The Exchanges

Q Right. Can you compare that to the numbers before for folks to understand?

A The Falcon nine. Mass to useful orbit is 17 metric tons, and the variable cost that, you know, I, I have seen Elon tweet about is somewhere around fifteen million dollars. So you are lifting more than five times the mass to orbit, and based on other statements, that hundred metric tons is a very conservative estimate, and you are doing it at call it 10 to 15% of the cost. So this is a, you know, we can all do the math, but we can envelope it, um, you know, roughly a 50 X. Huge. Change. And this, you know, massively changes unit economics for Starlink for sending anything into orbit. And as Antonio said, it's, it's great for SpaceX. It's great for America. And it's great for, uh, for everyone.

AI assessment note: “The Falcon nine. Mass to useful orbit is 17 metric tons”

Q bump. The election is over. The cabinet is being assembled, and we're seeing Doge, the Department of Government Efficiency, as well as maybe regulations being pulled down, and we'll, we'll get to our first topic about our new SEC chair, but just generally speaking, Gavin, as a market participant for a living, What's your take on what we've seen over the last three weeks since the election results came in?

A So if they execute on stated plans and there are some of the world's greatest execution machines involved, um, you know, Elon generally does what he says he is going to do. Um, like I, this is going to be awesome for America, for markets, for the world. And the analogy I keep coming back to is Satya Nadella taking over as CEO of Microsoft. Microsoft was a monopoly, incredibly advantaged. It had just been horribly mismanaged for years. All he had to do to start winning was stop doing really, really dumb things. And that's an incredible place to be, you know, America, like we're the greatest country, you know, we've got, you know, oceans on two sides, peaceful neighbors, incredible natural resources, you know, completely, you know, can produce our, our own food and energy. Like in many ways, most privileged country on earth, but sometimes with great privilege comes great, like stupidity and California to me would be a leading example of that. Most, in many ways, most privileged state in America and has bridged it away with bad policies. And I do think one thing that everyone of all political stripes agrees on is there are too many regulations that result in far too many administrators, far too much complexity, and an inability to build things in America. So, you know, it was used very effectively and as it should have been against the Harris administration, that they'd approved f…

AI assessment note: “this is going to be awesome for America, for markets, for the world.”

Q Mm-hmm. Gavin, any thoughts on specifically Nvidia being able to sell more chips into China material for the company? Good for America. I mean, it's obviously a pretty debatable issue.

A I think I disagree with Chamath on this. I think selling in deprecated Nvidia GPUs to China. Lowers the odds of them developing their own alternative ecosystem, which would be a lot power hungrier, um, because you use optical, you bring in optical a lot earlier for scale up fabrics. I think there's sound arguments that this is stabilizing for the world and is the best, highest probability path for keeping America ahead in AI and kind of keeping control of AI. And that's almost a shame we've had to have this debate because now people like me have said this many times in China. If they didn't understand it, they probably do really understand it. And by the way, that's not to say we shouldn't have had the debate, but, um, That is what I believe, you know, reasonable minds can disagree.

AI assessment note: “selling in deprecated Nvidia GPUs to China. Lowers the odds of them developing”

Q claims that this is a distraction away from the highest priority work that should be done? I mean, is the political stuff unique? And then his point of view on the political stuff, just like it was with Twitter, is that it's existential to society and civilization and our ability to function and to do, you know, to achieve the progress that he aims to achieve in his core function.

A So obviously he is succeeding being the CEO of many different companies. But, you know, I think a big part of what makes him kind of exceptional as a CEO is the perfectionism, the drive and, you know, politics is not a game of perfection. It's a game of compromise. So I am sure if he sets his mind to the America party that he will succeed, but I think there are other things for him to focus on and, you know, the boring company, Neuralink, these are games of perfection. These are, you know, this is about engineering. This is about solving a problem. It's not, you know, compromise in the art of the possible. So I think he would succeed. I just think that there are other things that Maybe a higher calling for him.

AI assessment note: “I think there are other things for him to focus on”

Q Gavin, what do you think should be the biggest business deal in 2025?

A I think the biggest business deal is that there are going to be deals. I think there's gonna, you're gonna see a tidal wave of M&A after four years of not being able to get anything done. I think there's an enormous amount of pent up demand. So kind of point number one, point number two, something will happen with Intel and that will be big and hopefully it's good for America. And then I do think you will see a lot of these frontier AI labs that are independent be acquired. You know, I thought your points were well taken, Jason, about, you know, everything that, uh, you know, Google is bringing to bear and X AI owns their own compute. I mean, you guys can choose whichever one of these three you like, but the ultimate AI winner will be the one with the lowest cost of infrastructure costs, the lowest cost of compute. And definitionally, you can't be the low cost provider if you're renting your compute from someone else.

AI assessment note: “something will happen with Intel and that will be big”

Q Throw a bunch of compute at the problem and these things learn really quick. I think is the high order bit there. Gavin, what are you, what's your take on Anthropic's recent success and their massive hiring binge?

A The success is extraordinary. It's undeniable. I think the fact that they are now, they were EBIT positive per the Wall Street Journal in the most recent quarter is a really important fact for kind of the whole AI narrative, because now there's You know, you could talk about circular funding, you could talk about ROI, and we could go look at the ROIC of the hyperscalers, but if OpenAI and Anthropic are at, call it a hundred billion dollars of ARR now with 80 percent-ish gross margins on inference, like the returns are there. And then if we add in, and they're growing really fast, if we add in Gemini, we add in Cursor, we add in XAI, we add in open source. You know, it's, it's not hard to see 203 104 hundred billion dollars of ARR at the end of this year at a high margin.

AI assessment note: “The success is extraordinary. It's undeniable. I think the fact that they are now”

Q I love it. It's a great, great contrarian take. Small chance, but big impact. Gavin, what do you got?

A Well, first we should use Gemini deep research to, um, just ask it to look at the total, you know, Pax Americana debt outstanding, apply the current market interest rate to it, and then look at that You know, interest expense relative to GDP and generate that chart over time. That would be a cool chart. That would be interesting. I don't, I don't, I don't disagree with Chamath. You know, any, anything is possible. And, um, you know, there, there could for sure be, be a problem at a big bank. I don't know that I would say it's likely, but anything is possible. My most contrary belief, so I think that America over the next, at some point over the next four years, will print at least one year of greater than five percent GDP growth, real GDP growth. I think productivity is going to go vertical because of AI and deregulation. And I think there could be a world where this is the late nineties and it doesn't sound like a big difference, you know, five or six percent versus two or three percent, but it is a massive difference. You know, at five or six percent, you know, the economy is doubling. You'll call it every 12 years, roughly. First 24 years at three percent. I mean, it's just, it's a massive difference in terms of, kind of, the wealth of the country and individual Americans. As far as a specific prediction for 25, uh, because I don't know when that will happen, I think you wil…

AI assessment note: “My most contrary belief, so I think that America over the next... will print”

Q Year of the Robots is your prediction. I love it. Uh, Gavin, what do you got?

A Profoundly agree. I think Year of the Robots, um, inclusive of FSD. You know, I think FSD works today and it's gonna, It's, it's gonna cross into mainstream adoption where I already notice, particularly if I'm taking an Uber late at night, I really, really prefer to have a Tesla. Just sometimes you get an Uber driver who's, who's tired. And I just feel a lot safer if they have the FSD running. And then obviously using it for yourself is amazing, but I think it's gonna continue compounding at an accelerating rate. I think the, in a lot of ways, what I'd say broadly is, I think for a while, it's going to be big businesses are winners, big businesses that use AI thoughtfully. And the reason is, is that with what is happening, what, what Oh three showed us opening eyes new model. A combination of reasoning and test time compute. Like, I think you're, if you're a big business and you can pay a million dollars to let an AI think for six weeks about the most important question for your business. That's going to be a profound advantage relative to small businesses that can't afford that. And by the way, we are the other, like inference compute is going to be the, the kind of derivative winner of that. We're going to run out of GPUs, accelerators, compute in 2025, the same way we did in 23.

AI assessment note: “Profoundly agree. I think Year of the Robots, um, inclusive of FSD.”

Q Exactly. We'll know exactly. That's right. Pull the ETF. Absolutely. I mean, Gavin, what do you think of this hot take here from, from Mr. Freyberg?

A Really cheap. I've had a, I would call it a no China guideline ever since the long top financial fraud more than 15 years ago. Or it was just amazing. It was, you know, uh, a lot of great investors owned it and You know, they, they talked a great game and they had a Western auditor, but then it turned out that the fraud was happening at the local post office where they were opening up the documents that the local auditors had signed off on changing them and then sending different documents, you know, you know, whatever, FedExing them. Yeah. So I just think it's, it's a hard place if you're, I think if you're not Chinese to make money, But I tend to agree with a lot of what David said. I think Trump and Xi both want to deal. I think there's a deal to be done that leaves Putin out in the cold, as I referenced. And if that happens, Katie, bar the door. I mean, there are really, really high quality companies, you know, that are, some of them are mid single digit multiples.

AI assessment note: “I tend to agree with a lot of what David said.”

Q is a great pick. Now the other picks, there's some whack pack picks in there. I'll be totally honest. I don't know what the strategy is, Joe. I'll ask you that in a minute. But what do we think, um, Gavin, of this sort of approach here, which is really thinking thoughtfully about what is sophisticated and, and what are we doing here? What is the outcome we're looking for?

A Yeah, no, I think it's great. And you could just, I mean, the reality is a hedge fund that runs with a lot of leverage probably is more risky. I don't know that it's more risky than a single security, but, um, Like at the end of the day, since he wrote that letter, it was an incredible 15 year run for private equity. And, you know, now all the big private equity firms are making a huge effort. Now, um, when probably the business is more mature and maybe the return opportunities aren't what they were to appeal to Main Street America. Like, it would have been cool if, you know, Blackstone or KKR or whoever in, oh, seven, oh, eight, oh, nine, 2010, you know, could have had their fund up on, like, the Fidelity marketplace for subscription. That probably would have been good for America. So I think his comments are well taken.

AI assessment note: “Yeah, no, I think it's great.”

Q They invert the X axis so that as you move to the top right, it's better, right?

A Exactly. And I mean, this is now, and this is on cursor bench and XAI, you know, SpaceX is, you know, almost certain, you know, they have a right to acquire cursor, they're working with cursor, so maybe this is grading your own homework. Well, right here on the right, this is from Databricks, which just raised money at a hundred and ninety billion dollars. Um, this is a very real company that is very sophisticated, and you can see, like, even on this, you're well ahead of, um, Fable Five, Which, you know, I think most people would, would, would kind of agree is the, um, You know, the gold standard. Like this is your, your, your higher quality. That's a slightly lower price. So I think Grok's 4.6 is, um, you know, the numbers are the numbers, you know, on X, I think we put it down. Um, people, you, you know, you, we do have to wait and see the vibes, but I think it's pretty interesting. Um, DHH, who I think created Ruby on rails.

AI assessment note: “Exactly. And I mean, this is now, and this is on cursor bench”

Q ago. Some were in quantum computing. Generally, a lot of it, though, was in the NVIDIA ecosystem, so I think that the, the US government is really trying to make sure that these next-gen GPUs don't go directly to China, but also don't go to countries that could then redirect them into China. Did you get a chance to look at that, or did your team look at that yet?

A Oh, yeah. And I think this will be part of You know, the upcoming negotiation between America and China, but it is hard. You know, it's, if we can't, you know, if America cannot keep illegal drugs out of, you know, out of our own country, these GPUs are arguably dramatically more valuable per Unit, you know, like a Blackwell is like, you know, the size of this, and it's just hard, particularly like, you know, America, we're trying to keep drugs out. Here, China is trying to bring the GPUs in, and it's kind of significantly, you know, I think harder to prevent than, than, you know, smuggling of illegal drugs. Now, everybody does their best, but, um, Yeah, the United States, I think it's always going to be kind of a game of cat and mouse until you get some sort of grand bargain, and right now they, they're allowed to sell certain kinds of GPUs into China, and I'm sure they'll be.

AI assessment note: “Oh, yeah. And I think this will be part of”

Q also is, you know, do we see an evolution if, if the kind of increment in performance relative to the investment in Net kind of training compute resources declines. Do we start to see a shift in, um, how the architecture of the systems are run? Meaning like, do we start to build models of models and that starts to resolve a higher level architecture that unlocks new performative capabilities?

A So I would just say we're already building models of models, you know, almost every application startup I'm aware of is chaining models. You know, you start with the cheap model, you check the cheap models work with, uh, More expensive model. You know, lots of very clever things are being done. You know, every, every AI application company has what's called a router so they can, you know, swap out the underlying model if another one is better for the task at hand. Um, As far as what the wall is, there's been a big debate that we were hitting a wall on these scaling laws, and the scaling laws were breaking down, and I just thought that was deeply silly, because no one had built a cluster bigger than, you know, 32,000 H-one hundreds, and nobody knew. It was, it was a ridiculous debate, and there were, you know, really smart people on both sides, But there is, there, there's no new data. Grok three is the first new data point to support whether or not scaling laws are breaking or holding because no one else thought you could make a 100,000 hoppers coherent. And I think based on public reports, they're going to 200,000, um, hoppers. And then the next chick is a million. It was reported. They're going to be first in line for Blackwell, but Grok three is a big card and will resolve the This question of whether or not we're hitting a wall. The other question you raise, David, is very …

AI assessment note: “So I would just say we're already building models of models”

Q from 2.4 to 2.7%, 30 basis point increase. Over May. So any thoughts, Gavin, on these macro issues around inflation coming back a bit, the, the stock market at an all-time high and unemployment? If you believe the unemployment data, and we've had a bunch of debates here about that, you know, how, how correct is it being at close to an all-time low in our lifetimes? Let's go macro.

A Yeah, I wouldn't. Read too much into the CPI bouncing up a little bit. There's base rate effects, I think on a month over month, uh, basis. And if you looked at, you know, what, what economists call core, super core, different, you know, ways to maybe smooth out CPI. I think inflation is still relatively contained. As far as, but I mean, A lot of very smart people are convinced that tariffs will lead to inflation. We'll see. There's certainly sound theoretical arguments for why they will, but it does look like maybe the, um, overseas exporters are eating a little bit more of the tariffs than, than people thought that they would, which is, which is good for US inflation. But, you know, the combination of a very weak dollar and tariffs theoretically should lead to a little more inflation. Firing Powell. I worry that the market only went down one percent. I do think the market would go down quite a bit more if Trump did fire Powell. I think it would be a mistake. I hope he doesn't do it. And, you know, there are very sound reasons for the Federal Reserve to be independent. And I hope he does not read from that trial balloon down one percent that that's all that would happen. Look, the world would continue spinning. America would go on. But it would be a mistake.

AI assessment note: “I wouldn't. Read too much into the CPI bouncing up a little bit.”

Q hear because they hear no job, but he does say, hey, universal basic income is probably going to have to come into, you know, place, and he said that multiple times. You have Dario, according to Chamath, talking his own book, scaring the bejesus out of people in order to get regulatory capture. What do you think is going on here, and how do we do better as an industry?

A I think Chamath outlined, like, a very viable and positive path forward where just, you know, real people who are not, you know, at the tip of the spear, these are the positive impacts AI's had on my life. I was at a I was at, I was at an event maybe 10 days ago, and someone who runs a hedge fund, his daughter was born with a very rare genetic mutation that effectively would have normally condemned her to a life devoid of joy, meaning everything. The neurons in her brain were not firing, so she wouldn't, you know, who knows what her life expectancy would have been or what the quality of her life would, would have been, and it's, it's a tragic disease. He said he didn't accept that as an answer. He found, he did an enormous amount of research with LLMs and found an existing safe drug on the market that they thought would have a meaningful impact on his daughter's condition, and it did. It took, I think the percentage of times the neurons were firing was 30 or 40%, and it took it up to 80 or 90%, and that means that she can live a normal life. She may not be as smart as she would have been, but she can live a normal life. And he's now figured out how to use AI, how to, to further tailor that drug. And, you know, there've been all sorts of advances in protein design, et cetera, et cetera. And he's reasonably confident he's going to have a drug in months that is a complete cure. An…

AI assessment note: “Chamath outlined, like, a very viable and positive path forward”

Q Do it right now. Are you going full conspiracy theorist? Are you, are you trying to get your seat here by being a little more Alex Jones? Do you believe those drones were aliens? What's going on?

A So I don't know, and I think it's very clear. If you look at all the, um, you know, statements from the New Jersey mayors and governors, you know, who've met with the police, the FBI, The Defense Department that they don't know. Now, you know, Trump said someone in the government knows what they are. He also said he was not going to go to Bedminster anytime soon, his resort there. I just think over the last, I'd say eight years, every 18 months, there's a big story in the New York Times, the Washington Post, the New Yorker, the Atlantic, very credible media sources. With dozens of interviews with fighter pilots and commercial pilots talking about seeing things with advanced sensors that made no sense to them. And then there's kind of a taboo that went away. There was an article in the New Yorker where they quoted from a bunch of people who were at, you know, kind of the, the skunk works laboratories in the 19 fifties. And so they absolutely saw Extraterrestrial materials that have been recovered from a crash. That's one reason the US made such big leaps in material science. And, um, you know, there was the Uamata thing where essentially a lot of, you know, astronomers said, Hey, this is, this is clearly a UFO of some sort, including the head of astronomy at Harvard. And then just the conspiracy theory I would have is if, if, if, if, if, These are actually UFOs and not, you know…

AI assessment note: “it seems like the most likely explanation in New Jersey is that some sort of a drill.”

Q But what does it mean, Gavin, when like the accounts receivable jumps from a 1,000,000,005 or so to like five and a half billion year over year? Is that concerning or that's just business?

A So look, it's never good, but if you think NVIDIA has gone through the biggest Product transition in history in terms of Hopper. You know, this is a business doing tens of billions of dollars at scale. There's never been a product transition like this in the history of semiconductors going from Hopper to Blackwell, you know, kind of the 2022 generation GPU to this one. The only precedent for this on planet earth is the iPhone. And There's something called the Osborne effect, which is, you know, when Apple comes out with new iPhone for the three months before, nobody buys the new iPhone. And I think the only reason NVIDIA was able to grow through this product transition is because of reasoning models like DeepSeq, which are just so compute hungry, and this product transition does come back to the day's receivables.

AI assessment note: “So look, it's never good, but if you think NVIDIA has gone through”

Q Gavin, can you say how long it's going to take to stand up the fab at TerraFab?

A Well, I mean, if it was a normal fab, it would be a two, three, three and a half year process. But, you know, we, we, we've seen what Elon has done to other construction processes. And, you know, he's starting with some, some advantages with the Intel partnership. So I don't think anyone knows, but based on past history, he's probably going to stand up terra fab faster than other fabs have been stood up. But it, what, this is really hard. It's really hard. It's the intersection of magic and science. You can't believe how complicated this is. So it's going to be hard, but you know, he has a, he has a track record of doing, you know, what Jensen called impossible superhuman. And so we'll see, we'll see how long it takes.

AI assessment note: “if it was a normal fab, it would be a two, three, three and a half year process”

Q It's got such an evocative name. What does it mean?

A It is a list. I, everybody, you could just Google it, but I mean, the questions are ridiculously hard. They're extremely hard questions from a wide range of disciplines. And you can see Grok IV when it came out and it is improved. Since, since its release roughly a month ago, was at 44.4%, and GPT five was at 42. And similarly, RKGI two, which is another benchmark, or something called the jagged frontier in AI, where things that are very easy for AIs, like maybe math, Are really hard for humans and vice versa. And RKGI measures tasks that are relatively easy for humans. Maybe you could almost think of, you know, there's a lot of like geometric puzzle solving, but very hard for AIs. And here again, Grok IV is, you know, significantly better than GPT-V. Now in fairness, and, you know, to be balanced and accurate, it is narrowly, very narrowly ahead in artificial analysis on the right. But this is, this is the first time they have not released a model that wasn't decisively better. And the progress for Grok four was exceptional. Grok five's coming soon. You know, they have more, and I, I should note my firm. Is a shareholder of XAI. So, you know, as much as I try to be objective, I am for sure biased because I am human and to be human is to be biased, but, um, they have the largest Blackwell cluster. So I'm, I'm optimistic for Grok five, but you know, we're, we're waiting on Gemin…

AI assessment note: “It is a list. I, everybody, you could just Google it”