies. I always get concerned when technology companies transition from founders to corporate types. I went to Harvard Business School. I like my friends who went to Harvard Business School. I just don’t want them running most of the companies I’m invested in.

I do have a strong bias toward that. But let’s be very clear, particularly at this moment in time: I don’t know any of the facts. I know what everybody in here has read, which is that he’s obviously made some personal transgressions, or appears to have, that he’s acknowledged. But everything else, I find, is a lot of speculation, a lot of reporting that we just don’t know the answer to.

I was quite concerned. I’ve seen a lot of boards, public boards, over a long period of time. There are some good ones, but there are a lot of shitty ones that make a lot of bad decisions. When you put the political pressure of the United States of America, the president saying, “Hire Tim Cook,” and Scott Bessent going on the Sunday show saying, “You have 10 days to get us an answer,” all of a sudden it just felt like this massive rush to judgment.

If I were on that board, I can see how all of them would feel pressured to do something they don’t have to do. They owe a duty and a responsibility to the stakeholders of Nvidia not to rush to judgment. Take your time, gather all the facts, and then make a call. All I said is, it’s your prerogative, but that’s what you need to do. If it leads you to the conclusion that Jensen Huang can’t be the CEO of the company, explain yourself and make that decision. If it leads you to the conclusion he can come back, so be it. But no rush to judgment. Don’t get politically pressured.

Some people interpreted that as, “A tech bro billionaire thinks everybody’s above the law.” Let me be very clear: Jensen Huang is not above the law. There is no founder running any company who is above the law. I’m on Team America. That’s the team I’m on. He needs to stand up, answer the questions, and be held to account. But if he’s cleared, I certainly hope some of these people who have been pounding the table, rushing to judgment, and saying all the shit they’ve been saying the last week step up with some apologies.

14. Invest America update

Ron Biscardi

Yes. Well said. Okay, so we started this session with a video about Invest America. The theme of this event is how AI can and should revitalize the American dream. You were ahead of us on this by several years, because I think there’s no better way to reinvigorate and revitalize that than by giving the children of America access to the product of capitalism by way of these accounts. Why don’t you give us an update on these accounts?

Brad Gerstner

I mean, Ron, you had me on stage just after, I think, Jared Kushner, maybe three years ago in Miami, four years ago in Miami. And it was a huge boost to our efforts. As the founder of Invest America, any founders in the room know this: you have to fake it until you make it. You have to create your own momentum every single day.

People said, “How did you do this?” We didn’t hire a lobbyist. We didn’t hire any lawyers. I was the founder with the founder mission, and I woke up every day and I worked 1,000 hours because I think it’s deeply critical, if we’re going to preserve the promise of the American dream and preserve the promise of American capitalism for another 250 years, that everybody has to feel like they’re in the game.

You can’t have a country where less than half of people under the age of 40 believe in capitalism, where 60 to 70% of people are disaffected, feel left out and left behind. That’s deeply destabilizing. What’s going to happen in that scenario? If anybody’s watched the fires in France, the outrage in Spain, that’s a preview of where you go if you don’t get everybody into the game.

The great news is America is a self-righting ship. We’re not Europe. Europe has snatched defeat from the jaws of victory. There is no innovation. There is no growth. As a result, the gap between them and the rest of the world will grow dramatically.

15. 1,000 at birth, $1 million by 55

Five years ago, I said we have to get ahead of this. I was very focused on the wealth gap. How do we attack it at scale? I don’t want to mess around with one city or one school. If we had a moonshot, if I was being Elon in this moment, what would I dream of? I would dream that at birth, every child gets $1,000 invested in the S&P 500 for life, that anybody can add money to: states, philanthropists, moms and dads, corporations, et cetera.

That would compound. A thousand bucks matched at birth, and you had $10 a week. I was thinking Apple Pay on an app. My son designed the first app. If you just had 10 bucks a week, that’s $50,000 at age 18, $175,000 at 30, and a million bucks at 55.

16. It sucks being poor

I grew up poor in rural Indiana, in a town of 3,200 people. I watched my mom work two jobs. She’s 90 years old today, and she is a saint. But it sucks being poor and figuring out how to go from zero to one is a desperate feeling. I watched that desperation. I lived that desperation. I don’t need people on the left coast and the right coast to lecture me about what it feels like, particularly the ones who went to boarding school and Yale. And yet they say, “You’ve got to just give all the money to government.” No ownership, empowerment, not more dependency, more independence. This is title and ownership for every single person.

Today in America, 10,000 kids will be born, and 10,000 new accounts will automatically be created with $1,000 in the S&P 500. Tomorrow, 10,000 kids will be born, and 10,000 new accounts will be created. We have 70 million accounts that have been opened in New York City alone, 2 million accounts funded with $450 million. We talked about it as a dream. The dream has now occurred.

I have the best partner in the world, and hopefully we’ll chat with them a little bit later. Michael Dell, he and Susan stepped up in such an extraordinary way. It’s really a refounding moment for Invest America, showing the art of the possible to the country.

From coast to coast, we now have philanthropists adopting ZIP codes. Elon came out and said, “I’ll add a share of SpaceX to every one of the 70 million accounts.” That’s $11.5 billion. He said it’s the first of many.

17. Social Security for the age of AI

It’s important to understand that in 1935, we passed Social Security. We had 22% unemployment in the country. Europe was falling into the grips of fascism, and the U.S. was going to be a fast follow. Imagine the fear people would feel today if we had 22% unemployment, like we did in 1934. Social Security saved the country because it said to people, “Your asset is your labor. We’ll insure your asset if you get unemployed. We’re going to give you a working wage so your family can survive.”

My mother grew up without running water and couldn’t go to college. Her father was a recipient of that wage from the government. It really kept the country together until we got through World War II. Then, obviously, we got on this growth path.

I think the Invest America Act of 2025 will be to the age of AI what universal capital accounts were to the industrial age in 1935. We’ll look back at this and say that was to the age of AI what Social Security was to the industrial age.

Remember, it was Keynes in 1930, on the eve of the U.S. breaking apart, who said the U.S. was afflicted with an ailment—one that we would talk about in the future. But the phrase you may have heard is “technological unemployment”: people fearful that the machines of the industrial age would take their jobs.

18. How everyone can help

I hope we’re way ahead of this. We’ve given the country the optionality to manage through the age of AI, but it’s up to all of us, in whatever ways we can. Our Little Leagues, our churches, our synagogues, our schools. Go back to your local town and find folks you can contribute to. Maybe just a friend, Ron: you can go up to them, pull up your QR code, and zap $100 or $200 into their accounts, or $10 or $20 into their accounts. Get them excited about capitalism. We all have to win this together. Thank you.

Ron Biscardi

Thank you, as always. I love what you’re doing and think about that. You’re doing a great thing for the country.

Brad Gerstner

Thank you.