Skip to content
All-In PodcastAll-In Podcast

How hyperscalers bet $725B on a grid that can't keep up

The $725B hyperscaler CapEx wave is chasing electricity, not model demand. No GPUs sit dark; OpenAI's user miss is a power problem, not a product one.

Jason CalacanishostDavid FriedberghostChamath Palihapitiyahost
May 1, 20261h 20mWatch on YouTube ↗

EVERY SPOKEN WORD

  1. 0:00 – 3:05

    Bestie intros

    1. JC

      Jason, do you wanna tell us about your new favorite podcast?

    2. SP

      [laughs] Oh, it's so good. My feed is now, because, you know, since cancel culture ended, Sacks, everybody uses the R-word and the F-word right now, my entire feed on Instagram is either gay or Down syndrome or bulldogs. It's one of those three. And then I stumbled upon the Miss Thing Pod, Miss Thing, and they do a bit called Gay Name, Straight Name. Here's gay name or straight name for David. This, uh, good news and bad news, Friedberg. Here we go.

    3. SP

      Gay name or straight name. David. David, to me, is straight. [gasps] Okay. But he has my perfect body. It can be confusing because I'm kind of like, "Are you gay?" And it's like, "No, I just wanna be you, David." Totally. Well, it's so, like- Mm ... the Michelangelo's David. Exactly. The male ideal. It's like- Yeah ... incredible body. Yeah. Kinda small [censored] . Sorry. Yeah.

    4. JC

      Oh, it's a little rough. [laughs]

    5. SP

      Oh, boy. [laughs]

    6. JC

      It's so great.

    7. SP

      What are you watching there, J Cal?

    8. JC

      They basically nailed these two, but okay, keep going. [laughs]

    9. SP

      I don't think Chamath is on their shortlist, but I know Jason will come up at some point.

    10. SP

      Gay name or straight name.

    11. JC

      Maybe this is it.

    12. SP

      Chamath.

    13. JC

      [laughs]

    14. SP

      What went on the count of three? Yeah. Three, two, one. Gay.

    15. JC

      [laughs]

    16. SP

      I'm seeing, like, Italian sweater, like really kind of like a loud- Ooh ... vibrant sweater.

    17. JC

      [laughs]

    18. SP

      He, like, wears it to, like, poker night with his-

    19. JC

      Yeah

    20. SP

      ... like, his boys, and like not, I'm not talking like straight poker.

    21. JC

      Yeah.

    22. SP

      I'm talking, like, gay poker nights.

    23. JC

      [laughs]

    24. SP

      Like at the bar. Yeah. Always talking about wine. Talks about wine. Always sort of like- So much ... swishing it.

    25. JC

      [laughs]

    26. SP

      Yeah, exactly. Yeah. Also, it's so like a guy at the gym taking off his shirt.

    27. JC

      Oh.

    28. SP

      Taking selfies. Photos. Mm-hmm. Yeah, and everyone else is kinda like, "Excuse me, Chamath, I'd like to use the mirror."

    29. JC

      [laughs]

    30. SP

      Yeah. "I'd like to see myself." See you at the next gay poker night. [laughs] Totally. You bring the wine? [laughs] We'll bring the sweater. Yeah.

  2. 3:05 – 20:02

    OpenAI misses targets, Codex gains on Claude

    1. JC

      OpenAI. They missed their targets for ChatGPT, Friedberg, both on users and revenue. Let's talk about it. The Wall Street Journal says in a, uh, a breaking investigative report on Tuesday that OpenAI expected to hit one billion WAUs, weekly active users, before the end of 2025. They missed that, and they still haven't hit the milestone four months into 2026. Also, Chamath, they missed their 2025 revenue target for ChatGPT. Exact number wasn't specified, but as we've talked about here, they're at a $20, $30 billion run rate. There's a little bit of accounting nuance that is yet to be worked out in the industry. Two reasons why this matters, Sacks. OpenAI has $600 billion in spending commitments for compute. Just to put that in perspective, that's about what they're trading for on secondary markets. In other words, the entire value of the OpenAI enterprise equals their spend commitments in the coming years. CFO Sarah Friar, who is coming to liquidity, is reportedly worried, hey, that revenue isn't growing fast enough to keep up with expense, and OpenAI wants to IPO later this year. This has put Friar and Altman in conflict or, uh, maybe there's some natural tension there. Friar doesn't think OpenAI is ready for public reporting standards, according to The Wall Street Journal. Altman obviously wants to move faster. So they released a, a joint statement, "This is ridiculous," yada, yada, yada. Let's go to you, Sacks. What do you think's going on here? Are these major headwinds, or is this just managing expectations as the leader of the pack in the most important race of our lifetimes, the race towards super intelligence?

    2. SP

      Well, I actually have a little bit of a contrarian take on this. I know that OpenAI had a really bad week. Like you said, they had that Wall Street Journal article which said that they've missed their numbers. They missed their one billion user growth target. They missed their revenue numbers. That's called into question whether they can afford the data center commitments that they've made. And then in addition to that, they've also had the lawsuit with, with Elon happening this week. So in the press, it ended up being, I think, a pretty bad week for them. But I have a contrarian take on this, which is I think that over the past week or two, if you look at kinda what's happening at the product level, it's been a pretty good couple of weeks for them. They released ChatGPT 5.5, and the reviews from, you know, people I talk to in Silicon Valley have been really strong. You talk to developers, coders, they're very happy with it. At the same time, Opus 4.7, which is the latest Anthropic release, appears to be a bust. People are complaining about it. They're, in a lot of cases, they're rolling back to 4.6. They're saying that Opus 4.7 is rationing compute. It's reducing thinking time, not as good. There were some bugs in Claude. So if you just compare ChatGPT 5.5 to Opus 4.7, it does appear that OpenAI has had a better couple of weeks at a product level, and I think there's reason to believe that-The product improvements will continue. GPT 5.5 is based on a new base model called Spud, which is the first base model upgrade they've done in, I don't know, over a year. And having a new base model will pave the way for future improvements as well. So I think OpenAI is feeling pretty optimistic about their product right now, and I think you're starting to see on X that some of the developer mojo is shifting. I'm seeing a lot of people saying that they are shifting their, their coding usage from Opus to GPT 5.5. So I think that Sam may end up being right but for the wrong reason. And what I mean by that is that when he made these big compute commitments, it was based on those estimates of hitting the billion users on the consumer side and hitting those revenue targets. The consumer business ended up being weak, so they missed those targets. But in the meantime, coding has become the all-important sector of AI. And because they made all these compute commitments and they built out these data centers, they have more compute than Anthropic right now. Anthropic is token constrained. It's reducing their ability to serve Mythos, for example. It's causing them to engage in compute gating with Opus 4.7. And I understand why Dario made that decision. I'm not saying... I mean, it was a prudent business decision. I'm not criticizing him for it. But I think, again, I think Sam may end up being right here for the wrong reason, which is he missed on consumer, but enterprise is going gangbusters and is giving him the ability now, I think, to catch up on coding-

    3. JC

      Here's your Polymarket

    4. SP

      ... which is the all-important market right now.

    5. JC

      Of course, and we talked about Groq and Cursor teaming up last week, Elon and the team over there. Polymarket showing now a 32% chance that OpenAI goes public by the end of 2026. This is down from 60% in December. And Chamath, you gave a bit of a warning, "Hey, there's only so many dollars to go around. SpaceX IPO is obviously getting out first. And now if OpenAI doesn't go out this year and Anthropic does, this sets up an interesting dynamic." What are your thoughts here, generally speaking, about the massive commitment that OpenAI has made? Are they gonna run off the cliff, or will it wind up being brilliant, uh, even if it wasn't strategically for the exact reasons?

    6. CP

      I think they're gonna be fine. I think this is a multi-trillion dollar company. I think Anthropic is a multi-trillion dollar company. I think the thing that's happening right now is, uh, a complete misunderstanding of what's actually happening inside of the world of AI, and there is one very specific choke point that is constraining everything, which is access to the power that's necessary to drive these tokens. To the extent that OpenAI missed, I think what that is is an insight to not enough compute capacity today, and that problem is only getting worse. You've already seen that with Anthropic as well, where they just found a way to economically induce Amazon to give them enough capacity so that you don't have to route through Bedrock to get to the Anthropic models. You're also seeing them do differentiated deals now with economic participation on top of what they already had from folks like Google to give them more capacity. What is my point? Everything in this market is power constrained. The reason that these folks may miss a number or a forecast have nothing to do with demand. It is entirely 100% due to the supply of the power necessary to generate the output token. There is a really interesting thing that was just announced today that will make this problem even worse, which is what you're starting to see now is backlogs build up of not just the access to the power, but then the componentry that's actually necessary. Not just resips and not just nat gas turbines, but now you're talking about transformers and all the actual tactical grid infrastructure. Why is this important? If you look at the actual amount of gigawatts that are under construction, we have a huge mismatch now. People have announced all these projects, Jason, but less than half of it is actually being built. Less than half. Most of it is stuck in red tape. Most of that is because there are these supply chain delays, so there's no credible strategy to turn any of this stuff on. Who will this hurt? It will hurt Anthropic and OpenAI the most. Who will this benefit? It will benefit the hyperscalers, specifically Oracle, Amazon, Meta, Microsoft, and Google. And now what you're going to see is a negotiation and a trade back and forth. How much equity do I have to give up? How much control do I have to give up to get access to the compute versus how badly will I miss my growth forecasts if I don't? And now what that means is, and we spoke about this last week, that's a huge lane for Groq to just run through and SpaceX to run through 'cause they have a ton of excess capacity. And so I think the Cursor deal was the appetizer. But if I were Elon now, I'd be running all over this market because if the models catch up in quality, I think he could also do something really crazy with Anthropic or OpenAI right now. Now, maybe not OpenAI because of the-

    7. JC

      We'll get into the lawsuit in a moment

    8. CP

      ... the baggage.

    9. JC

      Yeah.

    10. CP

      But man, he and Dario should do a deal tomorrow.

    11. JC

      So you're framing, "Hey, the, the limited resource here is compute. The demand is off the charts."

    12. CP

      No, the limiting resource is power.

    13. JC

      Power, which then powers compute-

    14. CP

      And by the way, sorry-

    15. JC

      ... which then provides tokens, which then services the massive-

    16. CP

      Exactly

    17. JC

      ... uh, developer and cowork and all these other projects that consumers and enterprises can't get enough of. Got it.

    18. CP

      And Jason, the other factor that complicates that for Anthropic and OpenAI isAll the stuff that's sort of sitting around thumb-twiddling-

    19. JC

      Mm

    20. CP

      ... forty percent of that is gonna get canceled because they've done such a poor job of creating a good positive halo around AI that forty percent of all the announced projects get canceled because forty percent of all projects in the last four years have been canceled.

    21. JC

      Yeah, and there's, there, there are some bad feelings about data centers, AI, jobs, et cetera, and that's causing some headwind. People are literally doing violent things in society and blaming data centers and AI for it. I don't wanna give it too much airtime. Friedberg, what's your take on the chessboard we're looking at here, either through compute, energy, or through going public on a business level, you know, the strategic nature of capital, compute, and energy now playing a role in this massive amount of demand?

    22. DF

      Still a ball in the air kinda game. BCG had this theory, I think I talked about this once before, called the rule of three, where they've shown time and again that any stable, mature, competitive market evolves to a four to two to one ratio of market share for basically ninety percent of the market. So there's a market leader that has four times the market share of the second place. That's two times the market share of the third place. This is the case in pretty much every mature kind of competitive market. So you can kinda think about AI probably evolving into the consumer market and an enterprise market. OpenAI, even if they're not at a billion, they're still at nine hundred million weekly users, which is well ahead of whatever Claude is at. I think Claude is, like, probably sub 100 million. Sacks, you may know. And then Gemini is probably closer to them at 700 to a billion, somewhere in that range, probably p- pretty neck and neck with OpenAI. So, you know, the consumer market looks like it's trending towards a ChatGPT/Google fight for first place and second place, and then probably Anthropic in third place, and maybe Elon emerges and takes off, enabled by his compute capacity. And then the enterprise market is a little bit of a different story, and that's its own market, which is kind of Anthropic or probably Google in the lead, actually, if you look at all the Vertex use. Google claims that seventy-five percent of GCP customers are active users of Vertex, so there's probably a pretty sizable market share that Google's captured on the enterprise side as well. This is also probably why Google stock has absolutely ripped over the last couple of months, is they're literally in first place or fighting for first place in enterprise and consumer. But I still think that there's a lot of opportunity, uh, to Chamath's point about the compute and energy capacity constraints, in improving how we actually scale and deploy models in both the enterprise and the consumer setting, and it is such early days. And I just wanna highlight this paper that came out from MIT from these two scientists, and these guys published a paper on pruning techniques in neural networks. This paper showed that you could actually reduce the size of these networks by ninety percent and get the same accuracy out by pruning very large models down to smaller models, and then you can make a selection on which model to run for inference. And by doing this, you can actually reduce inference costs by 10X. You can get 10X the output per energy unit that goes into the data center with no loss of accuracy. And so it's a really interesting, call it algorithmic technique that can be applied to the existing large models to actually make them much lower energy use. So if you think about it, you're firing up a very large model to answer a very simple question. You could actually prune away that model. Now, this is probably gonna be the case in AI applications as it is in traditional Google search. There's a long tail of searches, but there's a few searches that account for a large percentage of search volume. It's like, what is the weather? What are the movies times?

    23. JC

      Sports scores.

    24. DF

      You know, what's the stock price?

    25. JC

      Yeah.

    26. DF

      Like, there's a certain set of things that make up the bulk of consumer energy, and there's probably a certain set of things that probably make up the bulk of coding output as well. And so if you can get that eighty percent of searches or chat interfaces or coding requests reduced down through pruning techniques to smaller models, and then you have a whole set of smaller models that can be called dynamically, and you reduce inference cost by ninety percent, you can make much more use, call it ten times the use, on data center and energy capacity than we can today. So I would argue that we're still in the very early days of getting efficiency in terms of output and tokens, and we're just in the very kind of early stage of that-

    27. JC

      So there's-

    28. DF

      ... which also unlocks the opportunity for guys like Elon to reinvent how this is done and potentially compete pretty aggressively.

    29. JC

      There are two ways to win. You could throw compute at it, or you can do SLMs, small language models, and VSLMs, verticalized small language models. So if you had a verticalized small language model for the weather, [chuckles] let's say, that doesn't exist, but, uh, you can, you can use it as an example. They will have one for travel as an example. When you hit Google for flight information, it's obviously going to route you to something lighter and faster that uses Google Flights, and Google Flights has been incor-incorporated into Gemini. Gemini now is right behind, 700, 750-

    30. DF

      Fifty million, yeah

  3. 20:02 – 31:03

    AI cybersecurity: a market that's about to explode

    1. SP

      Let me just say one other area where I think OpenAI had a good week is in this red-hot area of cyber. Obviously, Anthropic made a huge splash with Mythos. It hasn't been commercially released. They're compute constrained, but as a proof of concept or training model, it hit a new level of capabilities with cyber. But now, OpenAI has released a new model called GPT 5.5 Cyber, which has just been through a bunch of tests, and they've shown, this was testing done by the AI Security Institute, that GPT 5.5 is the second model to complete one of their multi-step cyber attack simulations end to end. So it has the same level of capability as Mythos, and it does appear to be commercially ready. You know, they've got the compute to serve it. So I do think that that's a big accomplishment. I mean, look, we knew that other cyber models were coming. It wasn't just gonna be Mythos. In fact, within six months or so, all the frontier models are gonna have Mythos-level cyber capability. But it's impressive that OpenAI got this GPT 5.5 Cyber out so quickly, and I think 5.5 might be the first cyber model that cyber defenders actually get to use because, again, I don't think they're as compute constrained as Anthropic is.

    2. JC

      And this is an incredible opportunity, you know, for the CrowdStrikes and Palo Alto Networks of the world, both of which have been on the program. They come out, and they start attacking this space. Man, you could really see everything get tightened up, and this could be an incredible-

    3. SP

      Yeah

    4. JC

      ... revenue stream for everybody who's got, whether it's Cursor, Claude, or OpenAI, or Gemini. This is an amazing opportunity to tighten up as much as it is-

    5. SP

      Yes

    6. JC

      ... to get attacked.

    7. SP

      Can I make a point about that?

    8. JC

      Yes.

    9. SP

      'Cause, look, there is so much fear right now, almost a level of panic about Mythos. People are treating it like a doomsday weapon or something like that. It's not. It's simply that the frontier models have reached the point where they're capable of automating cyber activities, just like they're capable of automating coding, but that means that a model could power up a cyber attacker or cyber defender the same way they can power up a coder and allow them to discover a lot more vulnerabilities. So there is obviously a risk there, but I think it's important to understand that Mythos or GPT 5.5, it doesn't create the vulnerabilities. It just discovers them. The bugs were already in the code. They were sitting there waiting for s- some hacker to discover. If we can now use AI to find these bugs in advance, these vulnerabilities, and patch them, then you actually harden our infrastructure-

    10. JC

      Correct

    11. SP

      ... and, and you harden our security. I also believe that this leap from, let's call it pre-AI cyber to post-AI cyber, it's gonna be, I think, a big one-time upgrade cycle because, again, you're gonna find all these dormant bugs and vulnerabilities. But I think that once we get past that upgrade cycle, you're gonna reach a new equilibrium between AI-powered cyber offense and AI-powered cyber defense. It's gonna become a lot more normal. It's not gonna feel like this huge disruption. Which is to say, I think, you know, people are treating this as, like, some existential threat. I don't think it is, as long as everyone does what they're supposed to do, which is use the new capabilities to harden their codebases and infrastructure and security before the hackers get ahold of these capabilities.

    12. JC

      Yeah, and if, Chamath, if you were to look at this, to build on Sacks' point, there are about 5 million or so security experts in the world [laughs] . And we talked about token cost. 40 hours of tokens just pounding it, you know, a week, you could create another 5 million for $100 per chief security officer per security expert. So it's the volume of security expert agent, Sacks, to your point, yeah? You could have 50, fi- 50 million of them, 100 million of them. They're not finding-

    13. SP

      Right

    14. JC

      ... something unique.

    15. SP

      Well, also-

    16. JC

      They're just-

    17. SP

      Yeah

    18. JC

      ... they never sleep. They're relentless in their pursuit of these problems. It's a really great point.

    19. SP

      Just kinda just refine that. So yeah, there's probably 5 million people in the cyber industry, but there's probably only a few thousand really elite hackers.

    20. JC

      Sure.

    21. SP

      Those hackers didn't have the time to go after the entire surface area of every possible attack vector out there. And so if you train a model to do what they do, obviously, like you said, it can operate with a scale and speed that a human hacker can't. So obviously, you know, what you need to do is get these tools in the hands of the white hats, let themDo the cyberattacks themselves to then find the vulnerabilities and patch them before the black hats get ahold of these capabilities. But I think it's just, just one last point on this, then I'll stop, is just it's really important to understand that the Chinese models are gonna have these capabilities within approximately six months. So-

    22. JC

      Oh, they have them now in DeepSeek V for sure. They've got some level of them.

    23. SP

      Well, nah, DeepSeek V, I mean, DeepSeek V is impressive in a lot of ways, but its capability is not at the f- frontier. It's maybe 80 to... 80, 85% of let's call it the American frontier.

    24. JC

      Chamath, you wanted to get in on this. Let's get Chamath in.

    25. CP

      Two things. The reason that this is even possible is because humans are error-prone, and when humans code, they create holes. And so humans exploiting humans is where we've been for a long time. Now, we have computers exploiting humans because the computers go and seek out all these bugs that humans wrote. In the next phase, it'll be machines versus machines. And so I think the nature of cyber is gonna completely change. Probably in the next five or six years, there'll be so much reason to rewrite all of the software that runs the world. In one part because you're gonna be asked to show more operating leverage and revenue growth, but in another part because everything else that was handmade in the past is just fundamentally insecure. Either way, all roads will lead to all the operational software that runs the world will get rewritten. More and more of it will be written by machines. More and more of it will be impregnable as a result. But then the cyber threat actually will only increase because then you're gonna try to figure out how to use a machine to inject something into another machine so that some agentic loop injects some malware or injects a bad token, and I think that's a very complicated thing. What I will tell you is, is... I, I'm not even sure if I'm allowed to say this, but a very good, probably the best cybersecurity company in the world, run by one of the very best CEOs in the world-

    26. JC

      [laughs]

    27. CP

      ... who may or may not be speaking at Liquidity-

    28. JC

      [laughs]

    29. CP

      ... would tell you that they have penetrated and can essentially manipulate every model.

    30. JC

      Sure.

  4. 31:03 – 41:00

    Elon vs Sam Altman lawsuit

    1. JC

      In other OpenAI news, Musk versus Altman, the trial of the century, or maybe the decade, has started. Elon is of course accusing OpenAI of breach of charitable trust, unjust enrichment. He's accusing OpenAI of essentially flipping a nonprofit into a for-profit. He's seeking 150 billion in damages, that they revert back to a nonprofit, that Altman and Brockman be removed. And there were some fireworks between Elon and the OpenAI lawyers. Elon kinda leveled up the discussion. He said, quote, "If we make it okay to loot a charity, the entire foundation of charitable giving in America will be destroyed. That's my concern." Obviously, there's a ton of interesting nuances here, specifically Greg Brockman keeping a diary where he was journal maxing his plans, uh, like a Bond villain here, and, uh, the excerpts from his diary include, "Conclusion: We truly want the B corp. The true answer is that we want Elon out. If three months later we're doing B corp, then it was a lie. Can't see us turning this into a for-profit without a nasty fight. I'm just thinking about the office, and we're in the office, and this story will correctly be that we weren't honest with him. In the end, it's still about wanting a for-profit, just without him," yada, yada, yada. Friedberg, your thoughts on this case. Is Elon gonna win?

    2. DF

      I just don't know why Greg Brockman's got a friggin' diary where he's, like, literally documenting... I mean, I love the guy, but what the [beep] is he thinking? [laughs] Like, you're just sitting here at home-

    3. JC

      Yeah

    4. DF

      ... and like, "Let me write about the, the crime I'm committing," or, "Let me write it, like... And let me record it. And by the way, let me never delete it." I don't understand this.

    5. SP

      It's not just journal maxing, it's discovery maxing.

    6. DF

      [laughs]

    7. JC

      [laughs]

    8. CP

      Discovery maxing.

    9. SP

      It's, uh, Smoking Gun maxing.

    10. DF

      I don't get it.

    11. JC

      [laughs]

    12. DF

      I don't get it, man.

    13. SP

      [laughs]

    14. DF

      I mean-

    15. JC

      I'm-

    16. DF

      Did you guys remember from The Wire that scene where [laughs] the guy's like, "Is you taking notes on a criminal [beep] conspiracy?"

    17. JC

      Yeah, totally.

    18. SP

      Yeah. [laughs]

    19. DF

      This has got everybody in the room.

    20. JC

      Totally.

    21. DF

      Can we play that clip?

    22. JC

      Yeah.

    23. DF

      It's like-

    24. JC

      Yeah. Find that clip, Nick

    25. DF

      ... what the fuck are you doing, Greg?

    26. SP

      [beep] is you taking notes on a criminal [beep] conspiracy? [paper crumpling] The [beep] is you thinking, man?

    27. JC

      If you're going to commit a crime, you do not write down the date and time-

    28. DF

      Can I just ch-

    29. JC

      ... of the crime in your journal.

    30. SP

      Well, look, we don't know it's a crime. Let's not, uh-

  5. 41:00 – 52:44

    Big tech smashes earnings, Capex explosion

    1. SP

      those two.

    2. JC

      All right. Big tech smashed their [laughs] earnings on Thursday. Google, Microsoft, Amazon, and Meta all reported. I don't know why they do this on the same night, folks, but they do, and performance was spectacular. It was great. However, the CapEx announcements were really the story here. Let me just queue this up and show the chart. $725 billion in CapEx guidance in 2026 from but four companies: Amazon, Microsoft, Google, and Meta. Amazon leading the pack with $200 billion. $190 billion each for Microsoft and Google. $145 billion for Meta. You add Groq, you add OpenAI, uh, and some other players to these plans, and we haven't heard from the new Apple CEO yet, but he's gonna be taking over, and he's gonna have some plans here I'm sure. We are going to see the lar- a trillion dollars, a trillion dollars in build out over the next year. I don't know if this is even possible, but this is all being driven by AI and cloud computing. Google Cloud, which includes the Google Suite, that grew 63% year on year. Let that number sink in. 63% on $20 billion in revenue. That's in a quarter. Microsoft Cloud, that includes Azure, Windows Server, SQL Server, they bundled some things together there to get the number to go up, uh, that grew 30% on $34.7 billion in revenue. Amazon Web Services, the original cloud, that grew 28% on $37.6 billion in revenue. That's a bit of a pure play, just counts Amazon's web services. Obviously, these are all moving to neo clouds. These are all serving AI jobs and tokens now. They have a massive customer base, and the customers from the smallest startups all the way to the biggest frontier models cannot get enough compute, and it is going to the bottom line.But this is shrinking Chamath cash flow massively. These were free cash flow machines, the largest money printing machines in the history of humanity, but they are giving up on free cash flow, stock buybacks, and dividends, and the focus on those three to invest in infrastructure. Amazon's free cash flow down ninety-seven percent. Google, Microsoft, and Meta down twelve, twelve, and eight percent respectively. Your thoughts on this free cash flow, the end of the free cash flow deluge, and the massive, massive investment we're seeing in Capex, Chamath.

    3. CP

      I think we're seeing a very important structural shift in the capital markets. I think the last twenty or thirty years, well, twenty years, it's been that the Mag Seven just kind of ran away with it, that these big companies got bigger and bigger, and it absorbed all of these investment dollars. And the biggest reason was that it had these very asset-light business models, right? You just build some more software, and it just has all this leverage, and it all just kind of worked. Except maybe for Amazon because they needed physical infrastructure for warehouses and delivery and whatnot. But by and large, it was a very asset-light investment cycle. Now, all of a sudden, the pendulum is swinging violently in the other direction. And there's something that I think people misunderstand, which is as it moves back to these asset-heavy infrastructure investments, the hyperscalers are signing checks that, I mean, I suspect their body can cash, but there's a world in which they can't. I'll give you an example. You know, when Microsoft convinced the owners of Three Mile Island to turn their-

    4. JC

      Reactor

    5. CP

      ... nuclear site back on-

    6. JC

      Yeah

    7. CP

      ... do you know what their forward purchase agreement was? It was for more than two x the prevailing spot rate for energy. More than two x. The problem is that's not for an enormous percentage of their overall energy needs. So if you play that out and you think these five or six companies all of a sudden are not just spending, Jason, seven hundred billion a year of Capex, which they are, but then from an operating cash flow, they're gonna be spending two x the prevailing spot rate because they just want guaranteed demand into the future, where's all this cash gonna go? It's not gonna go to the shareholder, and it's not gonna stay on the balance sheet. These companies will now get levered. They're gonna get highly sophisticated around the financial engineering. They'll have more debt. They'll have all kinds of different vehicles and term loans and revolvers and all of this stuff. And so they're gonna look like this big, bulky industrial business in five years. And I'm not sure that there's a good valuation case to be made at that point. And so I think it may be simpler, and this is what I tweeted, to just follow the dollars, like a trillion dollars a year going out of the hyperscalers. Where is it going? Just follow those dollars and buy those companies because those companies are already underpriced.

    8. JC

      This is, uh, obviously reminiscent of something we all experienced. Uh, Nick, can you pull up the Cisco chart I just sent you and put it at max? Uh, we had a massive build-out of the infrastructure of the Internet in the late nineteen-nineties and into two thousand. And what that caused was a lot of aggressive companies to do massive amounts of spending, a lot of retail investors to embrace these stocks like we're seeing with people trying to get into these private companies. And, [chuckles] Sacks, look at the two thousand peak of Cisco. This is the most extraordinary chart ever. It took them twenty-five years to get back to that peak, and, uh, they had a lost two decades, and we had a massive amount of fiber that wound up getting bought. We talked about that a couple years ago on the program. But there's something for you to build off of here when you look at this massive infrastructure. You think it's gonna be Cisco systems all over again, WorldCom, et cetera?

    9. SP

      No, I really don't. The issue we had in two thousand was dark fiber. You had all this infrastructure being built out, and it wasn't being used. There's no dark GPUs today, as, you know, Brad Gerstner likes to say. So what's driving the Capex now is the voracious demand for compute, for tokens, and the demand is now pulling forward this additional, um, investment in infrastructure. So I think what's happened here is that the bull thesis for AI just got validated in a single afternoon. I mean, again, you got Microsoft Azure, Google Cloud, Amazon AWS, Meta. They're all basically exceeding expectations, exceeding guidance in terms of where their cloud revenue would be, and therefore how much they're gonna reinvest in Capex. This year, I think we were supposed to have six hundred and sixty billion of hyperscaler Capex up from three fifty last year. I think there's now the new estimate is, is gonna be over seven hundred. So this is, you know, again, it's more than two percent of GDP. This is a huge tailwind to GDP. There's another article saying that I think in the last quarter, AI was seventy-five percent of GDP growth. And by the way, this is just the Capex part. This is the physical infrastructure. This is not the economic impact of the tokens that are generated inside the token factory. This is the building of the factories. How do those tokens get used? Like we're seeing, they're being used not just to do research or to answer questions, but to create code. And so we're seeing this explosion of productivity in software development, and we're seeing an explosion of bespoke software being created, and that's gonna accelerate every part of the economy. Every business that now wants to get code will be able to get code for the first time. Before, they couldn't even hire the engineers they needed to generate it. Now they will be able to. So that is a huge unlock of productivity across the economy. Then you're getting into these new use cases like the, the co-working use cases and agents, right? So the, the workflow automations that are happening.It's still early. I don't believe that this is gonna replace humans. We had that, um, in the past week, we had that crazy case of an agent deleting a production database in nine seconds-

    10. JC

      So great [laughs]

    11. SP

      ... because of, because of a bug. Look, what that said to me is that it's not that agents aren't valuable, they are valuable, but they have to be supervised. You know, this idea that you're just gonna be able to like automate all the jobs away, it is a massive amount of hand-waving over the real technical problems and issues. The agents have to be supervised. Someone has to be accountable. It's not gonna be the CEO. The CEO doesn't wanna be a- accountable for thousands of agents. You need people-

    12. JC

      Yeah, despite what Jack, [laughs] Jack Block said.

    13. SP

      Yeah. You're still gonna need a-

    14. JC

      6,000 direct reports is a great like goal, but it's not realistic.

    15. SP

      You need-

    16. JC

      Yeah

    17. SP

      ... you need-

    18. JC

      Yeah

    19. SP

      ... IT people who are savvy, who can supervise this and make sure-

    20. JC

      Yeah

    21. SP

      ... it's working. They have to be accountable to the CEO. Someone has to drive the productivity. It's like Balaji always said, AI is not end to end, it's middle to middle. You have to have someone to do the prompting, and you have to have someone do the validating, and I would add the supervision and accountability. So anyway, the larger point though is I'm speaking to the fact that I don't think there's gonna be this huge job loss associated with this productivity boom that we're gonna get. And in fact, I think what's actually happening now is that AI is becoming synonymous with the American economy. I mean, the fact that it's generating 75% of GDP, you have this Capex explosion, this energy explosion that feeds it, and again, just the beginning of the applications that are being unleashed by these new token factories, I think it's all a very, very positive thing, and all these doomers who are trying to throw a wet blanket on it or constantly scaring the daylights out of people, I mean, what do they want the American economy to do, just to stop? I mean, they just don't want any progress? I mean, like again, you know, when you talk about stopping AI or halting AI progress, what you're really doing is stopping the American economy now.

    22. JC

      Yeah.

    23. SP

      You're basically saying you don't want economic growth. AI is now synonymous with the growth of the American economy, and if there's no economic growth, there's not gonna be money to pay for all the social programs, there's not gonna be money to pay down the national debt, there's not gonna be money to basically build up our national defense, all these things we wanna spend money on. We have to have a vibrant economy, and that is now syn- synonymous with AI. So I know that AI may not be popular, I see those polls, but having a strong economy is popular-

    24. JC

      Yes

    25. SP

      ... and I believe that those things are now synonymous.

    26. JC

      It's almost like there was some architect or czar who set up the chessboard in the first year of this to make sure that it was ultra-competitive. Uh, well done.

    27. SP

      You know, President Trump set the table on this.

    28. JC

      Absolutely. With some good advice, I think maybe. Friedberg, your thoughts?

    29. SP

      It's always good to have good advisors.

    30. JC

      It's always good to have good advisors.

  6. 52:44 – 58:33

    Vibecoding nightmare: AI deleted someone's codebase

    1. JC

      this stuff, and we had an interesting story referenced earlier in the show where, uh, [laughs] Claude ate somebody's homework. This is the nightmare of all nightmares. Somebody was vibe coding, uh, it was the founder of Pocket OS apparently. They make software for rental car companies. He was using Opus 4.6 through Cursor's AI platform, their coding platform, and uh, you know, which is like the most expensive tier, uh, and he said he configured it with enough safety rules. But the agent was working on a routine task. They saw some sort of credentialing mismatch, and they decided to fix the mismatch by deleting a Railway volume without user confirmation, and, uh, they pushed the code from a repo to a live app, and they deleted everything, including the backups. Literally a scene from Silicon Valley's HBO clip of Son of Anton. Hilarious.

    2. DS

      You gave your AI permission to overwrite code in the internal file system? Were you gonna tell me about this?

    3. DS

      No. I thought that was the company policy these days.

    4. DS

      Okay. Well, your AI just failed epically.

    5. DS

      That's unclear.

    6. JC

      [laughs]

    7. DS

      It's possible that Son of Anton decided that the most efficient way to get rid of all the bugs was to get rid of all the software, which is technically and statistically correct. But artificial neural nets are sort of a black box, so we'll never know for sure.

    8. JC

      How did they get that so right, Sachs? [laughs] Five or six years ago, artificial neural networks are a black box, so I guess we'll never know, but technically it was correct. Friedberg, when you blow up a hollow system with your vibe coding, which you were absolutely showing off in front of Jensen a couple of weeks ago about how much code you're pushing, who are you gonna blame? You gonna take responsibility yourself? Are you gonna blame Cla- Claude or Cursor? Who are you gonna blame when you blow up the entire stack over at Ohalo? Who you blame, Friedberg?

    9. DF

      All right. Yeah, blame Dario.

    10. JC

      You blame Dario.

    11. SP

      [laughs]

    12. JC

      Okay, that's what I thought. That's the correct answer. Correct answer. Blame Dario. He's the one who says it's a doomsday machine. Uh, come on the pod any time, Dario. 17 invites. [laughs]

    13. DF

      [laughs] On the pod any time, Dario.

    14. JC

      I might have invited the guy like 17 times. He is totally going to be... He wants nothing to do with this podcast.

    15. SP

      Actually, let, let me speak to that. So I think-

    16. JC

      Please

    17. SP

      ... I think that, um, there's maybe a misperception that this error occurred because of, you know, quote-unquote, "AI scheming."

    18. JC

      Hm.

    19. SP

      Like kind of in that video that the AI decided that the best way to get rid of bugs is to basically eliminate the codebase. This is kind of like the, you know, AI's gonna turn the world into paperclips type thing, where somehow it'll, like, misscheme. That's not really what happened here. This is a case of just, uh, of old-fashioned bugs occurring at an edge case. You know, you've got the fact that this API was not designed for permissioned usage. You've got the fact that a credential was left kind of lying around. Probably it should not be. There's kind of like a perfect storm that caused a AI to do something or the agent to do something that didn't quite understand it w- what it was doing. I think that if there is a systemic problem here rather than just kind of a, a d- like a random edge case, it's that AI still doesn't know what it doesn't know. You know, like a human would stop before deleting a production database and just say, "Oh, I'm about to do something, like, really serious, really destructive. Am I sure I wanna do this?" You know? And a human would have stopped and said, "Oh, wait a second. Like, I need to be more confident in what I'm doing before I take that action." And AI still has this issue where, again, it can be kind of overconfident. This is where, like, the hallucinations come from, is it doesn't know when it should have a low confidence in its output, right? But this is why it has to be supervised. You know?

    20. JC

      Yes.

    21. SP

      The longer the time horizon for a task, the more likely it is to go off the rails.

    22. JC

      It'll drift.

    23. SP

      And... Adrift, exactly. And this is why I think people are starting to realize that this idea of eliminating all software developers was the peak of inflated expectations.

    24. JC

      Yes.

    25. SP

      Right? There was actually a really good tweet on this by Aaron Levie, who's got the right take on this. Aaron retweeted Matthew Iglesias, who sort of sardonically tweeted that, "Five months in, I think I've decided I don't want to vibe code. I want professionally managed software companies to use-"

    26. JC

      [laughs]

    27. SP

      "... AI coding assistants to make more, better, cheaper software products that they sell to me for money."

    28. JC

      Just lower your prices, don't make me vibe code, is the translation.

    29. SP

      Yeah, I mean, I think, like, rare win for, for Matt Iglesias there. Anyway, Aaron Levie then says, "Agentic coding is a huge boon for software developers that wanna get more done, and it's fantastic for anyone curious to learn how to start coding. What it's less great for is casually building complex software that you have to maintain on an ongoing basis and take all the risk for. Upgrades, maintenance, keeping up to date with the latest security issues, you know, the bugs, cyber. Those are taxes on most knowledge workers who aren't familiar with the system."

    30. JC

      It's not a tax. It's a huge risk.

  7. 58:33 – 1:06:34

    Retatrutide craze: peptides go mainstream

    1. JC

      opportunity. Hey, uh, Friedberg, you have become Retatrutide curious. You have-

    2. CP

      Oh

    3. JC

      ... and also you.

    4. CP

      Tell me. Tell me about Reta, Friedberg, 'cause I want it. I wanna get on it.

    5. JC

      Mm, mm, mm, mm, mm.

    6. CP

      I wanna use it, and I need you to tell Nat that it's okay for me to take it.

    7. JC

      I, I have a friend who has some advice as well.

    8. CP

      Over to you, Friedberg.

    9. DF

      The coverage is coming out of this Phase III clinical trial data release that Lilly put out last month, so everyone's going crazy over the data, which continues to show pretty amazing results. So unlike tirzepatide, which is kind of Lilly's main product today, it's a, which is a dual agonist. It's got two peptides in it that, that bind to different receptors, the GLP-1, the GIP receptor. This other one now b- also binds to glucagon, which is a third receptor, and that glucagon receptor binding peptide causes the cells to increase their metabolism, which actually accelerates fat energy consumption over what would typically be muscle energy consumption. It's more likely to burn up fat early on, which causes more quick fat loss, but also reduces muscle loss. And some of the other data that's now coming out shows non-HDL cholesterol down 27%, triglycerides down 41%. A-

    10. CP

      Liver fat down 80%.

    11. DF

      The 80% reduction of liver fat. A1C drops from 7.9% to 6% in 40 weeks, which is amazing, by the way. If you're diabetic and your A1C drops that much in a couple of months, it's literally a lifesaving product. The average user in this Phase III trial saw their weight decline from 214 pounds, they lost 37 pounds. That's compared to six pounds on placebo in 40 weeks. And, you know, modest side effects. 20% people felt more nauseous than the people that were on the placebo. There's a lot of other separate studies that are being done now that are showing significant reductions in inflammatory signaling molecules. So systemic signaling of, like, hey, cells are in distress, triggers this kind of inflammatory process that can have a lot of other damage to your body. It can accelerate aging. And so one of the other conversations is that retatrutide might actually be kind of a de-aging drug as well.

    12. JC

      Oh.

    13. CP

      Hm.

    14. JC

      Hercules, Hercules, Hercules.

    15. DF

      Yeah. They're rapid. They're, they're... You know. And a lot of the studies, by the way, are done on the, the, the very high dose, 12-milligram dose, but you could probably get this thing dosed down to two milligrams and still see a lot of the anti-inflammatory-... maintenance and other benefits. I'm no doctor, but people are going nuts over this being more widely useful than just for clinical obesity or type 1-

    16. JC

      When will the FDA approve it

    17. DF

      ... or type 2 diabetes.

    18. JC

      When's the projected date, Friedberg?

    19. CP

      2027. Mid '27.

    20. DF

      That's what they're saying. Could happen sooner. I mean-

    21. CP

      Hm

    22. DF

      ... the data's in. The, you know, the FDA will take their time to evaluate it, but I think given the way this is all looking, could happen sooner, could happen some time later this year.

    23. JC

      Swim, Chamath, Swim said it's incredible and that, uh, it's living up to the hype in their experience.

    24. DF

      Who? Who?

    25. JC

      Swim.

    26. DF

      What is that?

    27. CP

      What is that?

    28. JC

      Someone who isn't me. Swim.

    29. DF

      Oh.

    30. JC

      This is a Reddit term. Someone who isn't me said, who has a guy, Swim has a guy, and has cycled, uh, on Retatrutide and does pushups and says muscle gain has been spectacular, no muscle loss, and a lowering of fat.

  8. 1:06:34 – 1:20:54

    Friedberg's Supreme Court experience

    1. JC

      Friedberg, guys, Friedberg had his own personal Super Bowl. You see me getting ready for Nick's playoff season. I get my courtside. Friedberg had the equivalence acts. He went to the Supreme Court in order to hear them talk about chemicals. This was a big deal for him. The Supreme Court coming together. Did you wear a-

    2. CP

      Oh, Monsanto. Was it the Monsanto?

    3. JC

      Yes, the Monsanto trial happened in the Supreme Court, and he went. He got courtside. He went to the Supreme Court and listened in the building.

    4. DF

      Have you guys ever seen a live Supreme Court hearing?

    5. JC

      No. We, we-

    6. CP

      No, I'd love to, though. I'd love to.

    7. DF

      Sacks, have you been?

    8. DS

      No, I haven't, actually.

    9. DF

      I mean, honest- honestly, I think it was one of the most amazing experiences I've ever had. There was a massive protest out front. We went through the marshal's office to get in. And that building, you walk in, it's, like, sacred. It's all marble. It's, you're not allowed to talk. You have to be super quiet when you're in the building. Like, they keep going, "Shh, shh, shh," like you're in some quiet library.

    10. JC

      I'm out.

    11. DF

      It's like people treat it with this level of kind of sanctity-

    12. CP

      Respect, yeah

    13. DF

      ... that, that, and respect, and they're like-There is no politics here. There is no bull [beep] . There is no freedom of speech. This is the court. When you come into this court, the justices tell you how you will speak, how you will behave, what you will do, and you will not speak, like, unless spoken to. You put all your stuff in a locker, you go up the stairs, you go into the, the courtroom. And the courtroom, it's just so amazing being in there. They have this amazing marble frieze above the justices that has some of the great people of human history, Moses, and these kind of amazing historical figures, and then below them are the nine justices. And the court case, if you guys haven't watched the case, you can listen to them, I think, online. So basically-

    14. DS

      Yeah, they're all online. It's worth listening to.

    15. CP

      Hold on, hold on. Wait, wait, wait, I have questions. So does Roberts sit in the middle 'cause he's the chief?

    16. DF

      Yes. Yes.

    17. CP

      And then do all of the right justices sit on the right?

    18. DF

      No, they're mixed. So they're, I think, um-

    19. CP

      It's mixed

    20. DF

      ... I don't know, I don't know the exact seating, but they're mixed in terms of-

    21. SP

      It's probably based on appointments to the court. Is that right?

    22. DF

      Yeah, I think that's right.

    23. DS

      Seniority. Seniority.

    24. DF

      I think that's right. And then, so yeah, that's right. App- and then it kind of goes out from the middle with Roberts in the middle. Roberts occasionally will name the justices and say, "Hey, do you have a question? Do you have a question?" If no one's talking, but otherwise the justices will jump in with their questions when they want, and they'll ask. Now, honest to God, watching this is like watching LeBron James play basketball. These lawyers are so mind-blowingly impressive on both sides that you would just, like, sit there, and I was, like, in awe. It was so... I, I felt like my energy was completely sapped from me at the end of this process because you are just so engaged and so caught into the way that these guys are thinking and talking. And if you're-

    25. CP

      Did you take a Rose Sparks? You take a Rose Sparks when you were there?

    26. DF

      No.

    27. DS

      [laughs]

    28. DF

      And if you're familiar, if you're familiar enough with the case or the case history or the law that's being debated... 'Cause again, when, when you get to the Supreme Court, you never debate the case. What you're debating is the legal interpretation of the, the decisions that were made on the case. And so is this constitutional? How do you interpret this particular act, this law, this federal law? What's the right way to think about it? So you don't actually talk about the case. You talk about the interpretation of American law, of our laws, of the Constitution-

    29. CP

      What was the specific case?

    30. DF

      ... of the Bill of Rights. Yeah.

Episode duration: 1:20:57

Install uListen for AI-powered chat & search across the full episode — Get Full Transcript

Transcript of episode fpC4sbawSzQ

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome