Anthropic S-1震撼披露：史上最伟大的生意？Google市值蒸发5000亿美元，中期选举预测

All right, everybody. Welcome back to what is, of course, the number one podcast in the world, tweaking all the mid podcasters out there, four or five years running the All In podcast. It’s episode 295, and Sacks is back. Sacks is back. Sacks is back. How are you doing, brother?

Good, good. I saw some fake news that—

Oh.

—you guys had stopped using the term “superintelligence.” You’re getting some shade for that.

Oh. Listen, we’re here for This Week in Superintelligence right now. I mean, the whole docket is superintelligence.

Did you see that article?

I did.

It was kind of ridiculous.

Oh, are they—

And maybe I shouldn’t Streisand it, but they were trying to say that we were using the term AI still instead of SI.

Who is saying that?

This is literally—

It was a real article.

What is the press spending its time on?

I think it was Semafor. Now, this is why it’s such fake news.

Fake news.

They have a photo of me. I wasn’t even on the October 23rd episode.

You do look great, though. You look very presidential there. Man, if only you were born here.

But look at this fake news.

I mean, they’re desperate for—

“The U.S.’s most influential tech podcast.” Well, that’s good. But actually, we’re not just a tech podcast. We’re the number one podcast.

We’re the number one podcast in the world. Give us our due.

“Has cooled on the use of the term superintelligence,” and then they try to say that I somehow was defying—

Wait, wait. They tried—

—the president’s instructions.

All right.

But I want to clarify: I was not on the October 23rd episode.

Gotcha, gotcha. Well, listen—

I’ll be keeping you guys in line on this from now on.

Absolutely. We will get focused on superintelligence. With us again this week, your favorite organic farmer, David Friedberg. How’s Ohalo doing? You’ve been pretty busy yourself.

Ohalo in Kenya this week. Lots of excitement.

Mm.

Lines of farmers trying out true potato seed.

Your energy looks pretty good. You look like you’ve got some good energy. Speaking of good energy, man, we had a fun time playing cards last week. That was a blast. A great dinner.

You didn’t stop talking for 7 hours.

Wow.

Sort of like the podcast.

Yeah, just like the podcast.

It’s unbelievable. It’s unbelievable.

We had a fun time.

You just kept—

Come on. That was a lot of laughs.

—you just kept talking. Affleck texted me after: “I never want J. Cal in the game again.”

You’ve got to get Affleck out as a member of the group?

It’s going to be Ben Affleck or you, and we pick him, so you’re out.

That’s fine with me.

You’re never getting another invite.

That’s fine with me. I mean, it’s an upgrade. It’s absolutely an upgrade.

I mean, you just kept chirping for 8 hours straight. It’s unbelievable. He could not stop talking.

I haven’t been at the game for a long time. I had a lot of material I’d stored up.

It was really fun, though. It was really fun.

We had so much fun. And Chamath Palihapitiya, everybody’s favorite enterprise sales executive from the 1890 Corporation. How’d you do this week? Did you hit the gong? Did you hit your quota, or are you going to have to work this weekend?

Q4 is off to a good start, but I never want to count the chickens.

Okay. That’s good, considering you whiffed Q3.

No, no, no. Q3 was okay. We whiffed Q2.

Mm.

Butchered it. Butchered it.

Butchered it. Wow.

Butchered it.

That’s right when I invested.

All my fault, too, because I forced these poor salespeople—

Yeah, no, you top-ticked it.

I forced these 3 poor salespeople to only focus on the gigantic deals.

You’re learning. You’re learning.

And—

Yeah, you—

I had to learn the hard way. I had to learn the hard way.

You want to get those singles and doubles, and the people who pay on time and start the projects on time.

We’ll see in Anthropic’s S-1, but you can build a ginormous business with singles and doubles.

Absolutely you can. Speaking of gigantic businesses, Anthropic dropped its S-1. That’s the document that outlines the nature of the business for an IPO, an initial public offering. And I’ve got to ask the question: This seems to be the greatest business in the history of entrepreneurship. Goldman and Morgan Stanley are going to lead the round, and they’re targeting November 16th, right before Thanksgiving, for this IPO. We’ve been waiting for a long time. It’s been delayed. The ticker will be ANT, A-N-T, and they’re aiming to raise—wait for it—$100 billion. In other words, about 3 months of compute. That would make it the largest IPO ever, surpassing SpaceX. They raised $85 billion back in June, if you remember. And Polymarket, our great partners at Polymarket, say there’s a 78% chance Anthropic raises at least $95 billion. The valuation is $1.5 trillion to $2 trillion. Here’s Polymarket: Right now, it’s saying between $1.75 trillion and $2 trillion. Market cap, 31%. So, yeah, assume it’s going to be slightly under $2 trillion.

The revenue is the story, Sacks. $16.3 billion in the first half of 2026. That is 14 times year over year. In other words, the first 6 months of 2025 were one-fourteenth of what they did in the first 6 months of this year. Q2 is going to be a whopping $11.6 billion, a 30-times jump over the last 5 quarters, and almost as much as they made in the first half of the year. They say the TAM, the total addressable market, for their tokens and services is $40 trillion. That’s extraordinary. I don’t know how they backed into that, but we’ll talk about it. They’re projecting $100 billion of ARR by the end of the year. It’s unspecified whether that’s annual recurring revenue or their annual revenue run rate, which is ARRR—ARRR, like a pirate. They’ve got $112 billion in cash and a very unique ownership structure. Seven founders, yes, they have 7, are going to hold 50.1% of the voting power through a single class of F shares. Sacks, is this the greatest business in the history of capitalism?

It could very well be. I think this is probably going to be the biggest venture outcome of all time, especially given the time in which this was created. This is only a 6-year-old company. I think the exit here is going to be on par with SpaceX’s IPO, but SpaceX has been around for 20-something years. So this is probably the fastest value creation of any venture-backed company in history.

I don’t know if you saw, but there was a tweet this morning about the distribution of wealth this is going to create among the employees. Hundreds and hundreds of employees are going to be making hundreds of millions of dollars, and thousands of employees are going to be making $10 million-plus. This is going to be an enormous wealth-creation event like San Francisco, and maybe all of Silicon Valley, has never seen before.

There are a few other data points in this S-1 that I think reflect the strength here. Number one is that this $80 billion loss number we’ve heard so much about, which the media has constantly been pumping and harping on as proof that Anthropic was either a fake business or that we were in this whole bubble, basically just got completely refuted. It turns out that the $80 billion loss is a phantom number. It resulted from Anthropic having issued convertible notes to its infrastructure partners, like Amazon and so forth. As Anthropic’s valuation increased, those convertible notes got recorded as a loss on the balance sheet, and that flowed through into the income statement. But it’s not a real loss. If they had just issued equity, that loss wouldn’t be there. It was not a cash loss. It’s basically a fake number.

In fact, what was revealed in this S-1 is that Anthropic has been an extraordinarily capital-efficient business. They’ve raised something like $110 billion, roughly, and they’ve only spent about $9 billion or $10 billion throughout the entire course of the business. They still have, like you said, JK, about $100 billion left in cash or cash equivalents on the balance sheet. Everyone thought, based on the previous reporting, that this was a huge loss-making business. In fact, it ended up being a very capital-efficient business.

The other thing you saw was that their margins have been improving. I think their gross margin is now something like 85%. For a given task or capability, they’re showing roughly a 10-times improvement year over year in their ability to deliver that task—the cost effectiveness of that task. It gets 10 times cheaper year over year, and that allows them to cut the cost of that model or task by, let’s—

Yes.

—call it 50%. The consumer recognizes a huge cost savings, but they’re able to realize an even greater savings because they’re able to capture the rest of that efficiency in their margin. That’s what we basically saw: They’re cutting the cost of their tokens significantly, but their margins are going up even faster because they’re able to produce these models that are more and more efficient. They call them compute multipliers.

It’s very interesting, the way Anthropic sees its business. The framework Dario has is that there’s raw compute, which they’re scaling very rapidly as a result of better and better chips and more and more of those chips. Then there are the compute multipliers, which are things like algorithmic improvements, better curation of the data, or better architectures—more MoE architectures.

Efficiencies.

All these things make the models more efficient.

Yeah.

So they’re able to give you the same answer for 1/10 the cost, or they can take those efficiencies and create greater capabilities. What we’re seeing is that the business is actually improving, and I think that’s a bit of a refutation of all the people who’ve been saying over the past year that they’re getting commoditized away, that open models are putting them on a hamster wheel where they’re going to zero on gross margin, or that it’s a business that only works because it’s heavily subsidized by venture capital.

In fact, I think there are a lot of reasons to believe that Anthropic has a pretty big lead. If you believe recursive self-improvement has kicked in, as they’re claiming it has, then their lead could actually grow.

Friedberg, your thoughts on the impact of this massive deluge of capital coming into Silicon Valley, this corporate structure with 51% ownership, and what impact this might have on the technology industry, the venture capital industry, and democracy here in the United States? We’ve never seen this much capital flow through to a nonprofit organization—or whatever they’re calling this Anthropic nonprofit entity. Eighty percent of the 7 founders’ shares are going into this nonprofit entity of some type, a benefit corporation. Who knows what it’s going to be structured as? It’s optional, but that’s another interesting shoe to drop. Take it whichever way you like, in whichever order you like.

1. The AI Value Stack

Can I just comment on the economic value arising from AI that was perhaps revealed a little bit in this?

Of course. Sure.

In the S-1, Anthropic shows how many customers they have spending $100,000 or more. If we take a look at the image here, it’s a pretty remarkable ramp in customers spending $100,000 or more: 200 customers at the end of 2024, 1,500 at the end of 2025, and, at the end of September, 11,900 customers spending $100,000 or more.

So I ask the question: How many customers are there for whom a $100,000 annual spend is less than 1% of their total annual spend? In the U.S., it’s about 200,000 to 300,000 businesses. Globally, it’s a bit over a million. So if you say, “Hey, look, Anthropic’s customers could spend $100,000 or more and have that be less than 1% of their total spend,” the addressable customer base is about a million customers. They’re only at, call it, 1% of their total customer penetration potential.

Now, is there real money to be made here? I took a look at the AI value stack to try to identify who’s making money where, and it turns out everyone’s making money. I think we’ve talked about this quite a bit, and I know Tomas has some views on this as it relates to the neoclouds.

What is this chart? Is this AI output, or did you make this?

I had Claude make this for me as an image.

Oh, got it.

Or I should say, my analyst made this for me in my shop.

Your bot.

I basically said, “Start with NVIDIA. How much does NVIDIA spend to make the chips, and how much do they sell the chips for?” For every dollar of COGS—every dollar NVIDIA spends—that turns into $4 of revenue for NVIDIA, a 75% margin.

Then that gets sold to SpaceX, and SpaceX turns around and makes an 83% margin on that chip spend because they’re looking at a 5-year depreciation. They spent $10 billion on Colossus 1, with a 5-year life: $2 billion a year in depreciation. That’s the cost per year, but they’re making $1.25 billion a month on Colossus.

Anthropic is showing in this S-1, pretty remarkably, to Sacks’s point, that they’re making about 85% incremental gross margin on every incremental dollar coming in. Obviously, there’s a baseline of compute spend, but 85% is the incremental gross margin. At the end of the day, a dollar that NVIDIA spends turns into $18 that Anthropic makes selling to customers.

The question is, are customers making money? If we’re seeing the customer ramp and the spend ramp that we’re seeing, and customers are making, call it, 2 to 10 times the money they’re spending—otherwise, there would probably be a retention issue and a decline in spend. Maybe I’m wrong. Maybe it’s early days of exuberance and people are losing money, and we could debate that. But generally speaking, if AI is delivering value to customers, I think this shows that there really is true productivity arising from AI. There’s so much margin to be had across the whole value chain. There’s so much value being created all the way through to the customers, and we’re only at 1%—Anthropic is only 1%—of its customer potential, and probably less than 10% of the spend potential per customer.

This is what I think Eric Schmidt used to say a long time ago, in the early days of Google: Google’s total addressable market is global commerce. For the first time ever, we’re seeing a business potentially starting to realize that theory: if you can create enough productivity gains for your customers, your TAM is commerce. It really highlights that maybe we’re almost growing the equivalent of the human working population by 100x, 1,000x, or 1,000,000x with this technology stack.

I just thought it was so striking how early we are, but also how incredibly profitable everyone is. That indicates there’s quite a bit of productivity being produced across the whole value stack. Those were my biggest takeaways from the S-1. I don’t know if you guys agree, but it all felt pretty remarkable to me. I’d be curious to know if you think I’m a little loopy or too exuberant in my reading of this stuff.

I think the thing that’s probably most important is the last box, where you wrote an assumption—

The customer value.

—where you said “assumption.”

Yeah, right. Yeah, totally.

Well, how do they measure that? By the way, I agree with you, but—

The last box I made up. I said, “Assume 2 to 10x.”

Oh.

I have no idea. But I’m a user of Claude. I think we get a lot of value from Claude. I think our people are able to do more because of Claude than they otherwise would, so our business grows faster. We can get more done in a shorter period of time. I tangibly experience that pretty regularly.

I’ve talked a lot about moving to open source, and we have our own cluster installed with our own open-source setup now. But I think what Claude is showing is that it has such a high margin, it actually has good elasticity on pricing. It could push price down, and you could start to see them segment the market a little bit: be more competitive, perhaps with cheaper tokens for cheaper, easier tasks, and charge a higher price for premium tasks that require more intelligence. They could start to stratify the value of the intelligence they’re providing by changing the price points. I envision that’s where everyone’s going to go.

Nick, I’m going to show you something that gets to Friedberg’s point about the potential downstream productivity gains everybody can theoretically have with AI. Put that on the screen. This is from Microsoft’s earnings. My research team put this together, and it’s really quite incredible.

Microsoft’s revenue went from $198 billion to $332 billion in the last 4 years, while staff increased by only 2,000 people. Isn’t that stunning? Before AI, you would have thought that if somebody said, “Great, we’re going to get close to doubling revenues at this kind of scale,” headcount would have doubled, too. The reality is that headcount stayed flat, and we don’t know how many people will take the offer Microsoft just gave, because I think they offered it to as many as 60,000 people. This number could be under 200,000, depending on how many take it.

The scale and the impact on productivity speak to your point. I don’t think the idea that it creates job loss is true. It’s more that you’re seeing productivity gains with the same amount of human input, and I think that’s interesting. I went through the S-1. Do you want me to give you my little S-1 breakdown?

Mm-hmm.

This is the most incredible business I’ve ever seen. I hope they go public after November 11th, because it will force them to update with Q3, and then we’ll get even more insight. This is some contraption I’ve never seen before. I’ve never seen a business like this. It’s really, really, really incredible.

Let’s cross our fingers, because after November 11th, you’ll see Q3 drop. What people have been whispering is that revenue growth moderated to around the 50% range. Q4 is to be determined, but let’s break this down in a bunch of charts, because I think it’s interesting.

Okay, so Nick, start with chart number one, and we’ll just go through them one by one. The first thing I always read in an S-1 is the risk factors. I had my team put together a word cloud.

This is great, Chamath. This is really cool.

First, tell me what they’re afraid of. When you look at it, the share of words in the risk factors allocated to different risks is very important. What this tells you is that the single biggest risk to the Anthropic business is compute. They mention it more than anything else, and specifically within compute, they talk about data centers, power, chips, the grid, electricity, hardware, et cetera.

A close second, but still second, is regulation. They care about government policy, sanctions, and export controls. Then it’s this idea of zero data retention, privacy, and IP. Then it’s AI safety. Then it’s making money. Then there are their competitors, and then governance.

What you notice is what they’re not afraid of. They’re not afraid of customer traction because they rarely mention it. They’re not really afraid of employees leaving, because you’d be an idiot to leave if you were inside. And they don’t really care about taxes or any of these other things. So this is a business that’s very much on the front foot and understands the binding constraint.

If you know that compute is their biggest risk factor, then you double-click into that. I had my team run a further analysis: How do they describe these risks? How many times do they mention them? How many words are they spending talking about this? What you find is that, within the compute problem, their biggest issue is data centers. Then it’s the financing commitments they have, and it’s power. All of these are really important because, again, they help us highlight the binding constraint to this business. As go data centers, so goes power.

2. The Politics of Power

By the way, this is a great setup for politicians, because it’s very obvious as a Democrat reading this—and I’ll speak about it in a second—what you should be trying to do to win in 2028 versus what Republicans have to do in 2028. Basically, the takeaway from all of this is that I think Anthropic is emblematic of the entire AI complex: it’s all about data centers, which means it’s all about power.

If you’re a Democrat, the best thing you want to do is slow all of this down so that there’s a recession going into 2028. The only thing you need to do is slow down power and data centers. On the other side, if you’re a Republican and you want to have the best chance of winning in 2028, this is the one thing that must be done right. You must get the power—

Interesting.

—the electricity and data center thing right.

Okay. So based on this, your premise is they need data centers. If they don’t get data centers, they can’t keep growing. However, Sacks, the truth is these are the most liberal of all liberals in Silicon Valley. It’s the largest amount of money that’s ever been made by founders outside of Elon Musk. It would make sense to me that perhaps the opposite of what Chamath’s saying is true. Shouldn’t the Democrats come in and say, “We are pro-data center and energy because this is going to help society”? They want to see Dario and his group win because they’re going to be the equivalent on the left of Elon Musk on the right. Wouldn’t it be better for them to steal the data center charge and make Anthropic even more profitable, so that 80% of all their money just flows into Democratic super PACs?

Maybe, but I don’t think that’s the Democratic Party we have today. The old Clinton Democrats—Bill Clinton Democrats—are long gone. I don’t think the abundance Democrats are picking up a lot of momentum. The Democratic Party, as Gavin Newsom said, is becoming DSA-dominated.

So what is their agenda with respect to superintelligence? Either to ban it outright, as Bernie Sanders, Elizabeth Warren, and AOC want to do. They basically want 20-year jail sentences for developers of superintelligence, and they want the corporate death penalty for companies that build superintelligence. Or they want to have some extremely draconian regulatory regime that would feature a federal Department of AI—or a federal Department of SI, I guess—that would review and approve every single model before its release. Basically, they want to have a DMV for models.

Let me challenge you—

That is the Democratic agenda on this. Now, could that be shifted through oligarchs like Reid Hoffman or maybe some of these new Anthropic oligarchs, like Dustin Moskovitz? Maybe they can use their massive influence in Democratic circles to effectuate a change. But I tend to think Chamath is right: Democrats want to do anything they can to sabotage the Trump economy, and this whole SI boom is what’s driving the Trump economy right now. So I tend to think that’s going to be the partisan—

Interesting.

—that’s going to be the partisan impetus.

Yeah. So you don’t need to do much, I think, Friedberg, to scuttle the Trump economy, since the war, tariffs, and all that other stuff have driven inflation and all those problems. Wouldn’t it be smarter for the Democrats to—

Well, hold on. I mean, just hold on—

That’s just—

—hold on a second. I mean, look—

—my opinion, yeah.

I think the latest numbers from Fed GDPNow were 3.7% in Q3, and it was 2.5% in the first half of the year. You could also look at all the CapEx being spent right now on this build-out, and it’s roughly 3% of GDP. So you could say that all of the economic growth in the United States right now is coming from this superintelligence build-out.

I concur with that. We can have fun with numbers, but the Trump economy is a disaster according to Trump’s own supporters. Let’s just call balls and strikes here. His own voters, the independents who flipped over and voted for him, are very displeased with the economy. You can see that clearly in his having the worst approval rating of his entire career, and there have been some doozies in that career. Friedberg, what should the Democrats—

I don’t think even those critics—even—first of all, I don’t know what Trump supporters you’re talking about, but—

I’m talking about the America First people who feel he didn’t deport 20 million people, and that they—

Yeah, I know, I know, I know there are—

—they still have inflation. I’m just talking about that cohort. You could put—

Yes.

—Tucker in there. You could put Megyn in there. You know, that whole sphere.

Yes, there are definitely people who don’t like the war, and they recognize that there is some inflation being caused by the war. But I don’t think they’ve said the economy is a disaster. Anyway, whatever.

I think it is—

I don’t want to equivocate about words.

—to the bottom half. If you’re not an equity holder, I think you look at this economy and say both parties—Biden and this Trump 2.0—don’t care about them. That’s what all the feedback you’re getting says, and we’ll confirm it all next Tuesday and see what happens with the actual voting. Chamath, let’s continue your teardown.

That was section one: you always read the risk factors first. Get a sense of what management is concerned about. Again, the takeaway is that management is really only concerned about compute. This is a compute problem.

The next thing I always look at is the revenue side of the ledger, and I try to understand what kind of business they have. This is an incredible business, exactly as Friedberg said. Why? Because, given their growth and revenue, they’ve barely scratched the surface of the commercial working world.

This is a more nuanced chart because it starts to break down what we think the revenue stratification is by customers inside Anthropic. There are about 40 million individuals and small businesses spending under $100,000. There are about 10,000 spending between $100,000 and $1 million. There are about 1,650 spending between $1 million and $10 million, and only 250 customers spend $10 million or more.

To your point, I think what Friedberg said is right. The setup is that these folks can drive negative churn among existing customers, grow, and attract new customers at an incredible rate, which I think is crazy. This allows me to think about the upper bound of revenue. You don’t need to get to 100% of what’s possible. You need to get to 25% or 30%, and then you need to have a reasonable distribution of negative churn assumptions. What you start to see is that this can be a very large revenue business, in the hundreds of billions of dollars, in this three-to-seven-year time frame.

What slows that down? As far as I can tell—and this is more me putting the 80/20 hat on—the reason you have so much traction in the $1 million and under segment is that when a company gets large enough, it allocates responsibility and budgets down. At the director, manager, and VP levels, there are millions of humans running around with kind of million-dollar budgets, so there’s a lot of experimentation that can happen inside that bucket.

They’re willing to sign up for 3, 4, 5, 6, 7, 8, or 9 things. The minute you transition from a million-dollar deal to a multimillion-dollar deal—and this data confirms it—it’s a much more thoughtful pitch. It’s much more cross-functional inside a company, and you have to show some form of ROI or total cost of ownership advantage, right?

What’s interesting about this, Jason, is that 1,900 companies have been able to make the case. That’s what I would say. I think there’s a ton of experimentation, and about 1,900 companies can clearly see an ROI where they can now start to write big checks. Of those 1,900, a big portion are going to be AI natives, folks that are venture-funded. So you have to strip those folks away, and what’s left over is probably a couple of hundred Global 2000-type companies.

I think the ROI case is still a little speculative, but the potential is there, and as long as the potential is there, I think the market’s going to price it up. So that’s revenue composition. The next thing I think about is capitalization. I said this before: Anthropic needs to have $200 billion, and I really hope they actually get close to $250 billion.

First of all, the CFO of Anthropic, Krishna Rao, has done an unbelievable job. The fact that, entering June, he already had $125 billion is unbelievable. There’s a chart I didn’t include, but their total net burn to get there has been, like, $9 billion or $10 billion. It’s infinitesimally small relative to the market cap.

I thought they had a lot less cash than this. When I said they need to raise between $200 billion and $250 billion, and they need to price closer to a trillion, I was assuming they had $30 billion, $40 billion, or $50 billion on the balance sheet. I was wrong. These guys need to raise between $80 billion and $120 billion to get to that quarter-trillion. I think, in order to do that, the price is probably now going to come in between $1.3 trillion on the low end and $1.7 trillion on the high end. Call $1.5 trillion the midpoint. If I had to price the IPO just eyeballing this, I think it’ll come in around $1.5 trillion.

The other thing I noticed is how much cash Alphabet has and also how much cash Meta does not have. I was shocked to see that Meta had negative $1.2 billion of cash. This must be a mistake, but if you look at the 10-Qs as of the end of September, this is what it shows.

Net cash, you mean?

Net cash.

Maybe it doesn’t include short-term investments or something like that.

I’m going to take a look at Meta’s numbers right now.

Yeah, it doesn’t make sense.

Yeah. I’ll tell you in one second.

But, Chamath, your larger point is right: Anthropic, like you said, has over $100 billion of cash sitting there. They’ve burned very little of the money they’ve raised. Again, their company has been a model of capital efficiency. They’re about to raise $100 billion in this IPO. I think they’ve got a debt facility with another $80 billion. Add it all up, and this is a company with over $300 billion of dry powder, or it will soon have that.

So then you’ve got all these Anthropic bears who are saying they’ve made all these CapEx commitments in the future that they’re never going to be able to live up to. Actually, this secures those commitments because—

Yeah.

—they’ve got the capital—

100%.

—now to basically back up all of those compute orders, all that—

Yeah.

—forecasted CapEx spending till, I don’t know, 2030.

You’re exactly right. And because they only have to raise $80 billion, or $90 billion, or $100 billion, that’s a lot different than trying to raise $150 billion, $175 billion, or $200 billion, which is where I thought they needed to be. The demands they’re going to put on the capital markets in the IPO are going to be less, which means they’ll have more pricing power. They’ll be able to manage the float better. All of these dynamics, I think, give you a constructive read going into the IPO.

Okay, now let’s move on, because I think here’s my biggest takeaway—

By the way, Meta’s net cash is negative, too. It has $80 billion of cash and short-term investments and $82 billion of debt.

Oh, my God.

Wow.

Okay, now go back to this. What is my big takeaway? It’s bigger than Anthropic. I’m going to build on what Sacks said. The entire United States economy is levered to the AI complex. Very narrowly, we’re talking about 7 or 8 companies. That sets the tone for the entire country. Whether we like it or not, that is where we are.

Why do I say that? One, the most important companies in the world are in a massive spending binge. These were highly profitable, highly cash-accretive, highly cash-flow-generative businesses that have said, “We’re going to put the foot down on spending because we believe in what’s coming into the future.” That’s what drives GDP, and it’s what drives growth.

Okay, next chart. As they do that, though, they deplete their cash, and they have to raise debt. This is what you see. All but Microsoft have increased their debt. By the way, what an incredible job Satya has done. He’s halved his debt while everybody else has gone to an unbelievably crazy high number.

The AI complex now owes $340 billion to $400 billion to the market. You’d say, “Okay, why is that important?” The problem is this: these companies are raising so much money, it starts to crowd out the capital markets. The most important part of the capital market that it starts to crowd out is the appetite for government Treasuries.

What you can see here is that when you look at these AI firms, the amount of money they need to raise through the end of 2027 is 30% of all that Treasury is about to issue in that same period. Whether we like it or not, they are going to start competing for demand. Isn’t this nuts? It’s like you have 7 companies crowding out the Treasury market.

You know, Chamath, a lot of people are looking at that number, and they’re trading these stocks down based on that. I kind of view it the opposite way. I don’t think these are stupid companies. I don’t think they’d be making those kinds of investments unless they thought there was ROI there. And so—

Well, there obviously is—

—in the future—

—ROI based on the token usage and the fact that only 1% or 2% of people in corporate America are even using these tools. We are greatly underestimating the uptick here. Claude hasn’t even launched a bot yet—a Groq bot. OpenAI just launched theirs, and it’s in a small use case.

When 100 million or 200 million white-collar workers who subscribe to, say, the Office suite get the equivalent of Office 365 in corporate America, and everybody’s got 5 agents running, I think the utilization of these things is going 500×. The token usage is—

I think that’s right. And then—

—going to go 500×. That means we’re going to have—

Yeah.

—not enough compute, and everybody’s got the buildout completely wrong. There won’t be enough, and people are going to look back on this moment and say, “Why didn’t you build more compute? Why didn’t you spend more on CapEx?”

You’re absolutely right, Jason. The problem is the regulatory pushback. The number of approvals, the ability for utilities to find the staff, the ability to actually turn these things on, and the ability to get the capital equipment you need just aren’t there right now. So we, as a country, are sort of clogged up. We can’t do it even if we wanted to.

Friedberg, the other issue here is that if we’re not spending enough to build this out, and a large percentage of the country believes in socialism and is against the oligarchs—Republican, independent, and Democratic oligarchs—there’s a… This is what they think of us and of the leaders in our industry and other industries. This is really dangerous, because we’re not going to be able to build out the compute we need. What does that do to the entire industry in the next 3 to 5 years if we keep talking about all this? How do we turn that narrative around so people are like, “Yeah, we should have data centers”?

I have a view that humans try to organize into groups, organizations, to try and create power that benefits the members of that group. We form tribes, we form communities, and in a sense, governments are a form of organization. It’s an organizational coalition of individuals to achieve an objective for, in the case of a democracy, the majority of the voters.

As a result, organizations in a world of, call it, pseudo-limited resources compete with each other for those resources. Nonprofits, for-profits, governments—they’re all effectively competing for resources, because the individuals whose interests they represent are going to be served if they get more resources, more power, and more influence.

To some extent, the power of AI and the influence it enables—in terms of media, capital accumulation, changing people’s minds, and aggregating resources through its competitive capacity, productivity, et cetera—makes it a bit of a threat to government organizations. To Chamath’s point, if one of the manifestations of that is that AI is now accruing more interest from capital than governments, governments aren’t going to want to let that happen. Governments, whether they’re right or left, are going to fight for AI to either become part of the government, with the government subsuming AI, or to control the AI organizations.

This is why I’ve argued the case. I think compute is going to end up regulated in the future, with compute capacity assigned by the government, and they’re going to pass laws to try to do that. Post-production, meaning they’re not going to control the chips; they’re going to control the actual compute layer and determine who gets access to what amount of compute, I think, for governments to compete.

One way this is going to play out, to Chamath’s point, is in the near term, through capital. There’s more capital going to AI organizations than to the government, and the governments aren’t going to want to let that happen. So they’re going to try to stop the data centers because then—

I’ve never—

Yeah.

…seen something in my adult working life where a complex of a handful of companies is effectively competing with governments for capital. You’ve never seen companies, in an organized manner, have one very narrow thing they need money for that is a one-way bet on something, attracting the same amount of money that would otherwise go to a traditional 2-, 3-, 4-, 5-year or 10-year bond auction. It’s kind of nutty, actually. And you didn’t see—

And by the way, the story you’re going to hear from both sides is about which side benefits the people more. That’s going to be the narrative argument, and every side is going to say the other side is lying. Because remember, if one side’s lying and another side isn’t lying, the side that isn’t lying says, “Hey, those guys are lying,” and they’re correct. The side that is lying, de facto, says the other side is lying because they’re liars.

You’ll hear both sides screaming, “We are the better path for society. We are the better path for humanity.” It’s either government or private-market AI, and that’ll be the huge battle that’s going to brew over the next half decade.

Well—

And we’ll see which one, you know—

Politicians are for sale, and Dario and Anthropic should start buying, just like the right did.

Oh, they already have.

I don’t—

Look, Anthropic, even as a startup, prioritized government affairs. They were very active during the Biden administration. All those former Biden officials who were working on AI in the Biden administration immediately joined Anthropic at the end of 2024.

My point, Sacks, is that you don’t—

No, I get that, but—

No, no, I get that, but you don’t see them doing what you guys did: hosting a fundraiser for Trump, being out there publicly in front of it. You don’t see them out there—

Oh, sure.

…doing it.

But they’ve created PACs and all sorts of entities and nonprofits, all NGOs, that kind of stuff, and they’ve been massively funding it. They have a huge—

But why aren’t they doing—

…government affairs operation.

…what you guys did, getting out there in front of it and championing it?

Well, I don’t know what you mean by “you guys.” I—

You were public about your support for Trump. I’ve never seen Dario—

I’m just a person.

…publicly come out for—

I’m not a corporation.

…any politician.

Yeah, exactly. I’m not a corporation. Can you let me explain? I’m just one person who has views, and, yeah, I support certain candidates, and I do fundraisers for candidates. I’m not seeking regulatory capture. I’m just one person—

No, no.

…expressing my First Amendment views.

I’m saying, why hasn’t Dario—

I’m not—

…or the other founders done that? Why haven’t they said, “We choose this candidate. We choose these candidates”?

We don’t know that they haven’t. How do we know that they haven’t?

We know they haven’t done it publicly because we’ve never heard them do it. That’s my point.

Sure, but—

They have given huge amounts of money to PACs, which is public information, and those PACs have supported candidates.

I agree they’ve done it privately.

There’s a huge revolving door where Anthropic has been snapping up every former official. First it was the Biden administration. Now, frankly, they’re hiring a lot of former Trump administration officials.

Oh, really?

Yeah, and this is all public information.

Yeah.

They’re building a massive government affairs operation. And look, I think you can expect that, as they power up through all of this capital, having a $2 trillion market cap—

That’s my point.

—they’re going to have even more influence in Washington, or they’re going to be capable of buying a lot more influence. I think that is concerning.

One thing I will say, though, is that as they go from being a startup to being a public company that’s one of the top handful of companies by market cap on Earth, it’s going to be very obvious that they’re a big tech company. Therefore, they should be treated like other big tech companies, which is to say, with suspicion, as opposed to being seen as some poor little startup that has innocuous views or isn’t the 800-pound gorilla they actually are.

I do think it’ll be clarifying for them to be seen correctly as what I’ve been saying: they are the most powerful monopoly or duopoly ever in human history. They’re creating AGI, right? I’ve been saying that for a long time. So I do think going public will right-size them in the public consciousness.

Yeah, that’s my point.

One thing you’re seeing already, actually, is that there’s starting to be a Doomer-Anthropic split. There was a very interesting article in Vox, this huge Doomer publication, over the past week. It quoted a bunch of these Doomer NGOs that were basically pointing out that if you separate Anthropic’s rhetoric from what they actually do, they are the ones leading the charge on recursive self-improvement. There’s no single company in the world doing more to race toward full RSI than Anthropic. They are hell-bent for leather on getting to full recursive self-improvement.

A lot of the reason they’re becoming so efficient with their compute, cutting costs and offering the same capability at one-tenth the price, is that they’ve increasingly been using their own models to generate algorithmic improvements and create these compute multipliers. Back in August, Anthropic published a very interesting blog about how they were using RSI. They had a chart showing that 26% of the tasks involved in trying to create the next generation of models were already using RSI. That was back in August, so who knows where they are now, in October?

They are pursuing RSI very, very quickly. One way to think about the pacing letter Dario published a few months ago is that it’s almost like they’re creating an alibi for themselves. They’re acknowledging, “Hey, we are pursuing RSI as quickly as anyone,” and almost saying to the world, “Well, we want it to stop. We want it to slow down. We tried to get you all to agree to pace, but you wouldn’t agree to do that, so therefore we kept going.” It’s almost like creating an alibi in advance if this RSI thing goes really bad.

All right.

I think they had to know when they published that pacing letter that they weren’t going to get the whole world on board with pacing, so the whole thing was sort of an exercise in virtue signaling. In any event, I think it shows again that an Anthropic-Doomer split is possible. There’s no way for them to maintain their lead as the number one superintelligence company without continuing to advance the frontier very rapidly and pursue recursive self-improvement. That’s what the Doomers are most afraid of. Of all the various risks, they’re most worried about a breakaway superintelligence.

I do think this may be the beginning of a divorce between Anthropic and the Doomers, but we’ll see.

Okay, let’s get back to your breakdown, Chamath. You had the greatest revenue, the greatest business ever built, yada yada. They have a ton of cash. They have the cash for the build-out. They’re going to raise some cash in the IPO, and all of this bond activity is crowding out Treasuries, is where you left off, I believe. So let’s continue.

Exactly. I think you have to go to these risk factors now and think: if you’re going to price the IPO or buy the IPO, how would you decide to buy it? I think at $1.4 trillion, $1.5 trillion, or $1.6 trillion, it’s kind of priced fairly, so there’s no deal, but they’re giving you an entry price into the future.

So then you have to have a view on the future, and I think there are 2 risks. The first risk is what I showed you on their revenue composition: How long does it take Anthropic to show those 10,000 customers the ROI math they’ll need to move from less than $1 million to single-digit millions to tens of millions? That’s one very simple thing you can check quarter after quarter.

As those numbers swell, the reason they’ll swell is that Anthropic will have defensible use cases, white papers, and analyses that make it straightforward for a corporate infrastructure team and a multi-person executive team to sign off on a multi-million-dollar spend. That’s an Anthropic risk, and I think you can price that risk. The second risk, though, is what I teased here: the risk you cannot price, which is market risk. The AI complex has gotten so big that it now competes with U.S. Treasuries. Why is that important to note, and why does that create market risk? Nick, if you go back to the last chart, please. When you look at this, what does it mean for Treasuries to have to compete with Facebook, SpaceX, Anthropic, and OpenAI for money?

Dollars, yeah.

Well, it means somebody is going to have to pay more. Let’s play out one example. If it’s the hyperscalers that have to pay more because people believe the flight to quality and the duration risk of the government are better, then all the money the hyperscalers and the AI complex, Anthropic, and OpenAI have to raise will have to come in at a higher coupon. If that’s true, then Friedberg’s chart is also true: those rates cascade through the ecosystem. The neoclouds will have to pay more. All these other folks have to pay more. That means you increase and pull forward the risk of defaults, which we talked about last time. So you have this negative reflexive cycle.

Let’s say the opposite is true. Let’s say the AI complex is able to raise money at reasonable rates, and the U.S. Treasury has to go through an auction and pay more. Now the broad-based cost of borrowing goes up, because every other part of the U.S. economy is levered to the 2s through 5s and the 10-year complex. That’s a huge market risk that I don’t understand, because I’ve never seen it before.

What is the market risk? Explain it crisply here as we wrap. What is the market risk?

If interest rates go up because the U.S. Treasuries have to pay more for people to buy their bonds—

Got it. Yes.

—then everything else in the economy has a higher cost of borrowing. That has negative implications. If the opposite is true, the AI complex has to pay more, and then there are a bunch of negative repercussions narrowly to the complex. The problem is we’re trying to buy an AI stock, so many roads lead to downside risk more than upside risk.

Now, there are versions in which this all works out and everything is off to the races. But I guess my structural takeaway, Jason, is—

Yeah.

—we’re entering this very weird phase in the U.S. economy. We’ve seen this before, and we’re seeing it now. This complex is too big to fail. When you start to see these guys representing 40%, 50% of all capital-market equity and debt flows, which in turn represents 30%, 35%, 40% of all Treasury issuance, this competition happens. These guys are too big to fail, which means there’s a different problem at hand: a more systemic problem. I’m not saying that there is one. I’m generally bullish.

I understand you’re highlighting that this is confusing to you as a market participant who’s on the inside. I get it. What I’ll say is, you could actually argue the opposite, Chamath. This is the greatest business ever created. It’s an American business. You could argue that the American economy is still the greatest economy in the world, versus China, et cetera, but we’re running the table on superintelligence. If you look at the history of the greatest franchises ever made, I did a little bit of research on the first 7 years of revenue to back up what you’re saying here. If Anthropic hits the targets they’ve set out, ballpark, it’s getting to $100 billion, then $300 billion in revenue, and eventually $600 billion in revenue. They’re well on their way. They’re at about $100 billion in revenue this year, which is about a third of Google and a quarter of Apple.

If you look at the cumulative revenue in the first 7 years of the best franchises America has ever made, those would be Amazon, Google Search, cost-per-click search, and the iPhone franchise. It’s going to become clear that the 3 best franchises in the history of business, in my mind, are the iPhone, Google cost-per-click search, and Amazon. This doesn’t include AWS; this is just books and retail. If you put AWS into it in later years, it gets even more interesting. It’s possible Anthropic’s business is bigger than the top 3 franchises in the history of business put together. That’s something we’ve never seen, which means you have no choice but to buy American debt and this debt if you’re a global participant. Where are you going to put your money? In France? In China? You can’t put it into China.

Jason, I’m confused. Haven’t you been saying for the past year that open models are going to eat Anthropic’s lunch?

Well, in the face of new evidence, you’d have to say that they’re lowering prices, and the opportunity in the enterprise is so great that they’re not being dramatically impacted by it. You’ll see anecdotally that startups and enterprises are exploring open models, but now that we know their numbers, it’s obviously not enough to slow them down. It’s not slowing them down yet.

Have you changed your tune on that?

I do think it’s still the biggest risk factor, but—

I’m not debating revenue traction.

I’m not debating revenue traction at all or business quality. What I’m saying is that there are moments of systemic risk in markets where everybody goes risk-off. What I’m saying very explicitly is that I haven’t seen, in the last 15 or 20 years, anything this close to the GFC, where you have a narrow part of the economy representing basically all of what’s happening in the capital markets. All that means is that everybody has to pay a lot more attention. It doesn’t matter what side of the AI trade you’re on, whether you’re long or short. I don’t think that’s important. I’m saying something more structural: about 10 companies collectively funding one thing will be competitive with global governments on borrowing. I don’t know what the implications are, because I haven’t seen that before.

To the point, Sacks, that you made in challenging me: their revenue is slowing down dramatically. Their quarter-over-quarter revenue is decelerating, so obviously it’s not a big number.

Right. We don’t know Q3 yet. The scuttlebutt on X—and who knows if we can totally believe this—is that things slowed down a lot in Q3. Some of that may be attributable to the fact that they eliminated 99% of distillation, right? So they voluntarily took a big hit to revenue—

From people stealing from them, yeah.

—to prevent copycat models from distilling them. Then I think the scuttlebutt is that since they cut their token costs considerably, which happened pretty recently, demand has reaccelerated.

Yes.

It seems like—

Jevons paradox in full effect, yeah.

Yes, it seems like they’re back to their previous growth after a couple of hiccups in Q3. But I think the really impressive thing is the way they’re able to cut their token costs. It’s not just because they’re giving up pricing power, right? They’re not losing margin; they’re actually gaining margin. They’re able to do that because of all the algorithmic improvements—

Efficiency.

—they’re making to make their models much more efficient.

Yes.

Now here—yeah, go ahead.

Just to dovetail with what Sacks said, what’s happening with the models is even more impressive. They’re lowering the price of tier 2 and tier 3 models, right?

Right.

Your Sonnet and Opus models are getting cheaper, but enterprises are saying, “That’s enough for me.” As they make those models cheaper, businesses get more value and solve more problems with them, and they don’t feel the need to use the latest, greatest model. If you’re a lab, maybe you need that to do science. If you’re a mathematician, you need it. If they can keep making their second- and third-tier models 10 times cheaper each year, the Jevons paradox is going to get even more pronounced, because enterprises will use them more. Token cost matters, right? We’re sitting here token-maxing.

Right.

Yeah, yeah, yeah.

No, look, what we’re seeing is that we keep seeing Jevons paradox play out at every level. As you cut the cost of tokens, there are more and more use cases, and enterprises use it more and more. The net result is that total demand keeps going up, and total revenue and market size keep growing.

Also, just to reiterate a point you made earlier in the program, the new applications that are coming all consume a lot more tokens than the ones we started with. Agents use a lot more tokens than prompting, and cyber defense is going to use a lot more tokens than just doing a pen test once a year. You’re going to need continuous monitoring of agents, attack surfaces, all that kind of thing. So all the new applications are going to be using exponentially more—

50 to 100 times—

—tokens.

—the number of tokens, yes.

Exactly. So the price will keep going down, but the market size keeps increasing. I think the biggest risk factor to Anthropic is some version of the open-source models—not necessarily catching them, but saturating intelligence at a low price point for certain—

Commoditizing it is, I think—

Commoditizing it—

—the word you’re looking for here.

—for things like coding, where there is a big chunk of the market. That is still a risk. I would say a second-order risk is that, in order for Anthropic to stay ahead and keep making its models more and more efficient, it has to keep using more recursive self-improvement, right? You see this in the charts they publish: they’re using their latest models more and more to discover these compute multipliers, to discover these new algorithmic efficiencies.

So they have to push the limits of what they themselves call risky activity. I don’t know if I fully believe in the doomer story about RSI, but they themselves say that RSI is very risky. They have to keep pushing RSI in order to keep delivering these algorithmic improvements, in order to stay ahead of the open models. I just wonder how sustainable that is. At some point, do you get a breach with the doomers that becomes irreconcilable? Do the doomers want to shut down RSI or slow it down to the point where the open models truly do catch up? I don’t know the answer to that question. It’s a more nuanced take on this idea of open models catching up.

Yeah, and I think this chart is super illustrative, if you can pull it up, Nick. Gallup’s been keeping a survey of who’s using AI at work daily versus frequently, defined as a few times a week. The people using it frequently versus daily—we’re still talking about 10% and 20% here. It’s nascent. A lot of the top enterprises are not letting their employees use this in an unconstrained fashion, and they’re certainly not letting them connect their email, their knowledge bases, their Notion, their Slack.

Big corporations are not letting people actually use the power of AI because they’re scared about cyber. So once the cyber threat, Chamath, gets mitigated, this is going to go from 10% of people being daily users to 60%, 70%, 80%. There’s so much growth left here when every single employee has an agent. I saw just last week Uber bought OpenAI ChatGPT for every employee. That’s going to be the next shoe to drop: Do you buy these things for every single employee?

I think we beat that one to death. Congratulations to the Anthropic team and all the investors on making an incredible business. Extraordinary. Shout-out to Sacks, who predicted this would be the greatest business ever on multiple programs. Turns out you got that one correct. We’ll see what happens.

But like I said, I said 2 things. Number 1, best business ever, creating this AGI monopoly. Second, they’re going for regulatory capture, and we should definitely not allow that to happen, especially in light of number 1, right? It’s just kind of unseemly and gross for them to be seeking regulatory capture when they already have such a good business and they’re in such a dominant position.

I think this IPO will be very helpful in clarifying that because, again, once they’re seen as a big tech monopoly or part of a huge duopoly—

Any kind of advantage will go away.

—the arguments for giving them regulatory moats just go away.

Yeah, yeah.

In fact, we should do everything we can to make this market more competitive—

How long before—

—and help it stay competitive.

—we have the “Break Up Anthropic” movement? Break up Anthropic, separate the lab—

It could happen. It could happen.

—from the tokens, from the subscription. It will. Who said it, Gavin, that they were claiming they were building the last business ever? When you look at the other franchising—

Yeah, Dario said that.

Well, Gavin said somebody said Dario said it.

Oh, sorry. Yeah.

Yeah, so—

Gavin said that Dario said that—

—and that actually—

—but Dario did not deny it, remember? He had this lengthy response by Dan—

Yeah.

—he said, “Might have been,” kind of said, “Yeah.” I mean, I think they obviously believe it. If you look at that chart I made with the other franchises, they obviously have an incredible franchise that’s theirs to lose.

When we get the OpenAI data, then we’re going to have all 3 of the major players in this space disclosing the key information we need. That’s going to be really interesting when we have all the hands turned over in this poker game when OpenAI files.

3. Google’s Search Problem

All right, Google got smashed on Thursday, down 10% to 15%. We record on Thursdays, obviously. Q3 earnings wiped out about $500 billion in market cap. Why? For 2 straight quarters, Google missed on search revenue. Even though searches are growing massively, the people clicking on the ads are not, which leads to a little bit of shenanigans in trying to get people to click on more ads.

They’re negative on free cash flow, even though Cloud is ripping. Cloud doubled year over year, $30.7 billion. That’s the AWS business they have. It’s called Google Cloud. It’s quite nice. Overall revenue is still strong; they’re growing 25% year over year. The issue is people are looking at that Google cost-per-click advertising network and saying, “Well, obviously that goes away at a certain point if you just give people the answer,” right? ChatGPT, Anthropic, Grok—they just give people the answer, and in some cases they go solve their problem if it’s an agent.

The search business is what pays for everything, and that grew 12%, down from 17%. So it’s decelerating, but it’s still growing nicely. Friedberg, this is your alma mater. You witnessed the cost-per-click greatest business ever. This was the previous one that we said was the best business ever created, along with the iPhone franchise. What are your thoughts here? Are you worried about that franchise, or do you think they’ve done a good enough job with the other businesses? Google Cloud and YouTube come to mind, Android, in terms of diversifying their revenue. Are you worried about Google as a buy opportunity, or is this a stay-away?

It’s a complicated story. I think the core services, the core Google ad revenue, are a real challenge right now in terms of whether they can keep the momentum up. Overall, the business trades at roughly 21 times operating income, and you get that sort of a multiple based on reliability of future cash flows and, effectively, growth of those cash flows.

I think the Google Cloud business is looking really tremendous. I don’t know if you guys remember, it wasn’t too long ago that this thing was not making money. Last year, the Google Cloud business for this quarter was generating $3.5 billion of operating income. This quarter, it’s over $11 billion of operating income, and this is a lot like the SpaceX AI story, where the value of the compute infrastructure that’s been built is just tremendous. The question is, does Google actually have the models that sit on top of it? In many cases, they do, so they can make a higher margin on that piece of the stack.

When I say it’s complicated, I think there’s a story you can build in the cloud business that shows the next couple of years of really progressive growth in that business. The core Google services are where it becomes challenging. Over the last year, it grew from $33.5 billion to $39 billion, and, as some people have pointed out, this may largely be driven by a change in the pricing algorithm for ads and how Google is now, in some cases, over-earning for those clicks. Over time, advertisers will see that. They’ll experience that. They’ll reduce spend, they’ll attrit some dollars, and it won’t look as good of a business.

So the question is, do they have a consumer AI business that catches up and accelerates to support that? That’s the thing I think everyone needs to be watching: What does the AI replacement look like on the consumer side for Google’s core business?

The core cost-per-click advertising network when you type a search into Google. That’s the core business you’re talking about, yeah?

That’s the core—

Yeah.

—consumer business. So the question is, is there a consumer AI alternative that starts to generate meaningful revenue to replace that revenue? What does the revenue ramp look like there relative to the flattening or decline that may arise in core search, and how quickly does that happen? That is a dynamic and challenging question to answer now. Internally, I think there’s a bit of turmoil. There’s a bit of, “What are we?” Not just overall as a company, but even within the business segments.

What is Google Search? Is it AI or search? What is Google Cloud? Is it the models, or is it just the compute—the rental of compute? I think we saw this battle play out recently, where Jeff Dean and others left Google. I don’t know this for sure, but the theory is that the computer scientists building models at Google may not be getting the resources they had hoped for because Google can make so much more money renting that compute out to others.

So it’s a really hard question: What do we want to be at Google right now? I don’t think they have a great answer, and I don’t think it came across well in the earnings report.

It didn’t come across well, obviously, and people reacted, Chamath. But they did know this was coming. They did plan for it. They renamed the company Alphabet, and they said, “They’ve got this great CFO. Let’s make all these other bets fend for themselves the way most of the world has to make its way in the world: raise outside capital.” So they did prepare for this, in fairness to the team over there.

YouTube is surging, Android is surging, Google Cloud is surging, and then you have the franchise slowing. Search is up, but revenue growth is slightly down. Do you think they can fix this, or should they just do a Bending Spoons and get as much money as possible out of Search and put it into other things, like consumer AI, corporate AI, and their future models and hosting of other people’s models? What are your thoughts here, Chamath?

I don’t think that’s what Bending Spoons does, and not everything is an analogy to Bending Spoons. This is the best business—

I’m saying, when I say Bending Spoons, I’m saying manage it for earnings, not for growing it necessarily. Just get as many dollars out of it as possible.

I think that’s what they’ve been doing in the last quarter. That’s why people didn’t like it. There was a very interesting tweet. Nick, maybe you can put it up here. I think a lot of the consternation came from people reacting to what Max Anderson wrote, which got 5 million views, by the way. He wrote this through the lens of somebody spending half a million dollars or more a month on Google for years, and he itemized all the ways things have changed.

The most important thing I pulled out of here was what Friedberg mentioned, which is the change in the auction model. Typically, in a Vickrey auction, Friedberg bids, I bid. If I’m the underbidder, Friedberg gets filled at my price because he was the high bidder. It’s a really great way of doing price discovery. At Facebook, when we built the first ad system, that’s essentially what we copied. We copied Google feature for feature.

But if they’ve now abandoned the Vickrey auction, Friedberg’s right: you’re overearning. If the market starts to smell that out, they’re going to start to wonder what’s actually happening under the hood. These claims, which are now well syndicated, have been out there since July, and I’m not sure Google IR has necessarily nipped them in the bud. I think what happened in the information war after earnings was that people recirculated Max’s post, saying, “This has been going on for a while.” When you look inside the comments, it’s a bunch of other people seconding his analysis as accurate.

Without knowing the details, Jason, because I don’t spend money in the Google ecosystem, I think it’s really important for Google IR to get in front of this. If this momentum builds—the idea that this franchise is essentially robbing Peter to pay Paul, pulling money from the future into the present to smooth out earnings—the market will react badly to it. You’re seeing it today.

Mm.

But then this other problem starts. To your point, the composition of ad revenue as a percentage of the overall will fall below a threshold, and then they do look more like a structured conglomerate, where you have a YouTube revenue block, a GCP block, an Other Bets block, and a Search block. None are dominant. In some ways, I actually think that’s better because then you’re more diversified and less levered to any one single business.

But they’ll have to go through this complicated process where a bunch of folks may not want to own that kind of business, and there’s a different buyer pool. You have to transition out the folks who wanted to own a single-stock bet on Search or ads or whatever it was, and now have to move to this other kind of business model. When you switch buckets like this, a bunch of folks say, “Oh, it’s too complicated.”

Problem number one, if you’re Google IR, is: please read this and refute it, or try to figure out how to get in front of it, because every analyst will now try to figure out whether it’s true. The second, though, is very narrow: How do you keep Demis at Google? That’s it. That’s the only task.

Okay. Sacks, looking at this business, it’s been a journey over the past 10 years to make Alphabet resilient in its revenue streams and diversify those revenue streams. Do you think there’s some great risk here, and would you be long, short, or neutral on the stock?

There’s nothing else. Pay Thomas Kurian whatever he wants. How do you keep Demis at Google?

I’m seeing people post that Google Search revenue dropped below 50% of total revenue for the first time, and they’re making it sound catastrophic, basically playing into the narrative that the Search business is going to be disrupted by AI. That’s what it seems like on the surface, but when I dug into the numbers, it didn’t really seem that bad at all.

It’s true that Google Search’s relative share decreased, but its absolute revenue still increased. If you look at Google Search revenue year over year, it was still up 12%. It increased from $56 billion to $63 billion, so that’s still roughly a $7 billion year-over-year increase in Google Search. Google Search is still thriving, I think you could say. It’s not growing that fast relative to other businesses, but you’re not seeing it deteriorate yet.

The reason its share fell below 50% is that other Alphabet businesses are simply growing much faster, and the main one is Google Cloud, which basically doubled year over year to $30.7 billion. It’s the rise of the Google Cloud business that has made Google Search seem like it’s on the decline. I tend to think it’s a good thing that Alphabet is finally getting some of these other businesses to work and grow fast, and to finally diversify its revenue. Diversifying from Search is something they’ve wanted to do for a long time. They now have all these other segments: YouTube, G Suite subscriptions. In any event, I tend to think Google’s a little bit oversold on this news today.

I agree.

Although I’m certainly not telling anyone to buy anything.

Since inception, and maybe over the past 10 years, you can pull both charts up. We’ll go through them real quick here. This franchise—the number of searches is still going up. The revenue per search, and that issue, is obviously the one people are concerned with, but the other side of the coin is absolutely correct, David. The other businesses: they’re now only 50% dependent on this business, and that will continue.

For me, I just literally bought some shares, so I’m long. I just bought more shares of Google while doing the show.

And, by the way, just remember Alphabet owns 6% of SpaceX and about 15% of Anthropic as well.

I think 85% of Waymo as well, right? That’s the other piece of this puzzle. Amazon also owns a big piece of Anthropic. They’re going to be a big winner. All right, listen, that’s our time here.

4. Midterm Election Predictions

Hey, so do you guys want to talk about the election ahead of Tuesday night and predictions or anything like that?

You’ve been shepherding this. Go ahead.

What are you guys watching in the midterms? Anything you’re keeping an eye on, Chamath? Any big contrarian predictions on what might happen on Tuesday?

It’s so interesting. I’m really confused. I’ve started to stop paying attention just because there’s so much misinformation. I thought we were going to have—and I still think—a spritz, as Sacks called it.

You know, a little spritz.

I think it’s more of a Democratic spritz. When you look at some of the early data, and I don’t know if Brett Baier was doing this just to hedge his bets or not, he did a clip that went viral on X where it just looked like it was going to be a bloodbath for the Republicans, and they were going to lose both sides, both the Senate and the House.

I think it’s still set up for the Republicans to keep the Senate and the Democrats to have a narrow majority. Going back to maybe where we started, the mandate from the Democrats then becomes obvious: How do you introduce FUD? There’s one way to do that, which is to start subpoenaing everybody, start doing disclosure and discovery, and release document after document after document.

And let the masses start to churn through this stuff going into ’28. So, number one, create FUD. Number two, I think an economic slowdown benefits the Democrats going into the ’28 election cycle. Whatever they can do to influence power permitting and data centers—because, as we talked about earlier, that’s the lever, but to me, it almost doesn’t matter what happens at the federal level, except for the FUD part. I think we should now be paying a ton of attention to what’s happening in these governorships and some of these other races.

Sacks, has your view changed?

Look, I don’t think you can trust any of these polls. First of all, a couple of these polling firms have come out and admitted that they’re fake. We literally have a couple of—

Oh, yeah, there was a hoax, right?

Fake polls. There was literally a hoax, and those—

Yeah, and people fell for it.

Those fake polls were included in the averages. Some of the polls are outright fraudulent, and the rest are well known to have a bias. Patrick Ruffini did a study of the last 3 elections and found that 75% of the polls erred in favor of Democrats. We see this as a recurring pattern. We saw this in 2024. Ann Selzer, who was supposedly the gold standard, said that Trump was going to lose Iowa by about 10 points, and it turned out to be total nonsense.

The pollsters have a history of bias, fraud, or, for some reason, oversampling progressives in their polls. People should shut that out. I think some of these polls are a tactic designed to depress and suppress turnout, and Republicans should just go out there and vote. The economy is not a disaster. That is total fake news. We know what the growth rates are: It was 2.5% in the last quarter, and it’s tracking at 3.7% right now.

Median household income is almost $90,000. It’s at an all-time high. Poverty is at an all-time low. I think the percentage of people below the poverty line is about 12%, the lowest it’s ever been. It’s not just a K-shaped recovery. You’re seeing real wage growth in real terms for the average household, which was not true during the Biden years. I don’t think people should get psyched out by this media operation to make people super depressed about what’s going on.

J-Cal, any views?

Well, I agree. The polling data has been suspect for a long time across many elections. Wedgewood did this thing they’re claiming was performance art, but they were kind of serious. I don’t know who they’re working for or what the shenanigans are, but I do trust Polymarket. I think when you look at Polymarket, it’s pretty clear what’s going to happen here.

By the way, Polymarket was wrong on Bolsonaro, although I think maybe he got robbed. It was wrong on Netanyahu. Polymarkets are just a reflection of public sentiment, or the public sentiment that’s willing to place a bet.

Yeah, that’s what I mean. It’s that latter.

And they can be totally wrong.

Yeah, I mean, it could be wrong—

So they were totally wrong about those 2 national elections. By the way, I’m not predicting some sort of miraculous Republican win here. It would be very unusual for Republicans to win a midterm election when they’re in the White House. You remember that Bill Clinton and Barack Obama both suffered horrendous midterm losses. Bill Clinton lost in 1994 to Newt Gingrich and the whole Contract with America. That was the first time Republicans took control of the House in 40 years. With Obama, you had one of his midterms that was the shellacking.

It’s quite customary for the out-of-power party to do very well in the midterms. I’m just saying that the media is already portraying this as a tsunami or a bloodbath, and I think it might be a more normal midterm result. That’s all I’m saying. In any event, people should just go out and vote, not get depressed or psyched out by whatever the media is telling them, because the media has a long track record of being wrong and biased.

J-Cal, any view that’s different from where we’ve been in the past?

I think Polymarket is going to get it right. I think the Democrats are going to win the Senate. It’s a 68% chance right now on Polymarket, and I believe that. For the House, obviously everybody believes that: 92% chance.

The reason I’ll back that up is that Trump’s net approval rating is at an all-time low. The country’s very upset, and the libs are going to come out in force to vote anti-Trump. The moderates are either not going to show up, or they’re not going to be Trump-curious like they were for the second term. Kamala was such a disastrous candidate that they were like, “Okay, gotta try something else here,” and they went for Trump. Reasonably, if you were going to have a choice between Kamala and Trump, I think Trump’s the better choice for most Americans.

Then you have the fractured Republican Party, where the America First people feel like they got rug-pulled by Trump because the economy didn’t improve, inflation didn’t get solved, the wars didn’t get solved, and he started another war. So that group’s kind of pissed off. When you put all that together, I think it’s a very easy sweep for the Democrats.

Then, if we want to look at what the next 2 cards are, I think we’re going right back to Trump being impeached. From what I understand, the impeachment process goes through the House and then the Senate, and it’s a legal proceeding. Both houses are going to do that. It’s going to be absolute chaos, and then the Democrats are going to start putting legislation out there and force the Republicans to vote against it in order to scuttle JD Vance’s chances. This is obviously going to be a chaotic—

Phase.

—and we’ll cover—

You and I—

Let me finish. Hold on. Let me finish.

You and I are alike.

Let me finish. Hold on. Let me finish. What’s going to happen is it’s going to be a chaotic 2 years, and I think what will come out of that ultimately is we’re going to need some new candidates on the Republican side. They’re going to put the Trump era to bed, and then I think that means they put JD Vance and Marco Rubio to bed. They’re going to need a new champion who’s more populist and actually executes on the populist agenda that people feel Trump represented but didn’t execute on. That’s my analysis. Go ahead, Chamath.

No, I think what I’m saying is that you think there’s going to be a Democrat flood campaign. They could control the House. They’ll just subpoena everybody.

They’re going to subpoena everybody. They’re going to look into—

Discover everything.

—every single accusation of grift they can think of, whether valid or not, and everybody in between. Yeah. And then, we have a really good group, by the way, coming. So join us for the All-In midterm livestream, 8:00 to 10:00 PM, YouTube, X, and Polymarket. Shane Copeland, Tim Dillon, Rahm Emanuel, Taylor Lorenz. It’s going to be chaotic and fun.

Taylor Lorenz, she’s on a redemption arc, I have to say.

Based Lorenz.

Based Lorenz?

She’s actually been great on certain topics lately.

What is going on? She’s so based.

Full redemption arc, I have to say.

Wow, Based Lorenz coming on the pod. Love it.

I like this new Taylor Lorenz.

I love Tim Dillon. He’s coming on.

I’m not going to be there.

What?

Oh, you’re not? Where are you going to be?

What do I need to do a midterm livestream for?

Because you’re—

You’re in politics.

—because your party’s going to lose, so you’re not coming?

If we outperform, I’ll pop in and spike the football. Otherwise, I’m not showing up.

Come on, Sacks. That’s not cool. Everybody wants to hear you.

Come on, we need you.

We can drink some tequila. We’ll have fun with it, Sacks. Come on. You can’t cry into coffee.

We should all be drinking, by the way.

Drink the tequila.

Maybe we should get together at Chamath’s house, drink, and do it live.

I agree with a lot of your analysis about what will happen if the Democrats win, JCal, because they have not changed their stripes. People are just forgetting who the Democrats are and what they were like during COVID: all the lawfare they perpetrated, all the crazy woke stuff. It’s all there, ready to come back. They have not really analyzed why they lost in 2024. They have not really changed or adjusted anything. If anything, they’ve become more extreme. The DSA is ascendant in their party, as even Gavin Newsom acknowledged.

I think that if the Democrats do take Congress, you’re likely to see how nutty they are, and the American public will be reminded of that. That will be really interesting to see.

But that’s why there’s a pendulum in politics, right?

There is a pendulum. I’m wondering who the champions are. If it’s not JD and Rubio, I wonder who the champion on the right will be. I think it’s going to be somebody we can’t even name right now who’s going to emerge, and I think the same thing on the left. Anyway, it’s going to be exciting. We’re gonna get Sacks to show up.

Love you boys.

Love you, besties. Bye-bye.

Bye-bye.

Let your winners ride.

Rain Man David Sacks.

And instead, we open-source it to the fans, and they’ve just gone crazy with it.

Love you, besties. Ice Queen of Kin Law.

Let your winners ride. Let your winners ride. Let your winners ride.

Besties are back.

That is my dog taking a piss in your driveway, Sacks.

Wait, no.

Oh, man. My haberdasher will meet me at Glenn Center.

We should all get a room and have one big, huge orgy, because they’re all—

Yeah.

There’s this sexual tension that they just need to release somehow.

Let your—

Feet.

Feet be. Let your feet be. Let—

That’s gonna be our... We need to get merch—

Besties are back. Let your feet be.