IPO 与 SPAC 回归，Mag 7 大对决，Zuck 上头，Apple 失手，GENIUS Act 通过参议院

All right everybody, welcome back to the number one podcast in the world. I'm your host and executive producer for life. Isn't that right, Dave Friedberg, JCal, not at all what you are. Make sure you tune in startups and apply to Founder University. You're something very different. With us again today, the Sultan of Science, David Friedberg. Can I just congratulate you on your 4th baby? If you double that number, you're going to be able to catch up to Chamath and his 5 plus 3 illegitimate children. How are you doing? You're tired and grumpy, aren't you?

It's a little transition for me. I didn't have to do the work.

Are you tired and grumpy? And how's Allison? How's the baby?

Everyone's wonderful. Thank you for asking. And a beautiful boy.

Beautiful. Nothing is more amazing than seeing a child. How's the baby?

Magnificent. Thank you for asking. Thank you for all the kind words.

And we sent over a gift basket, Chamath and I. Longhorn Pana Stakes, a 10-year membership for—oh, hey, congrats to Olivia Landon, by the way, of Long Hill Wagyu. She had twins. That means she's going to have more people to work on the ranch and slaughter cattle to send us our picanha. Congratulations. Shout-out. Congrats to Olivia Landon. It's so funny because we love these steaks so much. She doubled. We mentioned it on the pod and you idiots started searching for it. Lunatics, and they ordered out all the coulotte steak. So now Chamath and I are screwed. No crew. No, they ordered out everything. Everything was sold out. So now we have to gatekeep.

Okay, let's move on. With us again, your chairman-dictator, Chamath Palihapitiya—he of 2 votes in our fine organization. How are you doing, Chamath?

I love voting control. I'm doing great.

He starts Thomas Laffont with a tie, and then all of the gamesmanship happens between the team of rivals: me and Friedberg. With us again, Thomas Laffont, a gentleman and a scholar. I have no idea why he's here or how he wound up on this podcast, but he's a true gentleman, a true scholar, and the host of East Meets West, an incredible conference that I attended this week with our bestie David Sacks, who, of course, is at the White House and can't join us.

Thank you for including me. No box lunches, by the way. We took your feedback from a couple of years ago, so I hope that we met your standard.

You did. What were the highlights for you guys at your conference, Thomas?

I think, for me, obviously, there was a lot of news in AI this week. That was the centerpiece of most of the panels, pretty much up and down the stack, from SaaS companies trying to transform into AI to the big Zuck news on Scale AI, and then potentially—I saw in The Information today—the Nat Friedman news. It feels like there's a lot going on in the industry, so it should be fun to talk about.

We're going to talk about it all today. We got a really full docket. Rick Caruso, the mayor who would have saved Los Angeles from the fires, was there, and you actually hosted at his incredible facility.

We did. We talked about the state of LA. JCal, is that where you're at, right?

Yes. I'm at my LA home, aka the compound. I'm here in LA. But yeah, Rick Caruso—what a great speaker. Interestingly, today a friend just sent me a chart showing the recovery of restaurants post-COVID, and LA is 50% behind on the recovery per store location versus the national average. What do you attribute that to, or what did they attribute it to?

I think there are a couple of different things. One, the economy, which, unlike the San Francisco economy, being levered to AI and on the upswing, is more levered to entertainment. And I think there's a secular decline.

Someone mentioned at the conference that filmings in LA are down 50% from peak. That's just a massive move down, losing share to other geographies, both in the US. Georgia, I think, was mentioned. Ted Sarandos has explained exactly how aggressive New York is being, the UK is being, and Atlanta. I mean, so many different hubs for movies are giving much better deals than Los Angeles is.

Yeah. I think it's a combination of being levered to one industry that's in secular decline. I can tell you from MrBeast: for Beast Games, we had a deal in Las Vegas and in Toronto, and we got huge tax credits. In the 2nd season that we're doing for Amazon, we did an enormous deal with the Kingdom of Saudi Arabia. We're filming a bunch of episodes there, we're building the sets there, and we're actually going to keep them there after it's all said and done.

We would not film in Los Angeles unless we absolutely had to. We will stay as far away from California as possible. Regulations are such a big part of this. It's uneconomic. You can't make it work. Thirty percent more expensive, I think, is the official number.

But there's also speed, right, Thomas? How quickly can you stand something up? How much paperwork do you have to file? The James Beard Foundation, I'm seeing here from the research, has found that all these independent restaurant owners said they just can't get staff here.

In Los Angeles, it's just hard for people to live here, and it's hard to get through the regulations. If you make it hard, there are other options for people. This idea that California has a lock on anything other than incredible weather and beautiful people is farcical. There's a lot of beautiful people in other places with decent weather, and you can go do your productions there.

Another topic that came up that a lot of people were talking about—something that I know you've talked a lot about—is our debt issue and the debt-to-GDP ratio. There was a lot of talk on the flip side, on the GDP side. What if AI can increase productivity and regrow GDP faster than expectations? Perhaps that's one of the reasons why interest rates might not be quite as high as you might expect, given some of the trends that you guys have talked about.

There were a lot of discussions around AI productivity and what we could look at over the next 5 to 10 years because of the improvements we're seeing.

This is particularly beneficial to the US, right? If you think about where AI is going to accrue economic surplus first, it's likely going to be in the US, not global GDP. So the US either captures dollars or increases overall productivity, or both, ahead of the rest of the world.

If we do see advances from AI accelerate GDP growth, is that because of all the onshoring of manufacturing and industry that we outsource today? Do you think that goes hand in hand with AI acceleration?

I think that's part of it, and I think the other part is just getting significant productivity improvements even out of the knowledge-worker workforce.

One of the things that we showed in our keynote is the adoption of these technologies. Even taking doctors as an example, there's this new company, OpenEvidence, coming in and developing a diagnostic engine that's now used by a 3rd of doctors. Already, a 3rd of US physicians are on the platform, using it 10 times a day to help with diagnosis.

In oncology, as an example, it's seen significant traction. Multiply that by the legal profession and coding. I think we're already seeing it. What if we just see an explosion of productivity gains across both the physical and the digital economy?

Yeah, the doctor one's a good example. If someone had the opportunity to go get more regular preventative checkups, they would. The problem is that it's very expensive, it's hard to get an appointment, or insurance won't cover it.

But if the cost to a doctor goes down because they can leverage AI, throughput goes up by 10x. They can see 10 times as many patients per day, and then suddenly diagnostic care becomes more available. They can charge for that; they don't need to charge the same amount. The price will come down per checkup, but more people will be able to get a checkup per day.

So that grows GDP in diagnostic care. That grows the size of that piece of the economy. Anything where AI provides leverage to a service provider and their throughput now goes up, there's an example of that.

Dave, there was an LA dentist that went viral this week. I don't know if you guys saw this story, but he created an ad using V3 about a skydiving gorilla who ultimately needs to get his teeth fixed because he was drinking while he was jumping out of the plane.

It's a very funny viral ad. He probably made it for a couple hundred bucks, and now his practice is totally full. He's been flooded with requests for new dental implants. To your point about increasing productivity, boom—there's how V3 can help a local dentist.

We got a full docket, but we're going to rocket the docket because there's so much going on here. Zuck is tilted, clearly. This has been the big discussion in Silicon Valley for the last 10 days or so. According to reports, Zuck is super frustrated that Meta is falling behind in AI, so he is swinging for the fences.

Sam Altman said Meta has offered top OpenAI employees a $100 million—wait for it—signing bonus. That's not comp; that's a signing bonus. Who knows if this is true or not, but he's also offering $100 million a year in annual comp. He's clearly cut out tens of billions of dollars for this effort, not dissimilar to when he did his VR efforts that didn't work out so well. Here's a 30-second clip of Sam Altman talking about this on his brother Jack's podcast, Uncapped.

Sam Altman

They started making these giant offers to a lot of people on our team: $100 million signing bonuses, more than that in comp per year. Crazy. It is crazy. I'm really happy that, at least so far, none of our best people have decided to take them up on it. I think that people look at the two paths and say, “All right, OpenAI's got a really good shot—a much better shot—at actually delivering on superintelligence, and also may eventually be the more valuable company.”

Meta also just invested over $14 billion—I’m using “invested” in quotes—in Scale AI for a 49% stake, and this probably is better described as a shadow acquihire to get around antitrust scrutiny. You remember Microsoft did that with Inflection AI back in the day. Google did it with Character AI, and Amazon did it with Adept AI. I'm not sure if this is necessary anymore, since Lina Khan's no longer in the position.

Scale CEO Alexandr Wang and others will be joining Meta to work on a new superintelligence team. They're saying that Scale is going to remain an independent company and get a new CEO. Not sure if that's going to happen. If you don't know, Scale does data labeling. They get experts to help train language models. Two of their biggest customers are OpenAI and Google, and they both canceled their contracts.

So Zuck is taking that chess piece off the board so he can get all that data into his LLMs. He's also reportedly in talks to hire former GitHub CEO Nat Friedman and Daniel Gross to work on AI. They have an incubator investment fund for AI. Daniel Gross had a really cool startup incubator called Pioneer Labs. I had him on This Week in Startups a couple of years ago. Really smart cat. Meta has $70 billion in cash.

Thomas Laffont, when you see Zuck doing this, what's your take—not only on what Zuck's doing, but how big of an opportunity is this in terms of the prize of having the best large language model? What is he going for here? And what's your take on these really aggressive packages and 49% purchases?

I mean, look, I think, one, it feels highly rational, right? If you think about it, Meta's market cap is—rough math—$1.7 trillion. If you're the CEO and you ultimately believe that maybe 50% of your market cap is at risk because of AI—$850 billion—why would you not spend maybe 4% or 5% of that if you think it increases the odds even slightly that you're going to win the market?

So, to me, it kind of reminded me of a few things. Number 1, the scale and size of the opportunity. Obviously, people think AI is massive, but frankly, Jason, I'm even wondering, putting the regulatory scrutiny to the side, if it was time—he just didn't want to wait. And obviously, doing it this way, I think Alexandr Wang, literally the next day, who's the CEO of Scale, can show up to work at Meta. So I think it's urgency around a large opportunity.

I'm curious to get Chamath's take, because it reminded me a little bit of the pivot away from HTML5 and also a much smaller acquisition, but one that we really felt, of a company called Onavo. For those that may not remember, Onavo was a small data service provider, but what it did was have a panel of phones, and we as investors could see which apps people were using. The data was incredibly valuable because it was the only service that gave you true engagement data.

So obviously, as an investor, you felt, “Wow, this is an incredible tool.” Eventually, it sold to Facebook, and Facebook used it internally and didn't allow anybody else to use it. We lost one of our key abilities in the mobile app revolution to tell who was winning and losing.

So you're saying the Scale acquisition parallels that in a bit: there's this great service a lot of people rely on; he buys it, shuts it down for everybody else, gets the tool for himself, and gets the data for himself.

Correct. I definitely see parallels, and given their market cap and the size of this opportunity, I think it makes a lot of sense.

Chamath, your thoughts on Zuck's action? Obviously, folks know you worked with him as he went from tens of millions of Facebook users to hundreds of millions. You were there, actually, during the HTML5 wrapper app disaster.

I think maybe that was a debate at our executive team, at our management team, and I was on the side of apps. Without embarrassing him, somebody else was on the side of HTML5. I thought it was stupid.

Why was that?

Why was that? Because all of my political capital at the time was also wrapped into native apps—our own phone, an entire vertically integrated stack. Politically, I think I made the decision for them very hard because I was not a “play nice in the sandbox with others” kind of executive. I was more of a scorched-earth, get-it-done kind of person.

They made an enormous mistake, but then they admitted it about a year after I left. They said this was the single biggest mistake.

Okay, great. Chamath, explain in plain English why HTML5 wrappers versus native apps.

I can explain it. Native apps were obvious in 2010, and the only reason to use HTML was as an end run around different carriers and different ecosystems that were trying to charge us a toll.

In 2010, I went to Mobile World Congress, and I took a group of my most talented developers. We built an entire replica of Facebook that we called Facebook Zero, which was only available via URL. We launched it at Mobile World Congress, and I announced it there because, if you went to India, as an example, all of the folks there would try to charge us a tax, but if you could navigate through the browser, you wouldn't have to pay it.

That was a good example of what to do in a developing market when people were toll-taking. But the real solution was to build an extremely integrated app, from the software all the way to the hardware. The only way to do that was as a native application, and that has tremendous implications today.

But just to finish on that, my proposition was full phone, full stack, full app. All of this other HTML stuff should only be a side thing that we do in markets where they try to make it difficult for us. Instead, it became politicized, and it became a big bet on HTML5, which I thought was absolutely stupid and unjustifiable.

That was also when I said, “Okay, well, this phone's not going to happen, so let me leave.” A year later, I think Mark, to his credit, said, “This was really stupid,” ripped all the HTML5 stuff apart, went native, and the rest is history.

Yeah, there it is. The biggest mistake was betting too much.

It was, again, people politicizing what should have been an obvious technical decision. The other piece to that was that it was also a religious decision. People liked the open standards of HTML5. Certain developers felt like we had to support them. Only stupid, nontechnical people thought that. It was stupid. It was obvious. There were morons on the executive team who advocated for this. Anyways, we were right, they were wrong, and he was fine.

Okay, fast-forward to where we are today.

It's the exact same story playing out. Now, what do I mean? You have to look very carefully at Microsoft's deal with OpenAI. Why? Because what you see is the compounding of secrets.

There are secrets in the training layer. There are secrets in the model layer. There are secrets in how these things are tightly coupled to infrastructure and compute. What we have to remember is that what OpenAI got from Microsoft was an extremely competent partner that built an enormous Azure compute infrastructure to train everything from ChatGPT all the way up to the o3 model—everything.

Why is that important? Because you start to figure out these tricks: How do you really optimize these models to be extremely performant? If you look at all of the other models, they've also had some level of that advantage.

So if you look at DeepSeek, what did they do? Well, we don't know. But what we have been told is that there's very tight coupling to hardware. If you look at what xAI is doing, I think what you can bet is that there's an extremely tight coupling to hardware, infrastructure, and compute.

If you look at what Meta is doing, they generically train on Nvidia and launch it in the open source. I think what they need to do is more of the OpenAI and Google playbooks. Look at Google. Google's Gemini models are extremely tightly coupled to TPUs, and that enables and unlocks an entire stack of secrets and capability that then get manifested in model quality.

I think the first thing that Mark has to do, if I were him, is start to chip away at all of the sets of secrets. What secrets do you get from Alexandr Wang and Scale? What are the labeling techniques that allow these models to be more and more performant? What labeling techniques are used in the reasoning models? What labeling techniques are used in more traditional LLMs?

It is clear that Llama doesn't know this. Meta doesn't know this that well, because their model quality is meh. So now what you get is that set of secrets.

What do you get from Nat Friedman and Daniel Gross? You get what the apps are doing and how they're approaching writing agents—these agentic tips and tricks that make usability and value more obvious.

But then what's missing? I think the thing that's missing is the infrastructure and compute set of secrets. I think it's insufficient to buy stuff off the shelf from Nvidia and expect these models to fundamentally compete. So, if I were a betting man, he's bought the training secrets, he's bought the app secrets, and now he has to buy some infrastructure and compute hardware secrets. You put it together, and he's got a pretty good strategy here.

And also, just to add to that, Nat Friedman and Daniel Gross have invested in a lot of AI companies, and those companies have secrets of their own. Actually, I think they have some along the full stack.

Friedberg, your thoughts on this strategy as described by Thomas and Chamath, and just the data we're seeing on the playing field: aggressive acquisition of talent and companies?

I don't know if I have much to add here. One additional point, Chamath, by the way, that you mentioned: If we look at the winners in models over the past 12 months, Anthropic is the same. They've been very deliberate and have explained how TPUs—they've been a big user of them—have helped define their training models. So, I think you're 100% right: If we look at the models that have really performed, it's the ones that have that secret, as you mentioned.

When I first started, 8 or 9 years ago, one of the key bets I made—which was a mistake, and we unwound the bet—was: Can we build a transpiler? That is to say, can you take a CUDA workload and redirect it away from Nvidia to different hardware? Basically, what I learned in that process is that all of the attention mechanisms built into transformers that really differentiate how good the models are need to literally be hand-tuned for every single target of silicon that you have.

So, when Amazon just wakes up and says, "Here's this chip," it means nothing unless you can incentivize somebody to build to it. But the opposite is also true. If you have a model and just run it generically, you're not going to get the gains, and it's not going to be as special as if you have a dedicated infrastructure and compute architecture and say, "We're going to tightly couple these." It's been clear that OpenAI has had that, Anthropic has had that, Google has had that, and DeepSeek has had that. I think Meta needs to do that. Otherwise, they're always going to be floundering on their back heel.

One quick misnomer: When people hear "labeling," they assume a photo of a dog and someone saying, "This is a dog." That's definitely how it started, but if you look at serious business use, it's completely more than that. You could actually label the problem. For example, in simple terms, 2 + 2 = 4 is a reasoning data set, right? You've got to think of labeling not just in the simple terms of an image, but of massive data sets of outcomes, and that's what's really used to train these reasoning models.

But I think there's another story here, guys, in my opinion, and it's the performance of the Mag 7. I'm going to have to check with my data science team, but I'm wondering if this is the year where we've seen the greatest divergence among the Mag 7. If I just gave you this performance, you can see: Meta's up 18%, Google's down 8%, Nvidia's up 8%, Tesla's down 20%, Apple's down 21%, Amazon's down 3%, and Microsoft is up 13%. It's kind of interesting: In a market where, historically, over the past few years, we've felt the Mag 7 have been truly correlated, the market is now saying, "Wait, hold on. We might start to see diverging performance."

What I read from that, in one element, is that the market's starting to try to sort out who are going to be the winners and losers, who's well-positioned versus maybe falling behind. I think we're going to start to see some divergent performance from the Mag 7.

Can you put that back up there for a second? I think that's so interesting because if you look at the conditions on the field today, Google's down 8%, but again, I would tell you as a user, Gemini models are exceptional—absolutely, just bar-none exceptional. I think Anthropic is incredible for code generation. Incredible.

What I see is every single company on this list that isn't Nvidia baking and rolling its own silicon, yet Nvidia is up and the rest are down. I told you that I spent time last week at Tesla. I would not be sleeping on this business. I think that it is yet again back into the land of being misunderstood.

The only one that I understand being down this much is Apple, because it's not clear that they're even baking something in private. There's nothing public. There's nothing private. It just seems like they're transitioning into being a cash cow and getting into that cash-harvesting mode. But it's almost weird that the price action is what it is, because I would have thought that Google would be up. Meta would maybe be a little flattish to down. Nvidia is up, but maybe it could be down. Tesla's down, but it should probably be up. Amazon's basically break-even, and Apple is down. I think that kind of makes sense. That's sort of how I read this table.

Yeah, I mean, what I love, Chamath, by the way, is that now there are debates, right? You can argue whether you agree with Chamath or whether you don't; he's spending $20 billion because he's not afraid.

Correct.

Yeah. No, let's pull the chart up again here. I think this is an interesting way to look at it. The only reason Microsoft is not on this list is because of the limitations of the DOS-era interface of the Bloomberg Terminal, where it will only allow you to compare 6 charts and not 7. But we know that Microsoft is up 13%.

When you also look at these, there are some extenuating circumstances here. Tesla's car sales are down, but all car sales are down, and I think that's the piece that maybe isn't being accounted for. They're in a transitional period.

Apple, obviously, has a lot of regulatory overhead. Tesla is losing solar and EV tax credits. Apple is being told to onshore and stop buying from China, so its supply chain is being disrupted because of tariffs. Those 2 companies in particular are far more affected than the rest. Even Amazon—there's been some conversation about the tariff effect on Amazon—but obviously that's offset by some of the benefits they've been realizing and promoting, as Sacks spoke in his letter this week, from AI.

I think there's a variation here that's probably a little bit more tuned to these conditions that aren't necessarily natural market forces, but are influenced market forces associated with the new administration and some of the policy choices being made.

If we were looking at those number 1 and number 2, which one do you think gets to AGI first, Chamath?

Well, wait, hold on. By the way, the other thing you should note, Jason, which I find really interesting, is that nobody talks about AGI anymore. If you listen to the language of all the companies, it's all "superintelligence," which is a much more achievable goal because it's defined as being multiples more intelligent than a human being. I think if you actually did a search for the number of times AGI is being said today, it's meaningfully less, because people have realized that that's not in the offing.

Yeah. By the way, another lens, Chamath, that I think about on these is: Who controls their own destiny among these 7 companies in AI? I would argue Tesla does, Nvidia does, and then it's kind of interesting. Amazon doesn't have its own foundation model. They're kind of dependent on others.

Microsoft does, at 49%, because of this kind of relationship they have with OpenAI. They own a big share, but they don't control it. Maybe 6 months ago we would have said Meta absolutely does. Maybe Zuck's trying to question that a little bit.

It's fun, in my opinion, to bring different lenses to this list. There's the regulatory one that Friedberg was just talking about. I kind of think about: Do I control my own destiny in this market? I expect these companies are not going to want to be dependent on others and are at least going to want to say, "No, I'm going to control my own destiny," whether I win or lose.

Who's your number 1? Who's your number 2? If you could only bet on 2 here to achieve superintelligence—AGI, let's just say, win the AI big prize—in the midterm, 5 years from now, when we're sitting here on All-In episode 700, give me your number 1. Give me your number 2.

Look, to me, number 1, I still think Nvidia. I don't see the GPU getting displaced. I see additional architectures coming on board and growing the market, but at the end of the day, all roads still lead to the GPU for all of these models. So, I would still put Nvidia at number 1.

My number 2 is more of a dark horse, but I would pick Tesla. I do think it has the most potential for vertical integration, all the way from the silicon to the model to the actual hardware. That might become super important, not just in cars but in Optimus. So, Nvidia number 1; Tesla is my dark horse.

Wow, stunning, Chamath. Who's your number 1 and number 2 in the midterm, 5 years from now? We're sitting here on All-In episode 700.

Tesla's number 1 and Google's number 2. The reason is that they are the closest to having that vertically integrated stack that I spoke about.

I think Tesla has the best vision models. Now, with xAI, they'll have one of the best LLMs and reasoning models, and they'll eventually be able to stick that on Dojo. And then all of that will be in all of the physical AI that you will interact with in your daily life, whether it's a robot, a car, or a robo-taxi. So that's number 1.

And then number 2, for many of the same reasons, I think Google, because you'll have the Gemini family of models, which just absolutely kick ass. Veo 3, which we haven't really spoken about, is going to destroy Hollywood in the next year. Hollywood is done, I think, but they're landing model after model. They have the TPU, and the next-generation TPU, I think, is exceptional. They're baking quantum, and then they have an entire funnel of billions of people that they can direct experiences to. So Tesla 1, Google 2.

Chamath, quick follow-up on that. I'm curious about Google because I oscillate a lot on this particular name. Can Google win if search declines?

Yes. And I think what probably has to happen is—bear with me when I say this—but if you had to boil down Google's economic north star metric, not the value north star, the economic north star metric would be price per click. I do think that Google is extremely well-positioned to pivot that to price per token.

I think that they have some emergent classes of physical AI, but they have the largest pool of people where they can generate a price-per-token value framework through YouTube, through Gmail, through Workspace, I think through search, but probably it's a different kind of model. It just requires them to rip the Band-Aid off at some point. But, yeah, I think Google can do it.

I'm going to go with you, Chamath. My 1 and 2 are either Google or Elon. I'll just say Elon because, like you, I spent a day up at xAI and I saw what a magnet for talent he is. I got to sit in some meetings, and he was interviewing people and working with that talent. At 8:00 at night, there were a lot of people there on a Saturday, grinding it out. It was nuts.

When I first went to xAI, in the 15 minutes that I was in the parking lot finishing a call, the kinds of people who were walking in and out of there—you could tell they were big brains. I don't know how, you know what I mean? From every walk of life, they all just looked much smarter than the rest of us. Some of them were chain-smoking cigarettes and stressed out. It was crazy. I hit a couple of Zyns, I'll be totally honest.

But the reason I say Elon versus Google is I think Elon's in a unique position. I don't have any insider information here, and I haven't talked about this or been back-channeling with Elon, lest anybody aggregate this.

I think what Colossus has done and what Tesla has done—both of these things: Tesla, with its own stack of hardware, to your point, Chamath, hardware plus software plus the user application of FSD and Optimus. Then you put that together with the data, the real-time data of X, formerly known as Twitter, plus what he's building with xAI—and obviously those 2 companies merged.

I think the Tesla board and xAI board have to get together and put those 2 companies together. One's worth $1 trillion, one's worth $100 billion. Put them together and just have all that brain power going in 1 direction, as opposed to Elon task-switching between the 2. You do that, I think he wins number 1. You don't do that, I think he either gets 1 or 2, and then I think Google is going to have a better search product.

Chamath, I think it's a really important point. Do they lose search share? It doesn't matter. What I think matters is: are their ads more effective? Is their ad network more effective? And I think, based on what they know about you from your Chat searches and your discussions, and what they analyze in your email—just analyzing your Gmail and your surfing behavior in Chrome, if they get to keep it, your Android phone if you use it, your YouTube list, when you drop off All-In, and when you start listening to another podcast, whatever it is—all that data is going to lead to an ad network that performs so much better that even if they lose search share, their ad network is going to continue to grow. And I think it will increase in velocity.

So those are my top 2. Friedberg, I'm curious from your position. Which one do you think is number 1 and number 2? I saved you for last because you know what we do here? We save the best for last. Friedberg, go ahead.

I think there's a difference in how I would lump them. I think Tesla is probably the best place to invest if you want to have a shot at a massive new industry. They've got a baseline business in, obviously, the automobiles, but I think this humanoid robot opportunity is absolutely mind-blowingly ginormous. And I don't think that there's a better company on Earth positioned to execute against this humanoid robotics opportunity than Tesla. So it's sort of like I would call it a low-probability, high-upside call option embedded within that business. And obviously, you're paying a premium for that because it is still a very healthy premium you pay for that business.

I think Nvidia, to Chamath's point, the common thesis is that it is the most protected. The durability of the business is there. But I would argue that there's actually a low-probability but very high-severity risk to Nvidia in China. There was just a demonstration last month of a 1-nanometer semiconductor manufacturing process out of China.

I think the more that we continue to try and isolate China from a policy perspective, the more we are emboldening investment in China, meaning from the government and from private industry into China, to create alternatives to the chip stack where U.S. companies, particularly Nvidia, have a moat today. So I do think that there's going to be an emergent competitive threat coming out of China to Nvidia. And just like we were knocked over by DeepSeek, I think we will be knocked over by some semiconductor manufacturing processes coming out of China in the near term.

But, overall—by the way, Dave, just on that point, I think Sacks's work on the diffusion rule, just generally, I don't think has gotten enough attention. The rescinding of the diffusion rule, which essentially handicapped our ability to even arm our allies with our semiconductor technology, in my opinion, was a milestone and very important moment to try and offset exactly what you were just describing.

That's exactly right. There was a report a few months ago, and I mentioned it on the show—or maybe I didn't, or maybe I sent it to Sacks and we talked about it offline. I can't remember—but it was about a $40 billion investment being made in developing competitive semiconductor manufacturing full-stack solutions out of China.

So I do think that the lithography IP moat is being crossed in China. I do think that China is developing actually new technology for DUV and EUV systems. I do think that there's a risk to Nvidia's core. Now, look, Nvidia is such a durable business. There are great moats, great advantages, but we're creating every incentive for an alternative to Nvidia to emerge from China.

And then my third categorization would be: what's the portfolio solution? I think that's Google. I think that there's a diversification of high-beta bets inside of Google, any 1 of which could have, call it, a $1 trillion market-cap outcome, ranging from Waymo to quantum computing to the biologics work that Demis is working on out of Isomorphic Labs. There are a number of things that do not get a lot of attention at Google.

So, yes, there's a core business that may be at risk, Chamath, but I think that there's a portfolio of options you get at Google, and you just need any 1 of them to hit to make up for the loss. But I do think also Sundar, in my interview with him, which we put out a couple of weeks ago, is very thoughtful about where search evolves to, and he is being, I think, reasonably aggressive in trying to evolve the search product architecture to meet the market, to meet the consumer. I do give him credit for that.

So Google would be in a good place for me as an overall pick in that set of options.

So, just to be clear: Nvidia 1, Google 2, or Nvidia, Tesla?

Like I said, I think in terms of having the right Sharpe ratio, that's how I would think about it. The alpha- and beta-adjusted returns, I would put Google number 1. I would probably put Tesla 2. Tesla's valuation, I think, already has a premium associated with those options, so I don't know if I would really pay that premium.

I think, aside from the valuations, let's take valuations out of it. The game here is who wins the AI prize 5 years from now. That's how I understood it as well. Valuation irrelevant. Valuation irrelevant. Who wins the AI prize? Number 1, you're saying Google. Number 2, you're saying Tesla.

I think Google's in such a position. I mean, look, Demis, I think, has been fairly coy about where they are. They obviously promote Gemini 2.5, but there's a lot still coming. And as Chamath pointed out, it's not just LLMs. There's a pretty sizable family of models, including a lot of these graph-based models, that are being used in really novel applications that no one else is even close to, no one is spending time on.

Some of the weather forecasting might seem small and trivial, but it's a demonstration of Google's competency in core model development that shows an understanding and a depth of research and work that goes well beyond LLMs. So I'm pretty bullish on the depth of talent, the full stack.

Yeah, yeah. And whatever they learn there could apply to Gmail, could apply to search, could apply to ads, could apply to the YouTube algorithm. It just goes up and down.

Yeah. Yeah. From a product perspective, I do think you see this kind of multimodel emergence that we're now seeing, where no one talks about the single model that sits behind the application. There are multiple models that work together. Obviously, this agentic architecture unlocks another layer—not just solutions to complexity. There's quite a lot, I think, that's emergent here that Google will start to benefit from in the year ahead.

For those of us who love tech, if we step back for a minute, I really feel like, to use the analogy of this podcast, we're now at the WSOP, the World Series of Poker. We have 7 companies around the table. The stacks are trillions in size. All of us are going to get a front-row seat to see what happens over the next 5 years.

On top of that, we're going to get to analyze and bet ourselves on who we think is going to win. We know there are some other companies pushing to get to that table, with some sharp elbows. What a time to be doing what we're doing.

I don't know if I love the analogy, because I don't think, first of all, it's a zero-sum game where there's some fixed number of chips and someone ends up with all the chips. I do think you could see, as an example, just talking about the scenarios we described, Tesla developing an extraordinary humanoid robot business that's worth $1 trillion, Google building, to Chamath's point, a media empire based on generative AI in media, and Nvidia building an entirely new chip stack that everyone's participating in. All of them, in an ecosystem-based way, could be major winners here.

Yeah, you're right. I didn't mean it in the zero-sum nature of it. I meant it more in the stakes, and there's a lot of hands to be played and there is a prize pool.

One thing I just want to point out here, speaking of regime change: What is going on at Apple? Siri was just the early idea of an AI agent. It's just totally disgraceful. It's disgusting. It doesn't work. It's embarrassing. Then, at their biggest developer conference, they're redoing the UI. It's time for regime change at Apple.

No, this has happened many, many, many times in many industries before. Companies that were stalwart organizations transition themselves from being a growth business to being a cash cow. These are well-documented transitions, and it requires an extremely brutal reset if you want to shake that up.

Yes, I think the same thing you have to respect Apple for is stability. Some of their best, longest-serving executives have been there for 20 and 30 years. On the scale of innovation, it's a horrible thing. The reason is that we all just get old. Our skill sets become rusty, and we don't have the energy or the capacity to think about what the future actually looks like because we are not living it.

Then what happens is you task those decisions to people you try to hire. But you saw it in the clip with Sam. Even in all of that crazy recruiting chaos that's happening right now for these brilliant machine-learning and AI people, maybe that's a fight between OpenAI, Meta, and Google. But what you don't hear is Apple. So, who's Apple getting? I have to think that Apple is not getting any of those people. By the time you end up at Apple, it's just a different caliber of person.

That is true. And they're living inside of a cash-cow organization that's going to optimize for “don't make mistakes,” right? But it's happened to HP. It's happened to Lotus. It's happened to Intel. It's happened to General Electric. It's happened to companies. It's just happening to Apple.

So, we should just not sweat it and move on. I don't know, Thomas, what are your thoughts?

I mean, it's kind of shocking. With all that cash, they don't acquire anything. They had Project Titan—$10 billion to build their own car—and they just shut it down. Imagine if they kept going with that. Do you think it's regime-change time? Maybe Tim Cook retires and they put somebody who's a product person in charge of it, or maybe they should merge with Tesla and put Elon in charge of it all.

There just seem to be no new products coming out of there. It's absolutely confounding that they're optimizing for share buybacks and earnings per share instead of having some amount of that money go toward innovation and acquiring companies. Their biggest acquisition is Beats. Give me a break.

It's interesting for me, because I've studied Apple basically my whole career. Their defining competitive advantage was the integration of hardware and software. That led to the beautiful MacBook that we're all using, and it led to the iPhone. The fact that they were so coupled between hardware and software, the user interface, and so on, directly led to them winning, let's call it, the mobile era.

But back to Chamath's point, I think the analogy holds in AI. They're the opposite. They don't control the silicon, and they don't control the underlying models. So now they're back to, maybe using a historical analogy, the PC makers who didn't control the OS.

I think the good news for them is that they still have a monopoly on users, and they have $3 trillion of market cap to play with. I think it's way too early to count them out.

But I think the market—let's posit what's the most extreme thing that they could do, just for intellectual sake. Buy OpenAI for $500 billion. I'm just going to put a crazy thing out there. So you think, okay, that's the most extreme. Well, is it even that extreme? And what would Apple's stock do that day?

Go up. That's my view, too. I actually think it would go up, not down, even if they did something like that. So I do think they need to be aggressive. I do think, to your point, Friedberg, it's important that all 7 of these companies could actually win and do well. That is an absolute possibility.

I would love to see them be a little bit more aggressive. You guys remember when Steve Jobs bought FingerWorks, right? It was this tiny acquisition. They made this little trackpad that you could use your fingers on. No one figured out why they did this, and then it turned into multitouch and scrolling.

Chamath, that was a great question I was about to ask. If Apple could do 1 thing—either 1 internal project or buy 1 external company—maybe we could do both around the horn. What would we advise them to do? My number 1 is build a humanoid robot. How does Apple not have a humanoid robot? That seems like the next giant consumer market: having Optimus or Figure in your house.

Friedberg, I'm going to go to you first, since I went to you last time. Is there a product they could build that they would be uniquely suited to, that would turn this all around? If you could pick 1 thing on their roadmap, what would it be?

I do think there is. I do think they're doing it, and I do think they have a shot at winning, which is this kind of ambient AI assistant. I don't know about you guys, but I must own 30 freaking Apple devices. I have many Apple computers that I use in different offices. I have phones. I have many AirPods. I have everything—watches, everything. I'm ubiquitous on the Apple platform, so I'm an easy transition into this if it works.

As everyone races to build the agentic AI assistant that's sort of in my ear all the time, or available where I don't have to stare at my freaking phone like this, it is a great unlock for humanity. It's a great unlock as a consumer. It's technically feasible, and I'm sure Apple, of everyone we've referenced today, is best suited to access the consumer, design and engineer this solution, and make it truly transformative.

I think it references a little bit of what Jony Ive and Sam Altman have been talking about doing. I do think this is exactly the direction Apple is headed, and I do think they've got a very good shot at winning at it. I don't think they need to own the full stack to be successful here.

Got it. So we've got Optimus, and we've got the device you're talking about—this ambient assistant that's part Siri and perhaps part pendant that records your behavior in the world and gives you feedback. That's what they're calling a puck, perhaps, that Jony Ive has made, or these pendants that record everything.

Thomas, what's your thought on the 1 product they could create?

To that point, it's interesting to think that the AirPods business at Apple is 3 times OpenAI's revenue base today. That's right. And that's just the AirPods business. By the way, let me just say 1 thing about this. We all think about devices in the context of a single device being an assistant.

I think if there are more devices integrated into our lives and the assistant is ethereal and ubiquitous among the devices, it's almost like Star Trek: The Next Generation. You walk in, you say, “Hey, computer,” and there's always a device available that's doing things. There's always a device observing, and there's always a device able to take care of things for you.

Whether it's in your ear, whether it's your phone, or whether it's your watch, basically these devices, instead of acting independently, all know what you've been asking or talking about with the other devices. So you could get in your car and pick up the conversation you were having while you were sitting in your office in front of your computer to do work.

And so the agent effectively is almost like this ethereal, ambient assistant. So everywhere you go, the agent is there. They could even be in a candlelit bath with you, Friedberg. They could be in there.

Think about also having identity, so it knows who you are. I could be in your home, JCal—not that I would ever get invited to your home, but let's say I was there. I could walk into the living room, and there's your puck, and it starts talking to me because it knows who I am. And, yeah, it's like it knows me. Or you and I could have a bath for two, and it would know when each of us are fighting over what music we want to play. The assistant will hear out the debate playlist.

Do you have a device, before we go on to IPOs here? Do you have a device or an angle for Apple to go after if they were truly ambitious? Or maybe they are and it's just in stealth. What do you think? Do you think it's the goggles, the glasses? Do you think it's a pendant? Do you think it's Optimus? What do you think?

I don't think they have any chance of anything great.

Love it. I would take the exact opposite of what Friedberg says. Look at this chart, and I'll tell you why. Okay, here we go.

This chart is not a strategy. This is a chart of Apple's revenue, and what you see is that iPhone has completely stalled out. And so, to Chamath's point, where do you make money? You make money in other hardware. This is not a strategy of success. This is a strategy of inefficiency: I lost my AirPods, so I need to buy a new pair. Oh, the cables changed, so I need to buy a bunch of those. This and that. A this-and-that strategy is not a strategy. It's a tactical play for revenue optimization in the short term.

A company that focuses on this kind of revenue growth is not capable of creating something that's exceptionally unexpected. That will come from a new company that has no ties to the past, has no nostalgia for the fact that we're going to swap out the connector type and book another billion dollars. What Chamath said is actually an indictment of their ability to do it. When your AirPods business is 2 or 3 times bigger than OpenAI, what there is internally when you try to have a strategy meeting about what to do is derision about OpenAI, because you're like, “That's small, and even our AirPods business is 3 times bigger.” That's what some smart-ass MBA will say in that meeting, and it'll shut the meeting down. So how do you expect that culture to then all of a sudden get their act together? I think it's exceptionally hard.

Here's the clip. Play it, Nick. It's a great point. I'm Apple nostalgic.

Me too. Bring Steve Jobs back.

Watch this lunacy.

Marques Brownlee

You probably saw that Jony Ive is linked up with OpenAI to create some sort of future AI device.

Craig Federighi

Yeah, I don't know what that is.

Marques Brownlee

I don't either. Is this a space that Apple's looking at? Is this a space that goes beyond what you have in the current lineup of devices? Something that is more personal? Maybe you wear it? Glasses.

Craig Federighi

I think we have some extremely personal wearable devices. If you want something that's aware of your environment with audio, I think you're wearing one right now on your wrist. If you want something that you can capture the environment with and see and also receive visual content, you might just have one in your pocket right now.

Are there other form factors that can make sense for AI? Sure. But pretty hard to beat something that's with you all the time and glanceable, or provides a nice screen that you can interact with. So, yeah, I don't know what they're working on.

What do you think, Chamath?

Again, I think I want to be very clear about what I'm saying. That is a very competent Craig Federighi, a very, very competent executive, and whoever the person beside him is, that guy's—I'm going to assume—competent as well. They're competent at making money the way that they've made money for the last 17 years with no meaningful disturbance.

And I think it's just something to appreciate that after 17 years of unmitigated linear success, it's very difficult to retool yourself. It's like asking Michael Jordan to go and all of a sudden become an all-star in baseball. It doesn't work. And so I think it's okay, though. This is my point: it's okay, guys, to have creative destruction of companies.

There was probably a version of us blathering on about HP and being nostalgic about the transistor radio that they made and the HP 12C calculator that they made. “Oh my God, why can't they figure their stuff out?” And where are we today? HP doesn't even exist. It's okay.

I mean, Chamath, the fact that they launched Siri, they bought that company, and Siri can't do anything other than an alarm. It can barely play a song, and it barely can do directions. I mean, literally, we're in year 27 of Siri, and it can't do anything. And then I have the Gemini and Grok voice, and when I turn that on, it does whatever I want. It will load on my Pixel. It loads other applications, fires them off, and does specific tasks in them.

It's absolutely incredible on your Pixel. I have a Pixel. When I flip open my Pixel, I have the Pixel 9, Chamath. It's the anaconda of smartphones.

Pixel 9 foldable. Got it. It's the greatest assistant ever. It's what Siri was—what Steve Jobs showed Siri could be.

I had you at 9.

He had me at anaconda.

Yeah, I had you at 9.

Chamath, I would argue to you that I think this management team has done it once, and it's in the transition of their gross profit base, which doesn't show in the chart that you just highlighted, but was something that I lived as an analyst covering the stock for a long time.

If you remember, over a decade ago, 90-plus% of their gross profit was a one-time hardware sale on the iPhone. And no one thought that they would ever be able to get away from the drug of selling that one iPhone unit, right? And cut to over a decade later, it's 40%, right? And I don't think they get enough credit for actually transitioning from hardware to a recurring gross profit base. But look, you might argue that that was an easier pivot and challenge than what they're going to face. And so, let's see whether they can do it.

The other thing, guys, I wonder about—let's, I know we want to talk about IPOs—but I do wonder whether Zuck buying Scale AI for $15 billion gives air cover for other companies to really start being aggressive. And to me, as we think about Circle and CoreWeave, 2 companies that have gone IPO recently, it's kind of amazing, numerically, that the charts are almost identical, even on a dollar basis, on a share-price basis, right?

Because to me, what it says—we were talking about the dispersion of the Mag 7 before, right? Which are going to do well, which are not. I expect we're going to have a lot of opinions on this over the next few years. And frankly, they may change. We may think about Apple one way today, and it may change in a month, right?

But I do think the market is starting to realize that there is dispersion, that AI might create some winners and then some losers, right? And it's starting to think about, “Okay, how do I want to be positioned for the next 5 years? What are big, open-ended growth opportunities?” And here come 2 companies, 1 levered to crypto and the other levered to AI. So I don't think it's a surprise to me. These things are intertwined.

You're 100% on, because here's the thing: the average profit margin of the S&P 493 is—drumroll, please—12%. The average growth of the S&P 493 is—drumroll, please—single digits. So, to your point, why would you own any of these 493 companies that may turn around and one day just get decapitated by something you don't even know is getting cooked up by a couple of kids in a garage using OpenAI or Grok or what have you?

It just makes a lot more sense when you find investable companies in the big themes of the future to, at a minimum, hedge, right? Be less long the past and, frankly, make some bets about the future. And I think that's where you're seeing these IPOs just absolutely rip. What is a better comparison, in my opinion, are the companies that are truly levered to the future themes of AI and crypto versus any of these IPOs that have happened of companies that are not. And I think what you see is there's a dispersion there as well.

They are being treated almost as similarly, Jason, as the S&P 493. It's like, “Yeah, it's good. Yeah, it's fine.” They get some reasonable gains. But if you're levered to either of those 2 trends, you're off to the races because it's just so disruptive. People don't want to be bag-holding these old legacy companies.

We're already into our next topic, which is IPOs and M&A. Lina Khan is no longer in the building, and M&A is back on the menu, as are IPOs, as Chamath has pointed out.

3 IPOs: March 28, June 5, and June 12. CoreWeave, up 4 times after going public, with an $81 billion market cap. Absolutely stunning. Circle was 25 times oversubscribed, up 6 times from its opening price, with a $48 billion market cap. Chime—that's a neobank like Nubank, which is already public—was up 40% from its IPO price, but then it went down 20%, to a $12 billion market cap.

On the other side of the ledger, we have a ton of M&A this year. So when you look at what's happening under the Trump administration, look at what's actually happening. The game on the field is 3 major IPOs and then massive amounts of billion-dollar acquisitions. Obviously, we talked about Google acquiring Wiz for $32 billion. SoftBank bought Ampere—I don't know what they do—for $6.5 billion. OpenAI bought 2 companies, 1 for $3 billion and 1 for $6.5 billion. Developer copilot Windsurf: $3 billion.

Jony Ive’s io is making some sort of puck or hardware device. Databricks bought Neon for $1 billion. Salesforce did an $8 billion acquisition, and interestingly, DoorDash bought 2 companies. Uber made 2 smaller acquisitions. There is a ton of activity here.

What does it say about the market, David Friedberg, that we’re seeing so much M&A and these amazing IPOs coming out within the last 3 or 4 months? Let me just follow up on a comment Chamath made and ask Sacks for his view. I have a theory, and I haven’t looked empirically to see if it makes sense.

For most of the S&P 500, fundamental profit growth is pretty anemic, with the exception, obviously, of a couple of the big tech outliers—the Mag 7 and a few others. For the majority of the S&P, this is a pretty anemic environment relative to the transitions that are fundamentally underway in the world with AI and ancillary technology.

Are institutional fund managers hungry for access to some of these new, high-growth offerings, and have they been held off? Just to go back, I think it was around 2008 when public institutional fund managers started to do crossover investing into private equities. That scaled up and scaled up, and it entered a stage where there was a heavy flurry—a lot of activity and a lot of crossover late-stage investing—right until 2021, when things started to pop in 2022.

Because they were overexposed with their private equity portfolios relative to their public equities, they came out of 2021 and 2022 with the market declining, and they now had a higher concentration of private equities than they were supposed to have. They have been kept out of the private market for the last 3 or so years.

Is there now this pent-up hunger or pent-up demand for new issuances—for high-growth tech issuances? Is that what we’re seeing? Is there pent-up demand because they’ve had to stay out of the private market for 3 years? If there is, obviously it bodes well for late-stage growth startups that are looking to go public, because the demand will be there. I think the reports were that the Chime IPO was 18 times oversubscribed.

I think you’re right. Something that I’ve talked about with you guys, and that was a big conversation at our All-In Summit last year, was the health of the private ecosystem. We talked about the concept that if you put a dollar in, you need to get a dollar out. I do think that we’re starting to see a healthier market where we know a lot of dollars have gone in, but now we’re starting to see some dollars coming out. I think that’s one element, and that’s both in M&A and in IPOs.

I also think the second element is that we’re in the tailwind of the mobile and SaaS era. Even if you look at the SaaS companies, we put this together in our deck when we were preparing it for our conference this week. Chamath, I think you’ll find this interesting.

If you look at SaaS in 2021, the median growth rate for SaaS companies was 17%, and a quarter of those were growing over 25%. If you look at SaaS today, the growth rate has been cut in half, from 17% to 9%, and only 5% of that cohort is now growing above 25%.

So I think, Dave, what’s clearly happening is that other sectors that were predominantly seen as growth are now slowing down. The market can no longer just rely on saying, “I’m just going to own the Bessemer SaaS Index for the next decade, and I’ll do great,” because those companies have really slowed down.

I think the market is starting to look forward and think, “Over the next 5 to 10 years, what are the companies that can compound at maybe 25% per year over that time frame?” I think companies like CoreWeave, Circle, and Chime, by the way, and others are going to fill that gap.

I really like this chart. If I had to guess about what has changed from 2021 to 2025, it’s that most companies have realized that buying yet another vertical software solution is not going to help their business. It typically adds bloat, it adds cost, and it adds people.

I think starting in 2023, what people started to guess is that, at some point in the near future, you’re going to have some AI way of rewriting all of this vertical software. I think that’s why it stopped growing. I don’t think this SaaS market ever had the return on equity that it was supposed to.

I think so many companies have woken up from this hangover saying, “There’s got to be a better way. It can’t always be yet another tool, yet another program, yet another multiyear delay, yet another price escalator.” I think the jig is totally up for software.

You’re referring to Salesforce and the SaaS category, Chamath, and what you’re doing at 8090 specifically?

Yeah. Well, it’s not just us, but if you look at anybody that’s rebuilding software, it is so much easier to rebuild software from scratch today. My team of 30 people can transact hundreds of millions of dollars of work—not because we are so prolifically amazing, although I think the team is good, but honestly because the underlying tool chain gives you a level of leverage.

If you rebuild the software development life cycle using these tools, you can’t help but become much more efficient, and you can’t help but deliver custom solutions that are meaningfully cheaper. Jason, if you look at the entirety of the software that runs the world, we’re going to rebuild it soup to nuts—all of it.

The tool you’re referring to, just for the audience, is the AI copilots that are contributing 30% to 40% of codebases at Microsoft and elsewhere, specifically?

Those are good for individual people, but the software development life cycle is more the horizontal, end-to-end process of making things.

What we do internally at 8090 is have an entire process that starts from the PRD all the way out to functioning code. We use different techniques at each step, but what you get is a 50%, 60%, or 70% increase at each step, which then compounds.

You have the ability for a team that would otherwise be able to service tens of millions of dollars to become a team that can service hundreds of millions, and then a team that would otherwise service hundreds of millions can service billions.

Let me ask you guys your response to this theory. If there is going to be this accelerated, call it, custom software rebuild of business models, and you take the S&P 493, do you think that we enter an era where there is a similar dispersion as we’re talking about seeing in the Mag 7, with the S&P 493?

Are there going to be probably the biggest money-making opportunities for investors that we’ve seen in decades between those that do adopt and rebuild using AI and those that don’t?

100%. I had a call yesterday with one of the largest private equity funds in the world, with hundreds of billions of dollars under management. We’re doing something with them at 80/20 with one of their most important assets.

When you’re an owner of a business and you can direct specific change, rip out hundreds of millions of dollars of software licenses, and replace them with tens of millions of dollars of highly customized software, it’s an enormous lift to OpEx and business-model quality.

Why doesn’t it happen more? The reason it doesn’t happen right now for the S&P 493 is that the IT organizations inside all companies essentially speak a different language than the CEO, the CFO, and the board.

If the CEO, CFO, and board of directors of the S&P 493 speak English, the IT organization speaks Mandarin Chinese, and you get away with saying all kinds of things. I’ll give you an example: I went to a CIO conference, and one person I met had an $18 billion-a-year IT budget. What the fuck does that actually even mean—to spend $18 billion a year on IT?

I’m not saying that this is a Mag 7 company, guys. When you take that example and multiply it by 50, 100, and 493 examples of people spending money, there’s an entire cartel of influence that’s been built in software that’s going to get undone, because you’re not going to be able to justify it.

Absolutely correct. The response from the SaaS industry is changing from the per-seat model as the number of employees at these companies continues to get lowered. Obviously, Microsoft has had a lot of layoffs, and Andy Jassy is talking about layoffs. They’re moving from the per-seat model. They’re not taking this lying down.

They know that people are going to make custom software, so what they’re doing is moving to a consumption model. You’re seeing people charge per call, per customer-support call, and so on.

They’re also dramatically lowering the number of people and developers they have on their teams. A lot of what’s happening in the background is that they’re starting to do rollups, and people are starting to talk about how we can take 20 of these SaaS companies and roll them up, just like you’re doing with your 8090 playbook.

I’ll tell you why consumption-based pricing doesn’t work. You can have some adoption in the short term—the best example is Snowflake—but in the long term, it destroys your business.

The reason is that you don’t know which data is valuable, and you’re not going to put up with a variable business model that increases your costs more and more because you need to trap everything. What happens is that all of these other companies develop around you.

People go back to Postgres, people go to Supabase, and they find all of these ways of saying, “Snowflake makes no sense.”

And the reason is because, in this world, nobody's going to pay for consumption. They're going to say, “How do you expect me to hold, store, and pay for terabytes and terabytes of data potentially every day?” It's not sustainable. We'll see if Intercom, Salesforce, HubSpot, and Slack start losing their customer base, or if they lower their pricing to make it easy enough to keep those systems in. Chamath, your thoughts?

Yeah, two quick thoughts. Number 1, to put a mathematical frame on this: we know that Anthropic is kind of the level zero of code generation. They're doing incredibly well powering companies like Cursor.

I think—and this is order-of-magnitude correct—that Anthropic in Q1 added 70% of the net new ARR in the SaaS industry, defined by public SaaS companies. So, let's think about that: the company in AI that is most powering the disruption of SaaS added 3/4 of the net new ARR of the entire industry. That's point number 1.

I think, Friedberg, point number 2 is that what we're seeing in the Mag 7, where we're starting to have debates about who's well positioned and who isn't, and who's going to win and who isn't, is actually, as it was over the past 5 years, going to be a broader lens into the S&P 493.

Inside boardrooms and investment committees, you're going to see the exact same conversations we've been having about the Mag 7: who's well positioned, who can win, which management teams—maybe like Zuck—are being aggressive and bold and capturing the opportunity, and which ones are not. For me as a stock picker, over the next 5 years, I couldn't think of a more interesting time. We're actually going to see dispersion between winners and losers.

Do you think these rollup models make sense? You've probably heard that some fund managers are putting together pools of capital to go out and buy businesses that they can then apply their know-how to. They're bringing in smart people in AI to create a category killer and go after that market. Are you guys participating in that, and how do you view that opportunity? Are all the public companies basically too mature, or are some of them going to go after this type of model as well?

It goes back to whether you can attract the talent to go and do these things. My advice to this large private equity firm is that you can probably try to stand up your own AI organization, but I suspect you're going to get the person who didn't get an OpenAI offer, didn't get a Meta offer, didn't get a Google offer, didn't get an 8090 offer, and then didn't get an Apple offer. That's the person you'll hire. How good that person will be, who the hell knows?

I think the problem is that even if you take some of these mediocre industries and roll them all up, you ultimately have to find a buyer who wants to own that business after you. If you were to buy a bunch of accounting firms, law firms, or IT services firms and do an incredible job, who wants to buy that in 7 years?

If you went to the OpenAI Demo Day, there was this really interesting chart where Andrej Karpathy talked about integrating Google login into one of his apps. I think it was his MenuGen app. The comment he made, which profoundly hit me, was, “Why am I doing any of this? Why isn't this just 1 click behind the scenes?”

You could take that generalization and apply it to all of IT services. Why does any of that exist? Why isn't it all 1 click? Eventually, if these agents become smart enough, the fear that I have is that there is no terminal buyer for many of these companies.

But they could still be public, Chamath. They could trade at some multiple of cash flow, and you're basically arbitraging the cash flow.

But I'm not talking about the private equity trade. I'm actually talking about the public equity trade. If you look at the 493 companies, those are better positioned.

Here's what I would do: I would take the 493, and the filter I would apply is: what offline assets do they have? What online assets do they have? What percentage of those assets are defensible and unique and exist in a post-AI world? And what percentage of those assets disappear in a post-AI world?

I think where I would end up is, I'd own a specialty chemicals company or something. You're still going to need lubricants and stuff, and you can find some way to make it. But if you're a—

You need lubricants. Sorry, go ahead.

You know, I love the lubricants, but no Diddy. No Diddy. Baby-oil making, 5 by the crate.

Chamath, do you want to talk about your SPAC tweet?

Uh-oh. You know the market's back. Can we see this? Can you play the siren? Can you play the siren? As with all my tweets, it starts when—

Look, here's what X is: an incredible platform. I use it for a lot of things, but your villain phase right now, man—you're a full supervillain. It's so great. The retweet is more important.

Yeah, I love that quote retweet. Here we go. Here's the tweet.

Here's the tweet: “Incredible that almost 58,000 people voted in his tweet on whether he should launch a new SPAC.” So, give the people what they want, Chamath, or what?

I first started this because when I use X sometimes to sound off, it de-stresses me during the day. I'll troll people or whatever. Then I did this, and I was so impressed that 58,000 people voted.

But really what happened was that I had a lot of very smart-money people on Wall Street and some crypto folks call me—people that I respect—and basically what they said is, “It would be really good if you did it.” So, I don't know if I'm going to do it, but I'm heavily leaning toward doing it.

Well, the argument to do it is that you learned a lot since last time. There's a lot of inventory there, and you've got a lot of access to pre-market companies. I think what people need to understand is, when you're doing SPACs—and correct me if I'm wrong here—

Here's what I'll say, Jason: this poll and this Community Note will be in every single document I do. Nobody listening to this should participate in this. This is going to be for me and a handful of advanced, large pools of money.

You should stay as far away as possible. Whatever I do next, don't participate in SPACs. That's the rule here. Stay on the sidelines. Do something else. Don't come into the arena, because we're trying things.

Chamath, don't you have enough going on? Why would you do this when you have—

Fate loves irony. Fate loves irony, bro. Fate loves—

Absolutely. This will be hilarious. It would be the greatest IPO of all time. If the poll was “yes,” I'd be like, “Oh, this is the last thing I need.” All-In SPAC, let's go.

Thomas, are you going to buy the All-In SPAC? What's going on? The SPAC market's coming back.

I'm open to all great companies coming to the public market.

Love it. Love it. So, Thomas, can I ask you a question? Tell us about the state of liquidity, and actually about IPOs and SPACs in general. Where's your temperature on it? Give us a read on what you think.

Look, I think we're getting real-world data in real time—not just from higher-visibility companies like Circle and CoreWeave, but Chime also did really well. Caris Life Sciences, a company more in Dave's wheelhouse, is also just coming out.

Then wait until we see the flurry of S-1s that have already been filed. Figma is a potentially generational company that's going to be coming. So, I think we're going to see fantastic assets coming out, and I think the market is saying we're open for business.

The Mag 7 is controversial. To Dave's point, in the S&P 493, there are going to be lots of winners and losers. It's maybe not as obvious, and there's going to be some dispersion. So, bring on the new cohort.

I think it's the first time you could probably argue that you could go short the S&P and pick a couple of winners. It might be the first time in the last 20 years that I would feel comfortable doing that, because I'm pretty negative on people's ability to pick stocks. But I do think this is such a transformative moment that if you really have a sense for what's possible, you could start to see category killers emerge out of the S&P.

Totally. But do you care about how these companies go public? Do you care about SPAC versus direct listing versus IPO?

I don't. I only care about the quality of the underlying asset and what I think it can be worth 5 years from now.

Now, obviously, I do care about the liquidity that I'm getting in the IPO. Am I getting 1 million, 100 million, or 1 billion as the float? That's number 1. Obviously, I also care about the percentage that's floating, and I care about the lockup. Those 3 elements are really important in terms of a company going public and how we think about participating.

Give the listeners the guidance there. For the first thing, bigger is better than smaller, correct?

Correct. Number 1 is: can I even buy it? If the IPO is so small and we can't get a large enough position, it doesn't really make sense for us. That would be point number 1.

Point number 2 is: how much of the company is publicly floating? Better there as well.

Correct. You kind of get a truer price when a higher percentage of the company floats. It's also most likely going to be less volatile and less susceptible, Chamath, to pricing predation, manipulation, and things like that.

What's the percentage float? I think 20% is, in my opinion, kind of a minimum. Some have gone out—I think I remember, correct me if I'm wrong, you may know this—I think LinkedIn went out at around 10% or something. I remember it being really small, and a lot of us thinking, “Wow, that is a small float,” which ended up, by the way, being very volatile.

So, number 2, the float. Then number 3, the lockup. First, is there one? In a direct listing, there may not be one, so you may get to a truer price faster.

Friedberg, why do you think there have been no direct listings? Why has that totally fallen away? Spotify did one, we did one at Slack, and then where are they? Why don't people pursue those?

Here's a statistic. I actually had to double-check this because I couldn't believe it. If you look at the cohort of companies that went IPO in 2021—and I'm not including SPACs in this particular analysis—at T+1 year, the cohort was down about 40% on average.

Okay, fine. Maybe they went up too high. 2021 was a peak. They didn't do well in 1 year. At T+5 years, it's down 50%, which really shocked me. So I think there's scar tissue on both sides of the table: on the buy side, about, “Wait, hold on. What am I really buying, and how do I make sure that it's a sustainable company?” But frankly, probably also from boards who are taking their best assets public and may just want to pursue a more conventional approach in the beginning stages.

I can tell you, for us, direct listing versus IPO makes no functional difference. Each has a benefit, depending on how concentrated your ownership base is, how understandable your business model is, and things like that. But we just want these companies to come.

There's a market behavior in direct listings, by the way. I've mentioned this once, but I've been in 2 transactions with direct listings. The first was Slack, and in the execution of it, we misexecuted—we meaning me, because I had a huge ownership of Slack. I didn't know what to do with it, and I ended up distributing portions along the way. It then went through all kinds of turbulence, and then it got acquired slightly above the IPO price.

What I learned in retrospect was the best trade is actually the first-day trade on a direct listing. When it came back around and I got a distribution the day before of Coinbase, I mentioned this to Brian. This was not a judgment on the company. I said, “If this direct-listing process is going to map to what I've experienced at Slack, the right thing to do is to sell.” I sold that on day 1 at $335 a share.

Yeah, it's just—I think, Jason, it's still not at the IPO price. I think it might be getting close.

No, it's not back.

Yeah, so these direct listings are not what they're expected to be, either. If we look back on SPACs, I think SoFi is above the price, and that might have been one of yours. Joby is getting close.

These were venture investments. These were late-stage venture investments in your mind, Friedberg, and then retail tried to become venture capitalists, but they didn't have the 5- to 10-year horizon that we as venture capitalists have. Is that your assessment of it? Are there any great ones that came out of the SPAC movement?

I mean, the direct-listing era, as an example, let's talk about Spotify, which has basically 7Xed over that period. Again, it's hard to tell causation versus correlation. Ultimately, for me as a long-term owner of these businesses, I really just care about the quality of the business. Whether you chose to go SPAC, direct listing, or IPO is a mechanical decision. To me, the output is quality of business, and that's ultimately what wins out.

Okay, I want to end on this. You just shared a chart of AppLovin and the massive revenue per employee. This is astounding, Thomas. AppLovin, as we can see here, had $3.6 million in revenue per employee in 2021, now up to $7.6 million. They peaked at 1,000 employees and are now down to around 750.

In related news, obviously Microsoft—we talked about it the other week—let go of 3%. They're planning massive cuts again in sales. These are organizations that are at record cash and record revenue, in an industry where we had a tradition of not firing the graybeards and people who had been at the company for more than 10 years.

Andy Jassy didn't come up as one of the companies we think is going to win at AI, but it might be the company most impacted by deploying AI inside its enterprise. He launched a missive. I suggest everybody read it. When you send a missive like this to your employees, you're trying to communicate something to them and to the public markets.

He published it on his website. He talks about dozens of AI projects: AI tools for advertisers, GenAI for sellers, their product detail page, Alexa coming back with a brand-new version, shopping assistance, everything. But then he started talking about the workforce size. He says in this manifesto, “In the next few years, we expect this will reduce our total corporate workforce as we get efficiency gains from using AI extensively across the company.”

So, my question to you, Thomas, is: When you hear public CEOs talking about lowering the number of employees while they're growing 10% to 20% per year, this is obviously awesome for earnings and the share price, but there's going to be massive job displacement. Any thoughts on job displacement, job replacement, and society navigating that? Also, what do you think of Amazon as a business, specifically, and of them being a player in AI, with AI being a player in their business?

I think it's an important question, and I'll defer to what Jensen answered on this topic because, in my view, it's still the most credible and cohesive answer I've heard. Jensen, the CEO of NVIDIA, is an incredibly long-term thinker. He looks at a population that's getting older and wonders who all the young people are going to be who take care of all the old people, whether they're nurses, doctors, or something else.

In his view, we better get a lot more productive to deal with our inverted demographic pyramid. I ultimately think this is going to enable more young people to take care of more old people. It's going to create incredible opportunities for the economy. Knowledge workers are incredibly flexible; they can take their tools from one particular skill set to another. I think this is going to make us more productive and wealthier, so I'm definitely on the more optimistic side of the scenario.

Chamath, any thoughts on Amazon? They didn't come up, but obviously AWS is crushing it, and they're a major player. They have their own silicon they're making. You mentioned that being an important part of the stack. Then you have Optimus and Figure robots that are going to be in their factories. That's a lot of jobs.

Delivery robots. They're doing drones like Zipline. They have their own version of it, obviously, and they're doing Zoox. If you just look at their behavior and their investments, they're massively investing in robotics, self-driving, and chips. So they're pretty hardware-focused.

For physical AI, they're a kingmaker in part because they're a sink for demand. They'll just generate so much demand for robots. If Figure lands the BMW or UPS robot successfully, Amazon will buy a gajillion of them. If Optimus lands a successful robot that they tune inside the Tesla factory and then are ready to sell, Amazon will buy a gajillion of them. If there are drones that are delivering things, Amazon will buy a gajillion of them.

On the one side, there's a lot of typical OPEX lift that Amazon will get. I think the problem is more with AWS, which is that their success is actually their biggest bottleneck. They're not necessarily kingmaking. They're about being a purveyor of many, many, many different things that you can find inside the AWS Marketplace.

The thing that they'll have to embrace is: Do I differentiate my own hardware from NVIDIA's at some point? Do I actually make a real bet on models and try to, frankly, buy Anthropic, which is probably their only solution, and tightly couple it in and say that if you want to have next-generation code-generation experiences, they need to run inside of AWS?

These are the difficult decisions that I think Andy will have to face, and he's going to have to spend hundreds of billions of dollars. But the Amazon retail side is going to be a kingmaker for all of these physical AI things.

Friedberg, any thoughts on Amazon just as a company broadly? Chamath is saying, “Hey, they're a kingmaker.” That seems like a really interesting insight. Do you have any insights there on Amazon and their playing a part here in the future of AI?

I don't.

Chamath, any closing thoughts here on the old guard—Microsoft, Amazon—their employee count, and the cuts we're seeing there, and what these companies will look like in the future in terms of revenue per employee? They're not hiring young people. They're getting rid of the old folks. They're just advancing, it seems, at a—They're adopting AI pretty severely at these companies.

What are your thoughts there? I'm going to play the role of JCal and ask a question to all 3 of you guys. Oh, here we go. Microsoft's employee count peaked at about 250,000, call it about a year ago. Who here believes that in 5 years Microsoft will have more employees than it does today?

More.

I'm going to say the same. I think they'll have just about 250,000, plus or minus 10%. If I could pick push as the answer, I would pick push: they're going to get 10% better every year with AI, 20% more efficient. Therefore, they don't need to add people. But I also don't think they atrophy much more. So maybe they have 225,000 to 250,000.

Why'd you say more so quickly? I'm curious.

This chart, which I think is a very dangerous vanity metric, is why. What Microsoft touts is what percentage of code is generated by AI, without answering the more important question: Is that code useful and good? If you ask that second layer—and I sent you this tweet from Yann LeCun—I'll tell you that this is my lived experience as well: Most code generated by AI is crap, and most of the tools that we use—the reason we call these tools “app crappers”—are called that because most of the code they generate is crap.

It's great in a single-player mode, but transitioning from single-player mode to a complex enterprise environment is not possible today. I think Microsoft puts these metrics out because they want to seem like they're on the front line of it, but I suspect that this is just like how you used to hire McKinsey consultants to fire people because it was good air cover. It's probably just air cover to fire a bunch of folks that they probably wanted to get rid of anyway, but it's not related to that chart.

The reason is that Yann LeCun's tweet is true. When you allow these models to run over complicated tasks over long periods of time, the error rates compound to such a degree that the resulting output is not worthwhile. Until that problem is fixed—which I'm sure it will be, and I'm going to bet that it will be—the idea that people are getting laid off because of coding agents, I think, is a fallacy.

I suspect that Microsoft's business, on the margin, grows. Back to Dave's point, some of the S&P 493 shrink and go away. It'll be cheaper for Microsoft to bundle together a bunch of other products that are point features today. So they'll have more people. They'll indeed have more people. The people will be different, and they'll have different skill sets, but I suspect Microsoft's employee base grows.

Friedberg, what say you?

I think it shrinks.

Wow. So, by the way, it's pretty interesting to think about. We have one decisively more, one median—about the same, a push—and a less.

I only say that because I do think there's a real probability of revenue decline in the next 5 years. If you look at the enterprise install base, I think that cloud gets competed away. I do think that, on the application-software layer, they're going to have a really hard time in this new world because the old-school customers that buy Microsoft are going to die. They're more likely to die in their marketplace compared to the folks that are going to build native software and native workflows.

I'm not really where Chamath is. I think you may be right about where AI-written code is today. I don't think that's true 3 years from now, 4 years from now, given the pace of improvement. In a world where software-written workflows are built for you through agentic tools, I think Microsoft's core business is going to decline. The losers are their biggest customers, and the winners are not going to use them. So that would be where you, Chamath, maybe you're the tiebreaker.

I'm in Chamath's camp. I actually think the Microsoft business will be bigger, if anything, on its own, and that at the end of the day we'll just need more people to support it. I just think they'll be more relevant. They'll have more productive employees, but they'll still be more of them.

I'm predicting incredible growth and the same number of employees.

So you guys are predicting incredible growth and employee growth.

I think that's interesting. So, sorry—less revenue, less employees. Interesting. The thesis, as AI grows, is basically where the application dollars go. That's one way to think about this, right?

There are multiple clouds, by the way. I went to the Google Next event last year, and I ended up going to these special dinners or whatever—a couple of cocktail-dinner things—because I spoke there. They put me with a bunch of these people, CIOs of Fortune 50 companies, and all of them said that they're multicloud. No one's going to standardize on one cloud, so everyone has to be on Microsoft and Google.

I had never really recognized or thought about this as being a fact: It's not necessarily the best or the lowest price. At the end of the day, these guys are going to distribute their exposure. I think that maybe supports your case. I'm very easily convinced by other arguments today. I'm very convinced.

Here's the revenue. What a spectacular revenue run. I think all 4 of us would agree that if we could synthetically own AWS, Azure, and GCP—if I could somehow automatically create an index of all 3 of those businesses over the next 5 years—

Yeah. Yeah, you wouldn't need to own anything else.

You wouldn't need to own anything else. I wish Elon would take that.

Why don't you just not have to put up with the shitty part of the rest of their businesses and own all 3? That's it. Call it a day, because you've got to assume that if 1 of them wins over the other 2 or accelerates ahead of the other 2, it's going to more than make up for the losses that the other 2 might experience in their other businesses.

The multiples aren't crazy on those 3 companies, by the way.

Correct. Quite reasonable.

I think if Elon took what he did with Colossus and had an AWS competitor, he would be a serious competitor in the space. The velocity at which he can build out data centers is extraordinary. This is where Elon does better, because he can actually get better fundraising in the private market with xAI than what he has to deal with.

Yeah, he's really struggling with that. That's what I'm saying.

Yeah, no, no, I'm saying it's better for him, right?

Hey, guys, look who's here. Couldn't stay away. 11 o'clock happens on a Thursday, and you start jonesing for your besties. Welcome to the czar, David S.

Good to be back, JCal. Where are you? You in LA?

Mm-hmm. I'm in LA.

You're at someone's guest house?

Yeah, actually, this is one of your guest houses. You just lost track. I still have the key code.

It's a JCal Kalanick. JCal is at your guest house. JCal. JCal, here. I'm here. Come down the hill. He'll still get that reference. It's getting kind of dated now. Oh, God. Kalanick is ride or die. He would jump on a venti or a grande for you, for sure.

Let's talk a little bit, since I've got you, Sacks. Would you be willing to talk a little bit about the GENIUS Act? We just passed it in the Senate. I think you have your fingerprints on this. Is that true?

Yeah.

Tell us everything.

It's definitely something we supported, and I think this is a huge milestone. Basically, what happened is we had this GENIUS Act, which is the stablecoin bill, pass the Senate with 68 votes. It got 18 Democrats to come on board. We had to hit that key threshold of 60 votes in the Senate. That's the threshold you need in the Senate, unless there's a narrow exception for reconciliation.

It's very, very hard to pass any bill out of the Senate, and you need a significant amount of bipartisan support. We got that. Now, when you consider where we were a year ago, you realize what huge progress this is for the crypto industry.

A year ago, you had crypto companies being prosecuted. You had this whole regulation-through-prosecution approach, where Gary Gensler, who was the chair of the SEC then, wouldn't tell startups what the rules were. He would just announce prosecutions. This was driving all the crypto innovation offshore, and I think we were basically poised to lose the crypto industry in the United States.

What happened then is President Trump adopted this cause. He announced that he wanted to make the United States the crypto capital of the planet. He really campaigned on this, and in the very first week of his administration, he signed a new executive order making it clear that his administration supported crypto.

We've been rooting out all the Biden-era war-on-crypto rules and regulations at the agency level, and now we have this first major legislative win. I would expect the House to act in the next few weeks, and then the president will have a bill he can sign.

This is great work, and it's really important, because to your point, Gary Gensler's concept was, “Hey, there's an existing playbook. There are existing rules. Just follow those.” But none of these things actually match the existing rules perfectly, so you need some new rules. They need to evolve.

It was much worse than that, because he would say things like, “Just come into the SEC and talk to us.” In other words, you had to come in and talk to them and get their approval. But then, when startups would go in there and talk to them, there'd be enforcement people there writing down everything they said. The next day, they'd get a Wells notice, and they would get investigated, honey.

They were honeypotted, basically.

Yeah. And so the response from the industry was, “Okay, we’re just going to leave the United States.” That was what was in the process of happening until President Trump won the election and then changed the tone in Washington.

I think there was one other really significant thing that happened because, obviously, President Trump has gotten Republicans on board with this cause, but the question is: Why are Democrats on board with it? During the Biden administration, Elizabeth Warren really called the shots on crypto, and it was well reported that Gary Gensler was sort of her ally and her pick. I’ve kind of joked that Warren controlled the Biden autopen on crypto because she really did exert that kind of influence.

So the question is, what changed? I think one of the big things is that in this last election, Sherrod Brown, who was the chair of the Banking Committee for the Democrats in the Senate, lost his seat in a close election against Bernie Moreno. I think there were many reasons for him to lose that seat. He was far to the left of voters in Ohio. Nonetheless, he had been a successful politician there for a long time.

One of the reasons why he lost is because the crypto industry really got behind Bernie Moreno, because Sherrod Brown was just a total blocker to any crypto legislation in the mold of Elizabeth Warren. I think a lot of smart Democrats looked at that and said, “Why are we dying on this hill again?”

And I think it’s also extraordinarily popular, Sacks, with consumers and businesses. So there is a demand here. Clearly, we’ve got something like 50 million wallet holders in the U.S., and they’re voters. So that’s 1 out of 5 American adults, right? I think a lot of Democrats said, “Well, wait a second. Why are we just blindly following Elizabeth Warren on this? What exactly is so harmful about this?” Particularly when what we’re talking about here is creating a regulatory regime. It shouldn’t be hard to sell Democrats on new regulations.

In this case, the reason why there’s broad bipartisan support is because the crypto industry itself is calling for those regulations, because having regulatory certainty is better for them than the possibility of the return of a Gary Gensler-like figure who just prosecutes them without telling them what the rules are. So this is why I think you’re getting some significant bipartisan support. As you said, bringing this onshore is such a large portion of it.

There are tons of actors who some people might describe as bad, gray, or dark. Tether comes to mind, with a lot of regulation against it. Now those folks who are running away with the industry, like Tether, have to compete with people like Jeremy at Circle, which are totally buttoned up here in the United States, and it levels the playing field. So it’s an example of actually good regulation bringing this opportunity back onshore and taking it out of the gray area.

So it is true that the number-one stablecoin issuer on the planet right now is an offshore company. That is partly because there has not been a regulatory framework in the U.S. and there’s been hostility toward the crypto space, and the logical reaction to that is to either not get involved in the crypto space, which is what the banks have done until now, or go offshore. Neither one is good.

You can see in the wake of this GENIUS Act, the stablecoin bill, that the banks have now talked about getting into stablecoins. They’re going to issue one. Also, Tether, under this act, will have 3 years to come onshore. But the bottom line is they will have to operate in the United States.

That’s a good thing for consumers. It’s a good thing for them. They have 3 years to get compliance, but they have to move onshore. Now all stablecoin issuers under this bill will have to be audited quarterly, and by a real audit—not this attestation nonsense, but real audits by American auditors. It will verify that every stablecoin that’s been issued is backed or fully reserved on a 1-to-1 basis with real dollars in American bank accounts, U.S. T-bills, or money-market accounts.

What it does is—by the way, I’m not saying there’s anything wrong with Tether—but this does provide additional certainty and confidence, because you know that all the companies are onshore, they’ve been fully audited, and we know that they’re fully reserved. When you want to redeem and cash out your stablecoin tokens, there’s a real dollar waiting there to cash out. You prevent the undercollateralization issue.

And by the way, I’m not saying that there is, but what I’m saying is now we create total certainty and confidence, which is good for the market.

What happens if a stablecoin issuer does not comply? Can you issue U.S.-dollar stablecoins and not be governed under this system, or no? You’re saying because the U.S. dollar is a U.S. government instrument, then no matter where you are or where you issue from—

Yeah, all issuers will be governed by this. If you’re a legacy offshore issuer, you’re given this time period to bring yourself into conformity. But otherwise, what happens if they don’t? Well, it’s a good question. I guess the exchanges won’t be able to carry their tokens, and they won’t be able to set foot in the U.S. They’ll be in violation of U.S. law. It’s just not a good place to be.

Yeah. You don’t have to guess. There have been dozens of actions and accusations—legitimate ones—against Tether. New York’s attorney general did a major settlement with them in 2021. They’ve been banned from many jurisdictions. In Senate hearings, Tether should just go public in America and be done with it.

I should hear Tether founder’s Italian, Sacks. I’ve got to give you a lot of credit. We knew that you would bring an efficiency level and some expertise to this administration, but I’ve got to give you your flowers. We’re 5 months into this administration. We can disagree about many things. One thing we can’t disagree about is that this piece of legislation is here, and we’re only 5 months in.

Maybe you could speak to the velocity at which things are getting done, and then any other closing thoughts. I know you’ve got to get back to your day job.

A lot of people deserve credit for this. I just want to give out a couple of shout-outs. Senator Bill Hagerty from Tennessee was the principal author of the legislation. He did an amazing job getting Democratic votes and also bringing the Senate bill into greater alignment with the House bill. Hopefully, this can pass the House very quickly.

Chairman Tim Scott, who’s the chairman of the Banking Committee, was also incredible. The majority leader, John Thune, and then we had a few co-sponsors of the legislation: Cynthia Lummis from Wyoming, and then 2 Democrats who were really important, Kirsten Gillibrand from New York and Angela Alsobrooks from Maryland. All of them did a great job.

We’ve got great leaders on the House side as well: French Hill, who’s the chairman of the House Financial Services Committee; Tom Emmer, who’s the whip; and Mike Johnson, who’s the speaker. Kudos to all of them, because I think it really is a pretty incredible achievement that they’ve been able to get this through.

Again, just a huge sea change from where we were a year ago, when crypto was basically under attack. It was being driven offshore, and now we have it as one of the first major pieces of legislation by this new Congress. Again, that’s all because of President Trump’s leadership and prioritization of this issue. So thank you to all of them for making this happen.

Congratulations to you, David.

Hey, one tactical question I forgot to ask you: the float on these. This is how Tether is making billions of dollars a year, and this is how people anticipate they’re going to make billions of dollars a year. Are they able to split that with consumers yet? I remember reading in early legislation that you weren’t allowed to pass on the interest made from a stablecoin to consumers, I guess, so it couldn’t be an interest-bearing account.

If you buy stablecoins, you can’t get interest on them. But for the issuer, like Circle, that’s their main business model. Did that make it into the final? Maybe you can give us some background on that.

No, it did not. The way the framework works is that the stablecoin issuers cannot pass on interest to the token holders.

Why is that? I don’t know if there’s a great principled reason. This was a compromise that was necessary to get the support of the banking industry, quite frankly.

Ah, they see it as competition, I’m betting.

There was a lot of concern from community banks that if stablecoins were paying 5% interest, it would put them out of business. Personally, I think that concern, although understandable from them, would not have led to that outcome. But these are the types of compromises, quite frankly, that you need in order to pass legislation. I hope that at some point in the future, we’ll revisit that and allow stablecoin issuers to just do what they want to do.

All right, and that’ll be easier once the banks get into the act and they’re participating in this industry. Got it. But right now, they’re total outsiders, and you can understand the fear factor.

Sacks, I want to drop you off, man. I wish we could have you on for the full show, but you’re busy. You’ve got a lot of things to do.

Love you, dude. I shed a little tear, and I miss my bestie. See you soon. Thanks, guys.

How’s the All-In Summit going, Friedberg? You know, we might get Jack Ma to come from Alibaba. Who’s in touch with him?

I am. Thanks to Philippe.

I just want to do one quick shout-out to our friend and fellow bestie, Vinny Lingham.

Oh, yes, his movie’s coming out.

A friend of ours did a documentary on it. It’s great, Friedberg. You’re going to love this.

I denounce it. I denounce it. I love Vinny. I denounce it.

So great. Amazing. Anyway, it’s called Animal.

Oh, it’s a great doc.

Perfect. Can’t wait. Where can people watch it?

I think he’s got a couple of deals.

Come to your local slaughterhouse, put it on your phone, and watch it at the slaughterhouse while you’re there.

Here’s the idea: you’re going to consume a certain number of calories per month. We humans were designed to eat meat. That’s the number-one thing we should be doing as a species: eating meat.

All right, guys, I’ve got to go eat. I have a photo shoot in 2 hours.

Oh, you’ve got a photo shoot. Is it going to be you showing the legs or just the top this time? What are you shooting?

I’m going to blur out the anaconda.

You should pixelate the anaconda. I hope it’s Italian Vogue. What are you shooting?

Thomas is in the general neighborhood. I can’t comment, but just tell us—bleep it out.

Nice. Tell us—bleep it out. Chamath, give me a call. I’ve got to talk to you about this weekend.

Okay. Love you guys. Talk to you guys. Are you guys still doing the tequila launch?

Yeah, Saturday night. We’ll see you Saturday night. Absolutely. See you there.