Salim Ismail 与 Dave Blundin 谈 Bitcoin、AI 与全球权力新格局｜第172期

The experts you speak to, everybody on the street—they're expecting that we're about to see a bull run. We've been waiting for a long time for Bitcoin to peel off its tracking of the Dow and the financial markets, and I think this is finally the breaking point.

Today, Amazon announced, “We are partnering with Humane, Saudi Arabia's newly created AI innovation company.” NVIDIA is selling 18,000 of its top GB300 Blackwell AI chips to Saudi Arabia. The freedom to just sit back and act, having infinite or near-infinite capital and instant decision-making, is a pretty powerful thing.

I think that, at the macro level, the combination of U.S. innovation and capital coming from the Middle East makes it kind of unbeatable for what the future holds. The safest bet by far is to run like hell toward your targets. You'll be very glad you didn't waste a millisecond between now and 2030.

Now that's a moonshot, ladies and gentlemen. Everybody, welcome to Moonshots and a WTF Just Happened in Tech episode with my Moonshot mates, Salim Ismail, the CEO of Open Exo, and my longtime best friend and fraternity brother, the head of Link XPV, Dave Blundin. Two of the most extraordinary individuals, and we're here to deliver you the real news that happened this past week—the only news that matters in technology that's changing our lives. You give us a couple hours, we'll tell you what just transformed.

Dave, welcome to Moonshots.

Good to be here. Good to be here. Thanks, Peter.

Every episode, it's like “Where's Waldo?” You are a traveling probability function around the planet. Where have you just come back from?

I just came back from Palermo, with a canceled flight in Barcelona and a day and a half to get back. But it was nice to be back and nice to be there.

There's this thing called Zoom where you can actually go there digitally and dematerialize yourself.

This is an important point. We thought when video conferencing came out in the '90s, it would kill the conference business, right? You could get all the content online. As we've evolved digitally, we go to more conferences than ever because we value that human presence.

You and I both would kill not to have to travel ridiculous distances to speak to groups, but we end up having to do more and more of it.

Yeah. I don't know. I feel like this podcast has a lot more reach than whatever you're doing on those trips.

It has way more reach, but the problem is that the demand is still there. People say, “Yeah, anyway, they want you.” I know that.

Even crazier, I did this TEDx talk a few years ago called “Fixing Civilization.” I got called by a company that said, “We'd like you to come and do that talk.”

I said, “Why don't you just play the YouTube video?”

They said, “No, we really want you.” I got there, went to the venue, and it was literally a TV studio with cameras. They were broadcasting to their offices. They literally could have just played the video, and I was like, “This is insane.” That part I don't understand, but okay, we'll go with that.

Oh, because it can't be understood.

Well, Peter, you shouldn't bug Salim too much. I saw you in New York this week. You're back in L.A. now, you're going to San Francisco, and then off to Hong Kong.

I was supposed to be going to Florida tomorrow, and the meeting got canceled. I was like, “Yay! I'm so happy.”

Pot calling the kettle black.

By the way, congratulations on New York on Monday.

Thank you. We'll talk about that. We just gave out $12 million of the $111 million in our Healthspan XPRIZE, helping people live an extra 20 healthy years.

Your birthday is coming up this weekend, I think. Isn't that true?

I turn 60 in 48 hours.

Amazing. Amazing.

60 down, 180 to go.

180 to go. Exactly.

All right, let's get into this week. A lot of news is happening in the exponential world, in the moonshot world, and my moonshot mates are here to help me tell this story.

Let's begin with Bitcoin. A lot of motion. For those of us who are huge Bitcoin believers—I think all 3 of us—we're back into the $100,000-plus zone. Welcome back, everybody. I think we're poised for new heights.

This is interesting. Michael Saylor, who I look forward to having back on the podcast soon, has doubled his Bitcoin purchasing plan. Dave, you want to lay this out for the team?

Nothing slows down Mike. He's going to stay aggressive. Doubling down: 21/21 is $21 billion of equity and $21 billion of debt. That's a lot. And 42/42—I think, Salim, you calculated the fraction of all Bitcoin that that implies. It's a pretty material fraction that he would need to buy.

It's a big number, and he's making a massive bet. The bet seems to be paying off. If you're a macro believer in the future of Bitcoin, then the way he's engineering this is about as amazing as you could imagine.

Yeah, Peter, you mentioned last time that there's a pretty big arbitrage spread between the price of Bitcoin and the price of MicroStrategy stock. There's a little less than a 2x difference there. On the other hand, MicroStrategy is leveraged 2-to-1 because of the debt component.

If you're a believer, it's actually a leveraged way to invest, which I think is why it's attracting a lot of people. Why would I just go buy a Bitcoin? If you buy a MicroStrategy share, you're leveraged on the upside. The people buying the debt are predominantly hedge funds that are buying the debt and then shorting the stock concurrently, so they're making money either way. That's why there's zero interest on the debt.

The one thing I love about what Mike's done in 21/21 and 42/42 is that he's gotten rid of all the interest payments. There's no real existential threat to MicroStrategy anymore. When he borrowed the money originally, he was financing the debt from the cash flow of their core software company, which had scale limits because he could only pay so much interest. Now, in his new design, it's infinitely scalable because there are no interest payments. It's pretty brilliant.

Mike's also making a call to the world—to all corporations, both private and public—to go on the Bitcoin standard. Of course, the more companies he gets to buy Bitcoin, the shorter the supply of Bitcoin, the higher the price goes, and the more his plan works.

This plan of $42 billion of stock sales and $42 billion of debt—$84 billion worth of Bitcoin being purchased—is massive. Of course, he's also pushing sovereign Bitcoin reserves. There's only 21 million Bitcoin, and we discussed this in the last episode, or a couple of episodes ago: if all the millionaires in the world wanted to buy some Bitcoin, they would get access to only a fraction of a Bitcoin. There's not enough to go around, and the float is relatively low.

When we were talking to Mike in Miami a couple of months ago, Peter, the one thing he opened my eyes to is that we're talking about sovereign wealth funds and sovereign treasuries putting a fraction of their money into Bitcoin. When you looked at the pie chart as a fraction of global wealth, it was this tiny, tiny, tiny little slice—all Bitcoin combined.

He doesn't need every country in the world to move its treasury to Bitcoin. He needs 1% of the countries in the world to move 0.1% of their treasuries, and that would absorb all the Bitcoin in the world. It would actually drive it to the prices he's predicting—way north of $1 million, north of $10 million per Bitcoin.

The number we ran was that if 1% of the Fortune 1000 put 1% of their treasury in Bitcoin, that would drive it to $1 million per Bitcoin, which seems more and more likely.

It's so funny. I've been buying Bitcoin along the way, and every time I sweep some capital in from the sale of a stock or some deal I did, I'm like, “Okay, I'm putting it into Bitcoin.” I am putting a certain percentage in MicroStrategy as well. This is not investment advice; it's just what I'm doing.

One of my financial team members said, “Should we wait for it to dip?” I said, “You can't do that. You can't play that game. You have to just buy and hold.” Whether it's $103,000, $97,000, or it goes up to $120,000, just buy it and hold it. You can't try to guess the market, because what you see over and over again is that the Bitcoin price will jump, like it did last week, by $7,000, $8,000, or $10,000 over the course of 24 hours, and then it's flat for a long time.

If you miss those trading days by selling in and out, you've lost all the upside potential. I think there's something like 10 days in the year when Bitcoin does 70% of its upside. If you miss those 10 days, of course, the challenge is knowing which 10 days they are.

Yeah. Crazy.

Here's more news. This is from Brian Armstrong: “Crypto is about to be in everyone's 401(k). My goal is that, in 5 to 10 years, getting into the Coinbase 50 Index will feel as good as this.” That's a quote from Brian Armstrong, the CEO of Coinbase, who will also be on this podcast soon enough.

Coinbase joined the S&P 500 days after Bitcoin soared past $100,000. Salim, what are your thoughts here?

Finally getting legitimacy. I don't think there's any turning back. I think the macro world and the money world will be forced to start addressing this.

I was talking to one of the top wealth-management people in a particular country, and they won't let their advisers recommend Bitcoin.

It’s crazy. The advisers are going nuts because their clients are coming to them saying, “I want to buy Bitcoin,” and the company won’t let them because of legal liability or legal risk, as determined by their risk folks. So there’s a complete clusterfuck going on around what they’re allowed to do. That pressure will slowly ease as we get more legitimacy into the space, and I think this is great news overall. In the meantime, people are buying MicroStrategy or other companies that are following suit.

Yeah. Next time we circle up on Bitcoin, I’d love to talk more about the agent-to-agent AI economy. Oh my God—micropayments. Putting Bitcoin into a 401(k), I’ve been working on that, thinking about that for years. It’s clearly getting legitimacy now, but the new thing is AI payments and how that whole economy is going to evolve. That’s clearly going to require digital currency. You’re not going to transact in dollars.

Most of you know that the news media is delivering negative news to us all the time because we pay 10 times more attention to negative news than positive news. For me, the only news worthwhile that's true and impacting humanity is the news of science and technology. That's what I pay attention to. And every week I put out two blogs. One on AI and exponential tech and one on longevity. If this is of interest to you and it's available totally for free, please join me. Subscribe at diamandis.com/subscribe. That's diamandis.com/subscribe. All right, let's go back to the episode.

David Bailey and BTC-native holding company Nakamoto merge into KindlyMD to establish a BTC treasury. Salim, this just speaks to the fact that Michael Saylor has been operating on a lone-wolf basis, and now other people are going, “Wait a minute, why is he getting all the upside and good fun here? He’s getting a huge chunk of the pie.” I think we’ll start to see more and more of this. I think we’ll see a flood of this type of announcement in the next 6 months as people say, “We’ve got to get into this.” It’ll be a self-fulfilling strategy because then it drives the price upward. All the while, Dave, are you seeing this in any of your companies?

I mean, I told you I greenlit them all to do it, and I’m just waiting. The bellwether for me is when they call back and say, “Hey, Dave, is it still okay if we put part of our treasury into Bitcoin?” I’m just waiting for that call. I’ll let you know the minute it happens.

Please do, because that is an inflection point. We’re about to see a price spike. I mean, when my mom says, “Should I buy Bitcoin?” but then her financial adviser says, “You can’t do it here,” I say, “Hey, buy MicroStrategy instead.” That’s right. It’s crazy.

All right, let’s continue here. This is another piece of that puzzle. Corporate Bitcoin purchases now outweigh the supply of new Bitcoin by 3.3 times in 2025. Salim, this is what you were speaking to.

Yeah. I think it’s going to start to become mandatory for corporate treasuries. I look forward to the day that somebody will be deemed unworthy of that role if they don’t put Bitcoin in the treasury.

Ah, fascinating. I mean, this is basically saying we’re mining Bitcoin at a certain rate, and the availability of that new Bitcoin, plus what’s available for sale, is being outstripped by demand. That will drive the price through the roof, for sure.

It’s really interesting to think about the original protocol document. I don’t know if you ever read it, but the original document expands the supply pretty aggressively for a period of time, then it’s capped, and then it just trickles out with mining after that. The amount of foresight in that design is just unbelievable. Everyone, it should be mandatory in school to read that document. It really should be.

This was one of the most prescient documents. And, you know, we’ve talked about this before: There have been many attempts at creating a viable digital currency or store of value over the decades. Lots of people waited 2 or 3 years to see if Bitcoin would actually bite, because all the others kind of fell apart after a year or 2. The damn thing held together so perfectly. It’s unbelievable.

Here’s a homework assignment for those listening who want to play in this: Go and download the original Bitcoin white paper. Stick it into Google’s NotebookLM and just listen to a podcast about it, or ask for an understandable 2- or 3-page summary. It is a beautiful work, and the most extraordinary thing about it is that it’s working. It’s never been hacked, and it has become the dominant store of value.

It’s amazing.

Yeah. All right, here’s just a chart to back it all up. The experts you speak to, everybody in the street—they’re expecting that we’re about to see a bull run. BTC is signaling a comeback, reclaiming the key 1-year MVRV momentum average. I’m not someone who looks at the charts and studies them myself, but this seems to be self-evident. I think what’s noteworthy here is that we’ve been waiting for a long time for Bitcoin to peel off from tracking the Dow and the financial markets. I think this is finally the breaking point, and I hope this continues. If it does, we’re going to see a massive bull run.

Yeah. From your mouth to the universe’s digital ears.

All right, let’s jump into our favorite subject: AI. This is a huge week for AI. In fact, every week is blowing apart the previous week. It’s unreal. This is a global power grab in the AI universe, and it was centered not in Washington, D.C., and not in Silicon Valley. This week, we were in Riyadh, where the tech CEOs from the U.S. joined President Trump to do a deep dive into a Saudi investment forum.

Dave, this isn’t even the full list. Jenny Johnson, who runs Franklin Templeton, is there. Larry Fink, who runs BlackRock, is there. I mean, this is ridiculous. Dubai has historically attracted that crowd. Now Riyadh is catching up. There’s an unbelievable amount of activity and investment, but boy, did they get the best of the best worldwide to show up this week. Amazing.

You know, I’m on the board of the Future Investment Initiative, FII, which is under PIF, the Public Investment Fund of Saudi Arabia, and I help address the AI agenda at FII every year in Saudi in October, in Riyadh. Saudi Arabia wants to be number 2 to the U.S. in the AI world, and they’re prepared to bring a huge amount of capital and relationships to the table. So this was a massive success for the ministers and for the ruler of Saudi.

Let’s dive in a little further. This is a tweet from Andy Jassy, Amazon’s CEO:

“Today, Amazon announced, ‘We are partnering with Humain, Saudi Arabia’s newly created AI innovation company, to collectively invest $5 billion and to build a groundbreaking AI zone in the kingdom.’”

Let’s pair that with the next article: Nvidia is selling 18,000 of its top GB300 Blackwell AI chips to Saudi Arabia. It’s a $10 billion commitment by Jensen Huang. Humain is owned by Saudi Arabia’s Public Investment Fund and is working to develop AI models and data center infrastructure. This is a huge deal. Dave, how do you think about this?

For starters, there’s some great video coming out of the conference. A lot of it is public. So, if you have a free minute after this podcast, you might want to look at David Sacks’s meeting with His Excellency Abdullah Al-Swaha in particular.

We’ll show a clip on that next.

Great. Yeah, they’re basically taking a lot of the handcuffs off chip exports within this region as part of this conference, and that was kind of inevitable. Saudi also has big investments in CoreWeave and Cerebras, so they’re working on their own infrastructure as well—or, more accurately, unrestricted infrastructure. So this isn’t the only thing going on, but it’s very consistent with a couple of things.

The Liquid AI team was telling us that Arabic in particular has 1,000 dialects, and none of the U.S. foundation models are addressing that at all. So they really want to be able to control the regionalization of the technology and also the cultural values. We’ve been talking about this a lot: having a culturally matching version of these AIs, because what’s going to happen next is anyone under the age of, say, 20 is going to spend a lot of time talking to their AI. A lot of what their worldview is comes back from whatever the AI is telling them. It’s very important for various regions around the country to have some say in what it says and how it interacts with the next generation of kids.

I think one of the most relevant things here is that this places Saudi Arabia squarely in the U.S. camp as opposed to China. They’ve teamed fully with U.S. infrastructure, U.S. partnerships, and U.S. investments, making a massive commitment to the U.S. AI infrastructure. So that’s huge. And Nvidia agreeing to give them 18,000 of its top Blackwell chips is massive. I mean, countries around the world are frothing to get access to that.

Interestingly enough, we have both Saudi Arabia and the UAE making massive financial commitments to this. What we saw about a decade ago was these 2 nations making massive commitments to new energy and environmental investments. It's the advantage of being a nation ruled by a ruler who can say, “Okay, yay, verily, we're going there,” and not having to deal with a full Congress and Senate and democracy. Their ability to pivot so rapidly and say, “Listen, the most important thing in the world is AI, and we're committing $1 trillion—and it will be multiple trillions—into AI” is a massive advantage for them. Salim, you've had these conversations.

It is incredibly impressive to watch. We often talk about governance models, and it's always clear that the best model is a benevolent dictator. The problem is, in many cases, a dictator doesn't stay benevolent. But in these cases, they actually seem to be very much so.

I've spent more time in the UAE, and the stories you hear about His Highness Mohammed bin Zayed are unbelievable in terms of the character of the man and the sheer integrity that's coming through. Then that cascades down all the way, very quickly. It's super impressive to watch, and the fact that they're making these investments—I think at the macro level, the combination of U.S. innovation and capital coming from the Middle East makes it kind of unbeatable, what the future holds.

Yeah, for sure. Let's go to the next slide here. This is Minister Abdullah Al-Swaha having a conversation with David Sacks. I'm going to play this video in a moment. Abdullah Al-Swaha is the minister covering space and communications. He was at Cisco before the ruler of the kingdom tapped him to come back and join as a minister. In fact, Dave, you and I hosted him in Venice in a makeshift majlis. What, about a year ago?

Venice.

Venice, California, by the way.

Just Venice, California.

No, it was great. He's a brilliant guy, and he's close to Elon, close to all the tech CEOs. All right, let's play this interview between David Sacks and Minister Abdullah Al-Swaha.

Technologies have helped us deliver a passport within 5 minutes. You said to me, “We're going to soon see that in other economies as well: how you could renew your license, how you could do all of your different digital services in a cashless society as well, from the digital age.”

In the intelligent age, we've shown you firsthand how we're creating nanobots with CRISPR technology, leveraging generative AI to tackle sickle cell disease and bad cholesterol, and have targeted gene therapy, which is remarkable—reducing the cost to treat a patient from $2.2 million to just under $100,000.

Think about the remarkable work that we're doing in agentic AI, helping Aramco energize the world, which we need to fuel the U.S. ambition and Saudi's ambition in AI training and inference, because we're going to converge not only to the price point of electronics, but it's really the price point of electrons.

Amazing. We're seeing huge investments by the government across all exponential technology areas. What's incredible is that, by virtue of their size and their technology base, they're outstripping other nations around the world. They don't come close to what we're doing in the United States, but they're focusing their capital, driving breakthroughs, and moving quickly.

I think in that video clip—if you have time to watch the full video—David Sacks has some brilliant things to say, as always. But you can sense that they have complete freedom, tons of capital, and complete freedom to act quickly in this AI environment. You can feel that vibe just in that video clip.

David Sacks is within the constraints of congressional votes, wearing a suit and tie and dress shoes. You can just feel the difference in the ability to be nimble. I remember when I was a kid growing up, mostly in Iran as a young child, my mom would always tell us, “U.S. foreign policy is going to be a disaster in this country.”

She learned the language, spoke Farsi, and really got integrated into the culture, which very few Americans in the country did at the time. She said, “The problem fundamentally is that Americans—every 4 years, we completely change our mind on what we're trying to achieve in this country.” There's no consistency whatsoever over time. Boy, did she turn out to be right about that.

So you're right, Salim. This AI time frame is so short, and there's so much that needs to be done so quickly. The freedom to just sit back and act, having infinite or near-infinite capital and instant decision-making, is a pretty powerful thing.

During COVID, when we had to make decisions very quickly, every major democracy failed in navigating COVID. Well, actually, every major country failed; all the small countries did very well. So I think there's a hint there as to the future of governance as you go to a micro level.

Everybody, I hope you're enjoying this episode. Earlier this year I was joined on stage at the 2025 Abundance Summit by a rockstar group of entrepreneurs, CEOs, and investors focused on the vision and future for AGI, humanoid robotics, longevity, blockchain, basically the next trillion-dollar opportunities. If you weren't at the Abundance Summit, it's not too late. You can watch the entire Abundance Summit online by going to exponentialmastery.com. That's exponentialmastery.com.

You know, Dave, you said something that's really important a second ago. I just want to point it out: speed here. We're going to see the AI—I don't call it AI wars—AI dominance play out over the next 2 to 3 years, 2 to 4 years. At which point, I don't say it's game over, but who's on top and who's second and third is unlikely to change after that.

What we're seeing is really a jockeying for position between major nations, in particular China, the Emirates, and Saudi Arabia. Qatar will play in there as well. I don't think Russia is going to be relevant unless they team up with China in some fashion, and then whoever is leading the game is going to make the rules. I mean, do you agree with that?

Can I offer a counterpoint? If I give the counterpoint side, the speed of democratization is such that even if one domain, one kind of pairing of partners, goes ahead of the other, it'll level off very quickly after that. We've seen that over the last couple of years. You've got OpenAI, then you've got Gemini, and you've got DeepSeek, and that gets overtaken.

I think it's going to democratize much faster than in the past, and therefore you won't see one taking off in a winner-take-all type of outcome. That's my view.

I think it's a safer view if that's the case. But I think we'll have a conversation on this in a few minutes about how AI is beginning to self-code, and we have the potential for a runaway digital superintelligence race with whoever's on top. Dave, how do you think about that?

Well, Peter, your first statement was so right, and you can't overstate the winner-take-all effects that are going on. There isn't currently a force to democratize any of this, and it would take an amazingly huge amount of inertia very quickly to democratize all of this.

The forces of concentration of everything are just unbelievable because of that self-improvement effect Peter just mentioned. I'm telling everybody around here, everyone around MIT, everyone around Harvard, everyone around my kids: You have a window of opportunity that's exactly the timeline Peter laid out—about 4 years—to set the course of your entire life.

Nobody knows what's going to happen after that, but the forces of concentration are much stronger than the forces of distribution. The safest bet by far is to run like hell toward your targets. In the event that something happens that spreads out all the value, you're fine. But in the other event, where it's all concentrated, you'll be very glad you didn't waste a millisecond between now and 2030.

I think there are 2 different things here. One is exponential opportunities if you move fast, which I 100% agree with. But the other part of winner-take-all, I will disagree on, and let's see who's right over the next little while.

Well, we will see. Ultimately, it's going to play out over the next 2 to 4 years. Why do you think the entire AI leadership of our nation and the financial leadership of our nation showed up in Saudi Arabia for a week? Everybody—not just a few people—there was a significant enough meeting and a significant enough purpose of that meeting to say you had to be there.

I think this was—I think we'll draw a line from last week to consequences that will be dominant for decades.

Yeah. I cannot tell you how unnatural an act it is for a Saudi Arabian to wake up one morning and invest in CoreWeave or Cerebras. That doesn't just happen. I was like, “What?” There's clearly a really intentional and intelligent awareness of where the bottlenecks are, and those investments are turning out to be brilliant, by the way.

But chip design—Saudi Arabia? What's the connection there?

The connection there is that it's the bottleneck to self-improving AI. We can go on about this for a while. When I was meeting with Minister Abdullah Al-Swaha, he said something that really stuck with me: “Every week, I have a meeting with MBS to talk about AI.” It is his number 1 item on his agenda, and it should be.

It drives me nuts that it's not the number 1 item on every senator and congressman's agenda here—how it impacts health and how it impacts education.

It’s so transformative. All right, let’s go to this news clip from Marc Andreessen on Google and OpenAI’s visions. Let’s take a listen to Marc.

Kind of a big little company, kind of big-company thing, and I think that’s repeating itself. One of the ways to think about what’s happening right now is OpenAI is growing up to become Google, and Google’s reinventing itself to be OpenAI. There’s a similar thing there.

I think that’s fascinating, right? So OpenAI is going into the search business and wants to dominate, and of course Google’s been in the search business. It really needs to go Gemini-first for the next few years. It’s true, huh? What do you think about that, Salim?

I think it’s very eloquent, as usual, for Marc. The thing that I find fascinating is that Gemini is probably the best model out there for general use, and yet OpenAI just continues to soar in adoption.

Yeah. It’s like when my middle schoolers are like—all they talk about is ChatGPT and OpenAI. Seriously, they’ve captured the market. Let’s talk about the relative value. We have Google at $2 trillion and OpenAI at $300 billion. Just $300 billion. Dave, how do you think about the relative value?

Well, and then NVIDIA is at $3.3 trillion today, so NVIDIA is almost 2 times Google.

Oh, my God, that’s crazy.

Yeah. I tell you, there are some real interesting quandaries in there because Google is losing search. There’s no doubt about it. You can only debate the rate, but it’s moving. Google’s in a good position to capture a lot of that, but at a much lower value per impression, so it really cannibalizes their revenue machine.

On the other hand, YouTube and video content are killing it, and they benefit tremendously from synthetic video creation and also ad optimization using AI. Those tailwinds are very strong for Google. So you’ve got the core dying at the same time the new things are growing like crazy.

Then, within GCP—Google’s cloud—they have TPU 7, the new TPUs that have this incredible weight-stationary systolic-array design, which allows them to be much more efficient than NVIDIA at running these core AI processes. I could go on for hours about it, but I’m telling you, this is the future of AI chips, and they’ve got it ahead of the curve.

It’s going to be completely sold out, and they’re going to redirect it into GCP and use it within their own internal use cases. I don’t know that it’ll ever end up not being sold out before they can even manufacture it. Then you’re like, “Okay, NVIDIA’s in a race with Google over who controls the AI chip stack,” which no one’s even talking about yet. It’s all bottlenecked at what TSMC is willing to make, for whom and why.

This was going to be the best TV series ever. I was going to say soap opera, baby. It’s Silicon Valley soap opera. Of course, the whole Taiwan–China situation puts this in an incredibly crazy place. So what does China say?

They want to own Taiwan by 2028. Yeah. Crazy. And I think the federal estimate—they don’t talk about it much—is that you have to make the assumption that there will be no access to chips from Taiwan by December 2027, which is nowhere near enough time to get the U.S. fabs up and running.

So how’s that going to play out?

It will play out disastrously. We know 100% for sure that if China takes Taiwan, the fabs won’t be operating for China.

Yeah. They will destroy them. They’ll shut them down. All right, let’s go to our next article. This one does my heart well, having been in the pharma-related industry. The FDA announces completion of its first AI-assisted scientific review pilot: an aggressive agency-wide AI rollout. Here’s the FDA commissioner, Martin Makary.

The FDA is rolling out a generative AI tool across all centers by June 30th to speed up scientific review for new therapies.

One of the incredibly crazy hardships of our world is the problem with the development of new drugs. It’s the regulatory slowdown—the fact that we sometimes take years to make it happen. An FDA employee who approves a drug that then kills some patients has disastrous results. But what we don’t do is value all the lives lost from not approving a drug, right? It’s just a flip situation.

This is the safety-versus-efficacy problem.

Yeah, which is structural here. But I think the fact that they’re using AI for a bunch of stuff is amazing. I think it’s great. I think it’s the beginning. We should be rolling this out across all agencies, and we’ve discussed this before, right? AI can help streamline all agencies and all approval processes.

It’s going to be crazy. In the future, people are going to look back and say, “Really? Humans did this? Seriously? How long did it take? Decades?”

I’m looking for the administration to make real structural change that lasts beyond just 1 administration.

I think one of the greatest assets we have as a country is 50 different states that can try 50 different ideas, so that if you’re stuck somewhere, you can try somewhere else. We’re way underutilizing that core strength that we have.

One of the best things it can do is say, “Look, let’s deregulate like crazy and push a lot of this into 50 different shots on goal. Even if we make mistakes, we’ll be far better off as a country if we have chances for entrepreneurs to move forward in many different places.”

Yeah, you’re spot-on there, Dave. Here’s an example. For the last decade, the FDA under previous administrations was shutting down all stem cell clinics in the U.S. So what happened was, if you wanted access to stem cells or a sort of advanced experimental therapy, you would leave the U.S. You’d go to the Caribbean, Mexico, Costa Rica, or Panama for these treatments.

The problem is that you’re in a location that doesn’t have advanced medical capabilities if something goes wrong. This year—in fact, this month—we’re about to see new regulation in Florida specifically. I know this because Fountain Life has 2 centers, soon 3 centers, in Florida.

Florida’s about to pass legislation allowing stem cell treatments to occur in-state. So you’re going to finally have the data, and people will be able to stay there, be near their physicians, take the treatments, collect the data accurately, and then learn.

All right, let’s move on to our next subject. This one’s on CoreWeave. Dave, lead us here. Read the article and let’s talk about it.

We talked about this being a bellwether for the whole IPO market opening up again. I think CoreWeave is above $70 now, which is great. It’s up 40–50% from the IPO price, so far, the investors are all happy. I’m sure the investment banks all over Wall Street are saying, “Hey, we’re open for business.”

What else? Let’s go. For those who don’t know, talk to us about what CoreWeave does.

CoreWeave is a bellwether for AI. It’s a data center company. The thing that makes it a little weird to people is that they borrowed a huge amount of money. They have an enormous amount of debt, which they used to buy NVIDIA chips. One of their investors is NVIDIA. They financed the data centers with the value of the chips.

The reason that’s controversial is because nobody knows how quickly the chips will depreciate. I think the current logic is that we’re going to be compute-bound for years. Everything that can compute is going to be in incredible demand, in my opinion, forever hereafter for humanity.

We’ll look at this early wave of computing from 1970 to 2020 as kind of the test ground, where sometimes we use the chips and a lot of the time they’re lying all over our office here, not processing anything. Starting around now and forever hereafter, they’re all going to be doing AI all the time.

Anything that can compute is going to be in huge demand, and so I think CoreWeave will go up in value if that theory is true.

Computronium. For those who haven’t read enough science fiction or aren’t science-fiction geeks, computronium is what AI eventually turns all matter in the universe into, so it can compute all the time, everywhere.

The article here from CNBC says, “CoreWeave beats on revenue, reports more than 400% growth in the first earnings after the IPO.” We’ve been waiting for some red-hot IPOs to move the financial markets, really to break open the logjam, because we haven’t had many companies going public or even that many acquisitions.

For the venture business, Dave, as the head of our partnership with Link Ventures, this is amazing news.

It is amazing news, and the companies that are in the pipeline behind this are much less risky. I was worried about this being the bellwether because it should go up, but with all that debt, if it goes bad, it goes really bad, and that would shut down the markets again.

I’m really glad to see it went well. The ones that are in the pipeline are almost all either chip companies or vertical-use-case companies. They’re so low-risk, with such high returns. This is good.

Once the pipeline fills up and we have 5, 6, or 7 more IPOs, I think the train will be rolling.

I hope so. All right, here’s another one from Bloomberg. Next slide here: “Microsoft layoffs hit coders hardest, with AI costs on the rise.” Engineers make up 40% of the total layoffs as AI begins to take over parts of their work.

Not a surprise. Dave, how do you think about this?

Well, this is right in my wheelhouse. We’ve all been telling people for a long time now that almost any white-collar coding or other job is not going to exist in 2030. There’s a line between today and 2030, and very few people are doing enough planning to get ready for it.

Just to be humane to your employee base, you should have some kind of plan in place. I’ve had some really frustrating conversations this week with a couple of our companies because they’re all using Blitzy. Blitzy is supposed to be an opportunity to learn what long-horizon AI can do—an AI that writes 3 million lines of code in a single night. That is incredibly world-changing.

And then the reports I got back are, “Hey, it had some bugs.” I’m like, “Oh my God, you completely missed the point. Look at the rate of improvement. Look what it’s going to do by the end of the year. Can you slow it down?” I love Blitzy. It’s one of our investments, full disclosure, in Link Ventures.

But if you wanted to replicate another company, right? One of the biggest challenges right now is the ability of AI to basically go in and replicate a SaaS business, to just stand up competition. Where would you bring Blitzy in to do something like that?

Blitzy, in its current form, is useful if you have any legacy codebase—5 million, 10 million, 20 million lines of code—and you say, “Hey, I want you to rebuild this entire product, make it faster and better, and run it on the cloud.” That’s kind of their sweet spot right now.

But that’s not the real point. The real point is, if you look 3 months, 6 months, or 9 months into the future at the rate of improvement of the AI—not just to write code, but to build entire plans, like project plans—it’s on this crazy, crazy upslope. If you look just a few months into the future, Blitzy can come in and say, “Well, look, here’s a greenfield opportunity. I’ve wanted to write a piece of software that does something that’s been burning a hole in my pocket for years, but it was too expensive. Can I do it now?”

Yeah. Not only can you do it, you can do it in a night or a week. A lot of people right now use Salesforce.com or Outlook to do things that they’ve bastardized to their actual use case. It sort of works, but it’s not exactly what they wanted. It’s good enough.

Now you can just vibe-code, or Blitzy-code, your way to a perfect solution for exactly what you were really after originally. That creates this whole long tail of really, really good, high-quality software that does what you actually wanted.

You know your dream, Peter, of, “I get on stage, I’ve got all this IT around me, I’ve got avatars I’m talking to. Why am I talking to an IT person? Why can’t I just talk to the AI, all orchestrated?” That is imminent. That’s a good example of the kind of stuff you can suddenly build because the cost per line of code is down 1,000 or 1,000,000×. That’s what Blitzy is enabling.

I just want to walk into a business meeting where I’ve got a presentation on my computer and have all the AV work. That’s all I want. Our favorite joke there drops in, right? AI is easy, but AV is really hard, courtesy of Brad Templeton.

Something just struck me, Dave, as you’re talking about this. This means we can finally retire all those old COBOL systems.

Yeah. The world runs on legacy COBOL that nobody wants to touch and nobody knows how to maintain. We can essentially rewrite those for the modern world.

The CEO of State Street Bank was telling me that all the NAV accounting for all the mutual funds in the country is done on PL/I code on mainframes. Why? It’s like $300 million a year of maintenance for that tech stack.

Why? Because it’s $1 billion to rewrite it, and I don’t want to eat that cost in any given quarter. It’ll kill my bonus.

Wow.

Well, you bring that $1 billion down by a factor of 10 or 100, and it’s like, “Oh, now I’ll do it easily because it fits within my bonus cycle.” It’s just that simple.

I’ll just end on this particular article from Bloomberg on AI. I used to harangue my kids: “You’ve got to learn to code, got to like to code, got to learn to code.” Now it’s, “You need to learn to vibe-code.” You need to learn to think in a way that’s concrete, so you can describe it to your AI. Everybody’s going to be a coder. Everyone’s going to have this ability, and it’s pretty extraordinary.

Yeah. I’m so glad you said that because I’m working on a memo right now that I’m going to send out to all the CEOs that says, “Look, guys, be humane here. Some subset of your people are going to be masters of AI, and another subset are going to get crushed. Come up with some plan for the people who are going to get crushed so that it’s humane.”

For the people who are going to be masters of AI, free up half or two-thirds of their time so they can start the 10,000 hours of learning that they need to do to get on top of this wave. What you’re doing right now by turning a blind eye to it is dooming all of them by not giving them the time to become masters of AI and learn to use things like Blitzy and Cursor.

Yes, let them play. Let them play. Let them play.

Yeah, but you explicitly allow that. It’s not going to happen naturally. You’ve got to have a plan.

One of our clients had to lay off about a third of its call center because it automated it all. They gave the employees an internal 1-year UBI and said, “Go do whatever you want for a year. Here’s a chunk of money. Play.” The employees absolutely loved it, and something like 95% got a job very quickly.

I would say, “Here’s a year of UBI. Use it to come up with new businesses and new business ideas that our company could utilize.” That would have been equally good.

The other point you made is really, really important because if anyone does choose to take our advice, telling them to go play with Cursor and do some vibe-coding is not good enough. You have to get them into that new mindset related to creativity that allows you to use AI to do something productive that goes way beyond just vibe-coding.

We haven’t fully specified that skill set, but we know it involves creativity. We know it involves vision. We know it involves knowing how to use the tools, and what the tools do well and don’t do well. But it’s way beyond just saying, “Hey, guys, play with Cursor a little bit. Okay, now you’re an AI person.”

So take it a step further, buddy. What would you recommend?

I would say that the 3 of us specify what we’re learning about that skill set—who’s succeeding, who’s failing—and then codify it into actual programs. I suggest you use this product, do this training program, and try to build this thing. Put it out on a website or put it out in this podcast as a road map that people can follow. The people who choose to follow it end up being masters of AI.

It’ll evolve over time, but it’s not as hard as people think. It’s intimidating to everybody. No one’s trying, even though it’s not that hard, actually.

You know what? If we had a dedicated podcast episode just to that, I’d love that.

We’ll record one session. Peter will talk. Let’s talk to Nick and the team about this.

Sure. But we should have one dedicated session running through it and saying, “Here’s how you build your own future and envision it. Here are some tools, and let’s run through some sample use cases.”

Give it a brand, like a badge of honor. People love these alternatives to university badges of achievement. We can give it one of those.

This reminds me of when electricity was first entering the manufacturing and factory ecosystems. People were hesitant to bring in electrical systems. They would finally see an electrical motor running some part of a factory, learn from that, and then put it into a different part of the factory. Very slow. Very slow.

Can I give a quick anecdote here?

Of course.

The worst example of this was in the 1600s, when a Scottish man discovered that the definitive cure for scurvy, which was a vitamin C deficiency disease, was eating lemons or oranges. You lost your teeth. It was horrible.

He figured that out and told the British Navy. The British Navy refused to believe him for 50 years—that such a simple thing could do this. They then took 50 years to deploy it throughout the Navy. They took 50 more years to say, “Oh, maybe we should tell the merchant navy.” It took the merchant navy another 50 years.

Between the known cure and full deployment was 264 years. Forty percent of sailors—thousands of sailors—were dying every week. It’s a testament to how bad we humans are at adapting to new things.

Luckily, that’s shrunk a lot. Most of that is down, I think, to 17 years, with most of that being longitudinal testing from a known cure to a disease treatment. But I think these are the cultural issues we have in adapting to things.

My biggest insight about human beings ever is that they’d much rather be comfortable than happy.

Wow. Sad.

I have a student intern at Blitzy this summer, so I’ll lean on him. I’ll learn a lot from that, but I’ll lean on him to come up with this: Okay, he’s in high school, coming into college, so what’s the perfect roadmap for staying ahead of this curve from your perspective? We can weave that into our plan.

Fantastic. Let’s do it.

Everybody, I want to take a short break from our episode to talk about a company that's very important to me and could actually save your life or the life of someone that you love. The company is called Fountain Life, and it's a company I started years ago with Tony Robbins and a group of very talented physicians.

Most of us don't actually know what's going on inside our body. We're all optimists until that day when you have a pain in your side. You go to the physician or the emergency room and they say, I'm sorry to tell you this, but you have this stage three or four going on. It didn't start that morning. It probably was a problem that's been going on for some time. But because we never look, we don't find out.

So what we built at Fountain Life was the world's most advanced diagnostic centers. We have four across the U.S. today and we're building 20 around the world. These centers give you a full-body MRI, a brain, a brain vascular, an AI-enabled coronary CT looking for soft plaque, a DEXA scan, a Grail blood cancer test, and a full executive blood workup. It's the most advanced workup you'll ever receive, 150 gigabytes of data that then go to our AIs and our physicians to find any disease at the very beginning when it's solvable. You're going to find out eventually. You might as well find out when you can take action.

Fountain Life also has an entire side of therapeutics. We look around the world for the most advanced therapeutics that can add 10 or 20 healthy years to your life, and we provide them to you at our centers. So if this is of interest to you, please go and check it out. Go to fountainlife.com/peter. When Tony and I wrote our New York Times bestseller, Life Force, we had 30,000 people reach out to us for Fountain Life memberships. If you go to fountainlife.com/per, we'll put you to the top of the list. It's something that is, for me, one of the most important things I offer my entire family, the CEOs of my companies, and my friends. It's a chance to really add decades onto our healthy lifespans. Go to fountainlife.com/pater. It's one of the most important things I can offer to you as one of my listeners. All right, let's go back to our episode.

Our next story here is from NPR, which I guess still exists for the moment. I won’t go into that, but the story is “Family shows AI video of slain victim as an impact statement, possible legal first.”

So, what does this mean? AI was used to create a video of a victim speaking at his perpetrator’s sentencing. And it says “slain victim,” so I guess the victim is dead. For the first time, AI was used in a U.S. courtroom for a victim impact statement. Legal experts see this use of AI as powerful and ethically acceptable.

Fascinating. I’m just fascinated to see how AI is going to play into different parts of our ecosystems, our regulatory and social worlds. Any comments on this?

I’m surprised they allowed it, also. That’s really forward-thinking to allow that. But I guess it’s the voice of the victim, and if you can simulate it to give some semblance of reality, why not?

Well, the “why not” is because the defense will then say, “Well, if you’re using AI testimony, I’m going to produce AI.” I mean, it’s going to be like, “Wait, this is deepfake versus deepfake now.” So, it’s a very slippery slope.

I assume in this case the judge and everybody wanted the outcome to be—you know, it must have been pretty obvious, but they were worried the jury wouldn’t see it that way. So, having the victim talk on their own behalf is going to get you the outcome that they want. But again, that opens Pandora’s box like crazy.

All right. Dave, I would love if you would cover this one: “Specialized AI systems deliver breakthrough results: domain-adaptive pre-training.” What is this all about?

Well, I’ll preface this by saying one of the listeners last time posted, “I love this podcast. It’s amazing. It’s just like the All-In podcast for geeks.” That’s the highest compliment I think I’ve ever gotten. So, I apologize. We’re dealing with the real news that matters, right? This is the stuff that’s impacting our lives for the long run.

This is called domain-adaptive pre-training. The way this evolved is, I was actually, 2 weekends ago, trying to design a circuit—a neural network circuit, as usual—and I said, “I wonder if Gemini is any good at designing circuits?” It was terrible.

So, I talked to Alex Wissner-Gross, the physicist, our mutual physics friend, and he said, “Hey, have you heard of this thing from NVIDIA? They took an open-source model—in this case, Llama 2, a couple of years ago—and they retrained it to be a circuit designer.”

I pulled up the paper, and I was like, “Wow, let’s try this thing.” Apparently, it’s an order of magnitude better at circuit design than the raw model is because it’s been retrained to be a circuit designer on specific circuits.

The reason I wanted to bring it up is because, in this particular case, the cost of training it to do that specific task dramatically better was only a 1% increase in the overall training cost. This matters a lot to upcoming startups because the new Llama 4 model—the 2 trillion-parameter model—is a massive, massive investment by Meta that is free and open source.

Here it is, and the tools to then retrain it to do things like drug discovery, self-driving, chip design, or speak some obscure language—those tools are open source online. There’s a whole cookbook in this paper on how you can do it yourself.

A lot of the companies that I look at, a lot of the teams, even the high school teams, are afraid to retrain. They’re just putting wrappers around the core models. They’re using the models all day. They’re talking to either GPT or Gemini all day long, but it never occurs to them to go and retrain it and make it an order of magnitude better at some specific task.

Oh, sure, go ahead. So, let me get this straight: you take one of the generalized models—in this case, Llama—and then you give it a training set for a very specific niche use case. At a very, very low cost, you get an unbelievable outcome.

That’s exactly right. I skipped one step in there: first, you distill it, make it smaller and much faster, and don’t lose any of its capability within your domain, but throw away a lot of the irrelevant capabilities. It’s been trained on 15 trillion tokens—every language, every conceivable thing—and you’re just trying to use it to either do math, do chip design, or speak Swahili.

There’s a lot of overhead that you can strip out. The distillation process can shrink it by a factor of 10 to 100, make it much faster, and then you can build it back up again with new data that’s specific to that domain.

So, it’s called domain-adaptive pre-training. It makes total sense given what you’re doing, and very, very few people are trying to do it even though it’s so straightforward. It’s just a cookbook laid out for you by Nvidia: here’s how you do it.

I did notice that after Nvidia did this, they went completely dark and internal on it, which tells you how well it works. They’re using it internally to design the new chips, but they say, “Holy crap, this is so good. We can’t just throw this out there anymore.”

Wow. You know, we just don’t know how many of these are out there—these hacks, these improvements. We’re going to see one in a moment. Let’s go there.

In fact, this is an article that I read. I just thought it was extraordinary. It’s news from DeepMind: “Google DeepMind introduces AlphaEvolve, a coding agent for algorithm discovery.”

So, here’s what we learned from DeepMind, an Alphabet company. AlphaEvolve uses Gemini AI and evaluators to evolve high-performance code. In doing this, Google DeepMind was able to improve Gemini’s training speed by 23%, which is massive, and improve GPU performance by 32.5%.

So, we’re talking about new models that allow an AI to self-improve—to improve its own algorithms. We’re seeing really the knee of the curve in this self-referential, self-improving AI future that really drives us, exponentially—super-exponentially—toward digital superintelligence. I mean, it’s all happening right here, right now.

Yeah. There’s a cool video of Sam Altman over at Sequoia’s office this last week. One of the questions he got from the crowd was, “Hey, is coding just another capability within all the OpenAI capabilities that you’re working on, or is it somehow special and isolated, and you work on it more?”

He said, “Oh, no, it’s definitely special. It’s very, very special.” It’s special because we use it internally. Then he bites his tongue and says, “Let’s move on to the next question.”

I asked Alex Wissner-Gross about that, and he was like, “Yeah, that’s the 3rd time I’ve seen that.” The problem with talking about self-improvement inside these big foundation-model companies is it attracts all the dystopians, and then all the regulators come right after that.

But it’s really obvious that the reason they care so much about this use case is because the thing is going to improve itself, and that’s going to accelerate the 32.5% GPU performance. This is just generation 1 of this thing, right?

Yeah, that’s unbelievable. It’s not just—I mean, when I say it’s super-exponential, it really is. We’ve gone from 2x every 2 years to 10x a year to 100x a year to something beyond that in terms of the improvements that we’re seeing.

It’s not slowing down. I mean, there are no physical limits right now that we can see. Here’s another one from Meta: “Meta’s FAIR shares new research to achieve advanced machine intelligence.”

A number of things were announced from Meta this week: Open Molecules 2025. It’s a massive data set enabling breakthroughs in molecular discovery with large atomic-system simulations.

So I love this stuff. This is where we can discover new materials, and we can discover new drugs and drug interactions. This is humans designing with atomic Lego sets. And then, A Universal Model for Atoms—another model out of Meta—is a foundational machine-learning model for accurately predicting atomic interactions across diverse materials.

Again, Salim, you and I have discussed this: materials science is at the fundamental core of everything. Materials science is going to drive energy, environmental breakthroughs, and space exploration. It's going to drive all of these things. There's this amazing thing of going from atoms to bits, right? And then, once you can use bits to affect atoms, you end up with this incredible feedback loop. If we can actually manipulate the physical world with these advanced AI models, then holy moly.

Yeah, it's crazy. I don't know. Did I tell you, Peter? Alexander Keesling, our much younger frat brother, is now one of the top quantum computing guys in the country. Quantum computing is good for 2 things that we know of. One of them is cracking cryptography.

Yes.

But the other one is simulating biological systems.

The hell were you guys doing in this frat? What in God's name? What was in the water?

Actually, MIT has just become a ridiculous place since we were there. The concentration of entrepreneurs is just unbelievable today, so I'm not surprised. But I think we should follow up on how that interacts with this, because the AI gets faster and faster and faster, but it needs that simulator in order to keep it on track. If you can take the whole drug-discovery thing or the whole molecule-discovery thing, and the simulator is actually quantum-computer-based, well, this is what Jack Hidary is doing at Sandbox AQ, right?

He's basically using superfast AI models to run quantum equations and being able to model things in a way that is understandable on a quantum basis, but doing it with conventional AI. I don't think we've seen anything yet coming out of quantum compared to what we're going to see over the next few years.

I don't know if you've ever watched 3Blue1Brown—unbelievable videos on all things science, math, and technical. It's just the most incredible explanation of everything. The transformer AI explanation is epically great.

Would you text me that? I would love to watch it with my boys. That would be great.

Yeah, for sure. The quantum-computing explanation couldn't do it. You just cannot fathom quantum computing—what it does well and what it doesn't do well—in an hour or 2.

Do you want to share our favorite conversation about quantum computing?

Sure. Ten years ago, Steve Jurvetson spoke at Singularity University, and we asked him in the Q&A, "The amount of computation that these things can do is off the hook. Where is all this computation coming from?"

He goes, "Well, we have a consensus answer amongst the top quantum physicists in the world, but you're not going to like it."

We're like, "What? Just tell us."

And he's like, "Okay." He goes, "It seems the current consensus is that we're doing the computation in parallel universes and bringing the answer back."

At which point everybody's brains exploded, and we're like, "Okay."

He goes, "I told you you wouldn't like it." So that was about 10 years ago. By the way, that's true. It's still the consensus.

Well, this is the point. Last year, we had Hartmut Neven, the head of Google's quantum-computing lab, and I said to him in the Q&A, "Steve Jurvetson said this 10 years ago. Now we have 10 years more data and experiments, and thousands of teams have been working on this. What's the consensus?"

And he goes, "You're not going to like it, but it's the same." In fact, he said the existence of a quantum computer may be evidence that we're living in a multiverse, in which case everyone just goes, "Okay, come on."

No, before that scares everyone away, we really ought to get Keesling on the pod for a little bit. The reason there's lots and lots of hope and enthusiasm around this is because we're very close to an AI-quantum compiler, where you can just say, just like you do with Gemini, "Hey, here's something I want to do. Does it fit quantum or not?" It will actually tell you and then compile what you were saying into the actual quantum code.

The usability for a startup, or for just a weekend hacker, is really exciting. Also, the quantum computers are on the cloud. It's not like you need to go anywhere; it's just another login on the cloud. So it's really practical and exciting, and coming soon.

You know, I'm always looking for tipping points: when do we know things are hitting an inflection point and accelerating, and things are changing? This article, I think, is one of them. The New York Times headline is, "Top Priority: Pope Leo Warns the World of the AI Threat."

Pope Leo XIV warned that AI poses a serious risk to human dignity, justice, and labor, and considers that AI must be developed "for the good of all" and calls for ethical use. It's not just his commentary; it's the fact that this is a New York Times article, and that the Pope is focusing on AI in the first month of his papal leadership, if that's the right terminology. That's pretty extraordinary.

Yeah, I think it's great, because there's a billion Catholics all over the world, including my wife, and the underreaction is rampant. When you get somebody as renowned as the Pope saying, "Okay, here are the priorities," it's very similar—on a much smaller scale—to Joe Aoun, the president of Northeastern, who said in his commencement address last year, "AI will affect every one of you, no matter what you majored in. You must do AI." He's the only university president I've heard say that, and it's like, wow, thank God.

Now you've got the Pope giving you the heads-up that this is coming, and hopefully that will create some motion. It's great to see them getting in front of this, as opposed to waiting 300 years to acknowledge that the Earth goes around the Sun. This is at least much more up to speed.

I've got to admit, the couple of times I've been to the Vatican over the last few years—Peter, you've been there a couple of times as well—they've been way, way more up to speed with what's happening in technology than I ever would have given them credit for.

Yeah, we had sessions on regenerative medicine and longevity there, and I was pretty impressed that those topics were allowed. But I think it's going to be interesting if an AI claims to be conscious, right? They would have a reaction to that.

Yeah, maybe they'll let women actually become priests someday, and then AIs will follow after that. We'll see.

Well, just another note on this: the last Pope had this defining quality, which was tenderness—one of the most extraordinary, humble, wonderful men you could ever meet. Usually, when they vote in a new Pope, they go the other way, and it's amazing to see them continuing the same path in this new post. So it's great to see.

By the way, one of the most extraordinary things you can do with large language models is take very dense and complicated information, like the Bible, and make it understandable to you. The Bible is in every large language model out there in detail. Talk about Bible study—I think there's no better Bible study than a conversation with your favorite AI.

Can I give you the retro version of this?

Sure.

In the 1500s, there was a Swedish monk called Swedenborg. He was a psychic, and what he did was basically decide that the Bible was a set of coded rules, and you had to apply these filters. One of the rules that he applied was, "Any verse that has water in it is truth. Any verse that has fire in it is false, and you should just chuck it out."

He basically rewrote the Bible based on the rules that they defined. When you read the resulting text, it's unbelievably beautiful. So this is a new version of that, I think, applied when we say AI can help us interpret some of this stuff.

Amazing. All right, let's turn to other exponential news beyond AI. This is a story that I love. This came out from Archer Aviation. I've had them on my stage at the Abundance Summit. Archer is the new official air-taxi provider for the LA28 Olympics.

Here is the tweet: "We're extremely proud to announce that Archer is the official air-taxi provider for the LA28 Olympics and Paralympic Games and Team USA."

I've been saying this for a bit now: that is the target deadline. Archer is beginning operations and has built manufacturing plants. We also have Joby Aviation and BETA in the United States. These are sort of the flying cars. I've had this conversation: they're officially called eVTOLs, electric vertical takeoff and landing. That name just rolls off the tongue and hits the floor.

I think flying taxis or flying cars is better. The 2028 Olympics is when manufacturing operations and FAA regulations all need to be in place. I fly my airplane out of Santa Monica Airport here, where the runway has been shortened from 5,000 feet to 3,000 feet. Eventually, it's supposed to be closed down in 2028.

Hopefully, what will happen is it will become a vertiport for these to take off and land. So, if I want to go to Malibu, a Dodgers game in downtown LA, or Santa Barbara, I can hop in and avoid the 405, which is like hell on Earth. If there was one breakthrough I've been waiting for forever and a day, it's this one. Because this will open up and enable so many things.

Just consider how many amazing little plots of land are high up on the side of a mountain that you can't get to, right? Now they become accessible because of huge geographic arbitrage. I've talked about going to buy island real estate off the coast that doesn't have easy access but is beautiful, and just set up eVTOL service. A lot of islands out here on the East Coast.

Peter, say again.

A lot of really nice islands out here. Actually, I look at a bunch of them every day.

All right. Well, let's do it.

So, let's do it.

Yeah. I think the barrier is safety, obviously, but each of those electric motors is completely independent, and so the power safety is going to be pretty doable. And the air traffic—it's all AI-controlled.

Yeah, I think it's more infrastructure because you need specialized gates and specialized facilities to deal with this, all that stuff. But the safety is definitely—I think I'd feel safe even landing in New York with one of these right now.

I think my real point there is there are absolutely no barriers to this, and it's all regulatory. So having this forcing-function deadline is going to put the manpower on it. People are working on it. It's going to happen.

Yeah. It's really exciting, actually, to have a deadline.

It is, and I mean, it's the number-one driver, I think, and I'm super excited to hear this announcement. I spent 2 hours yesterday trying to get home from Kennedy Airport and 2 hours last week trying to get to São Paulo Airport. So, God help us get this thing going faster. The alternative is helicopters, which are damn dangerous, right?

Yeah, they truly are.

All right, let's move on. The next subject here is longevity, one of my favorite subjects. I want to hit on a couple of articles and subjects here.

Dave, you were there. You were someplace on Earth, even though you're a director of the XPRIZE Foundation. The big news here is that, in November 2023 in Riyadh, once again, Saudi Arabia was putting its money behind advancements. We announced the $101 million Healthspan XPRIZE. This is an XPRIZE directed at trying to add a minimum of 10 years, with a goal of 20 years, of additional health in immune function, cognition, and muscle.

For the first time, what we're doing is providing milestone awards to help teams make progress. We've had 623 teams register for this XPRIZE. I'm super proud of that. Last week, Dave, you and I were sitting together, and we awarded 40 of the teams $250,000 each, totaling $10 million, in this Milestone 1 award ceremony. We had pretty amazing teams, with winners from 12 to 14 countries, from Australia to China, throughout Europe, the U.S., and much of Asia.

What do you think of it, Dave?

Well, right out of the gate, it's amazing how the original XPRIZE that you conceived of was trivial to measure compared to what you're doing now. The impact of this is so incredibly massive.

In terms of the XPRIZE's overall mission, it's just the best idea ever: to try and make it a measurable, manageable prize. It must have been incredibly hard. So congratulations on that. This is clearly working.

Thank you. I have to pass the congratulations on to my partner in crime, Dr. Jamie Justice, who is the executive director of this competition. Yanni Pallis helped me find her, and she's been running it, working with an incredible scientific advisory board.

Mehmood Khan, the CEO of Hevolution—which, think of it as health and evolution—I'm invested in a whole bunch of different technologies here, and I advise different companies, so I am conflict incarnate. But, to be clear, I'm not involved in any team selections or team judging. I'm a bystander, watching and rooting for a winner sooner rather than later.

I think it's awesome. This is the biggest public prize ever given in the history of humanity.

Yeah, it is.

Compared to Elon, we just awarded Elon's $100 million prize, and this one is a $101 million prize specifically. So it would be bigger than Elon's XPRIZE.

Yeah. The other thing that jumped out at me on Monday, down in New York, is the original XPRIZE: a $10 million prize and, I think, maybe $100 million of effort to win the $10 million.

Yes. It was a 10x return.

So then the carbon-removal prize was kind of the opposite of that: a huge prize from Elon Musk's coffers, and somebody actually made a lot of money on it. This one is back to the original mission, which is as a force multiplier: a $101 million purse. The number onstage, with some massive amount of effort having gone into winning these stages—I think it's 30x overall leverage that the prizes are delivering now, right?

Yeah, 30x. We've generated—we've launched—about $550 million of prize purses, and we're close to $10 billion in R&D spent cumulatively by all the teams over the last 30 years to do this.

It's a great idea. If you go to XPRIZE.org, folks can see all the XPRIZE competitions. One of my favorite ones going on right now is the Wildfire XPRIZE. I did a Moonshots podcast with Palmer Luckey, which was super fun. Palmer was the first person to register for this wildfire prize. I've seen his technology going after it, and it's amazing. I'm super excited for that.

All right, let's go to a related article on longevity. A lot of people are concerned that if we extend the human healthspan—in the United States, average life expectancy, how long your heart is ticking, is about 79 years, which is the average age of death. Average health expectancy, how long you're healthy, is 63. So there's this 16-year gap between when you stop being healthy and when you die, and it's just miserable for a multitude of reasons.

People say, “Well, if you extend the healthspan and lifespan to 100 or 120, aren't we going to have an overpopulation issue?” Well, it's just the opposite. I used this article to bring this conversation to us: “Turkey declares 2025 as the Year of the Family to curb declining birth rates.”

This article just came out. In 2024, the birth rate in Turkey was 1.48 children per woman. You need 2.1 children to replace your existing population. But Turkey is doing pretty well compared to other countries. So I used my favorite LLMs to come up with the 10 countries with the lowest birth rates, and they're not insignificant countries.

Check this out: South Korea has 0.72 children per woman. South Korea is sublimating. It's vaporizing. It's going away. Hong Kong, 0.8; Palau, 0.88; Singapore, 0.9; Puerto Rico, 0.9; Taiwan, 1.1; China, 1.16; Spain, 1.19; Italy, 1.2; Ukraine, 1.22 children per family. It's a problem.

Elon has been very vocal about this, and I have as well. The biggest issue is not overpopulation, right? The old population-bomb theory isn't playing out. As we create increased abundance, the opposite is true: it's an underpopulation issue.

The South Korea one is nuts. It's saying that the birth rate is three times less than it needs to be to maintain its population.

Yeah, it's really ridiculous. So what does a country do? It starts offering massive benefits and incentives for having children.

Dave, any thoughts on this?

Yeah, I've read up on the incentives, and they're not even close to covering the cost of raising a child. No one is reacting to the incentives.

I also haven't modeled out how the declining birth rate and increasing longevity interact. Is there a dip and then a climb, or is it smooth? It would be worth trying to map that out. It's hard to predict the longevity curve, but it would be worth trying.

One thing I can say is that there's a perception in the U.S. that China is trying to beat us in AI and take over the world. But what's actually driving China's incredibly fast push into AI and robotics is a ridiculous demographic time bomb that hit Japan. They have the exact same problem, times much worse, coming to China. It's survival—fundamental survival.

By the way, the other side of the equation is India, right? With 1.41 billion people, the majority of whom are at the absolute bottom of the pyramid. They need AI and robotics for education and health care. They can't survive as a nation without this.

Are you seeing that, Salim?

I have to go a bit tongue-in-cheek on this. I was in India, and I introduced 2 random people to each other, and they knew 5 people in common. Then I introduced 2 other random people to each other, and they knew 10 people in common. I'm convinced that India has, like, 50,000 people in CGI to simulate 1.5 billion. It's like everybody knows everybody. It's insane.

I've not lived in Bombay since I was 10 years old, and I can go to a party and know 5 people. It's ridiculous.

Anyway, the broader point is dead-on: the opportunity for robotics for elderly care, education, and health care in general. There's so little of it.

It's like water in a desert. Any little thing is going to totally change the landscape, and I'm super excited by the potential there. It's incredible. It's a good segue into Elon updates.

One of the things I'm excited about in the direction things are going is that globalization is kind of falling apart in a big way, actually. Globalization was almost entirely driven by cheap labor. Robotics doesn't happen naturally in the U.S. because you can always do something very cheaply with Philippine or Indian labor.

But if that becomes less true, then we have to invest in robotics, which, in the long run, is a much smarter direction for the world to go. That should kickstart investment in the category because the technology—the vision systems, the tactile systems, all of that—is solved now, and it wasn't just a couple of years ago.

Let's put the numbers on the table here. The minimum wage in California today is $20 an hour. I was just with Cyrus Sigari. We do a monthly exponential mastermind for the Abundance members, and Cyrus gave one yesterday on humanoid robotics, flying cars, and spaceships.

We were talking about this: The projected price per hour for operating a humanoid robot is going to be $1 an hour. At a base cost, it's 40 cents an hour, but with maintenance, electricity, everything else, and insurance, it's a buck an hour.

This arbitrage between a $1-an-hour, GPT-5-level, Grok-4-level humanoid robot that operates 24/7 and a teenager from California—there's just no comparison. It ain't going to happen until the robots become conscious and start demanding equal rights and equal pay. If we have robots doing all of our drudgery work, they're not going to be happy about that.

We actually added a special section this week on Elon company updates because there are a lot of them, and we'll wrap up with this segment today. I think it's important: Elon is still our moonshot mate, just really driving extraordinary progress. How he does this while doing everything else, right? He's in DOGE and in the news all the time, but he's still operating 4 other crazy moonshot companies—5 if you want to. Five, right?

Let's talk about 3 of them here. The first is Elon in Saudi Arabia. He was by Trump's side the entire time. Let's listen to a conversation about Elon announcing robotaxis and Starlink in Saudi Arabia.

My prediction, actually, for humanoid robots is that ultimately there will be tens of billions. Potentially, we could have an economy 10 times the size of the current global economy where no one wants for anything. Sometimes in AI they talk about universal basic income. I think it's actually going to be universal high income, where anyone can have any goods or services that they want.

It would be very exciting to have autonomous vehicles here in the kingdom. I'd also like to thank the kingdom for approving Starlink for maritime and aviation use.

So there he is with Minister Abdullah Alswaha, who has been a key point of contact making announcements on robotics. I'm going to be amazed if we don't see Optimus manufactured in Saudi Arabia as well.

Here's the next article: SpaceX deploys 28 additional satellites. There are now 7,135 Starlink satellites in orbit. The FCC has approved 12,000 of Gen 1 and 7,500 of Gen 2, but SpaceX has proposed expanding the constellation to 42,000 satellites.

That's either 42 because of Douglas Adams, or it's 420 because Elon likes 420. I don't know.

What do you guys think about Elon's companies in Saudi?

Well, I mean, Elon—like you said—his ability to manage 5 concurrent mega-companies plus DOGE is a case study and a role model for how you use, 1, AI assistance, and 2, integrators: people who are incredibly like-minded, finish your own sentences, but are entirely focused on operations. It's very similar to Google in the early days.

I try to encourage everyone I bump into to study Elon very, very closely, at the individual-action level. You take the part you want to take, but replicate a lot of it. It clearly is a path to success.

The part I love is that every week he has 5 companies, so he goes 1 day a week to each one and says, "What problem do you need me to help solve?" You're solving a problem a week—50 problems a year per company. Amazing.

The other thing—which is where you wanted to go exactly—is that there's never been a point in history where the tech titans have hung out with the politicians and world leaders. This is a lifetime world first. I don't know if anyone's really appreciated the magnitude of that.

I'd love your guys' opinion on Tesla. Tesla's taken a massive beating because of Elon in the government, in DOGE in particular. I tweeted a couple of weeks ago that the value of Tesla is not in the car sales; it's in the robotaxi and in Optimus, which are trillion-dollar opportunities. How do you guys feel about Tesla stock?

To me, this one's really obvious. We all lived through the Apple cycle, where Apple was the greatest company and enabled us like crazy. Then Apple products didn't work at all, and it was torture. Our actual daily lives turned into this hell of needing to migrate to Sun or to a PC, and then Steve Jobs came back and magically life was great again.

It's a very real thing, but you just have to track the talent. If Steve Jobs comes back, it's not Steve alone. He brings this incredible wave of talent and inspiration.

Yeah, yeah, yeah. So if Elon can attract that talent to Tesla still, it'll become a great robotics company. It's cheap. If you look at LinkedIn and you see that Elon is so distracted that he's not working on the recruiting, then it's a car company.

You mean, then it's a car company.

I've lived through the early years of Tesla, when he was about to run out of money every quarter. It was hard to put money into it, and lots of us didn't put money into it. Earlier, when SpaceX's rockets were crashing, you didn't put money into it. In the early years of Starlink, you're like, "How is this going to work?"

The lesson that people have learned over the years is: Don't ever bet against Elon. His manic focus, his engineering ability, his ability to distill key problems and just go after them and solve them—the cluster approach they use at Grok to solve things—is legendary stuff that's going to go down for centuries in terms of somebody making massive turning points.

I think the opportunity for Tesla to totally turn it around with robotaxis is huge. Also note that all these million cars are carrying around energy. That is a huge marginal impact on things. There's the fact that you can integrate that with Starlink, with solar panels, with other stuff. Then you bring in the humanoid robots, right?

I think this is a monster opportunity to buy a stock that will never be lower than it is now.

Yeah, I agree with you. When we think about Elon's companies, we think about SpaceX, of course, Tesla, X, xAI, and Neuralink. People forget that he still has the Boring Company.

I remember the tweet. He's going from Hawthorne to Beverly Hills to his home, from SpaceX headquarters to his home, and the 405 is jammed. He goes, "This is insane. There needs to be a better way." And he tweets, "I'm going to start tunneling under L.A." That's him. When he has a crazy idea, he goes, "Yep, that's what I'm going to do. It's going to be a boring company. I'll call it the Boring Company."

For the first time ever, the Boring Company has continuously mined in a zero-people-in-the-tunnel configuration. There's no one in the machine, which is simultaneously advancing forward and erecting a ring—in other words, building a tunnel behind it. Each full ring weighs 24,000 pounds.

You can see here in this video a completely automated process of drilling, extracting the earth, and building a tunnel behind.

In my conversations with Palmer Luckey, Palmer was saying that while we have a Space Force, an Air Force, and an Army, eventually there will be an underground force as tunnel technology gets more advanced.

An underground force. Yeah, crazy. I think The Boring Company has the greatest name for a company in history. I agree.

So, that's the articles for this week. Pretty intense, guys. I can't wait for us to get together again. If you've been watching this episode of Moonshots and WTF Just Happened in Tech, our mission here is to deliver to you and contextualize the most important news in the tech world that's impacting every company, every industry, every parent, every aspect of our lives. Please subscribe. Give us your feedback. Tell us what articles you enjoyed listening to, what you want to listen to next. Tell me, do you like Dave better or Salim better, and who else would you like on this podcast?

Can I do a small plug?

Yeah, of course.

We did an ExO workshop last year for 100 bucks. It sold out. People said it was one of the best workshops they ever had. We're doing another one on May 20th. Anybody interested, it's what's being covered in that workshop on May 20th, which is my—how do you build an ExO?

It's how do you build an ExO? So, Peter, you should drop in for 30 seconds, say happy birthday to you.

Yeah, send me the details, but we cover how do you build an ExO, and people loved it. So we're going to just keep doing that every month roughly. The last one sold out, so it'll be great.

Amazing. An Exponential Organization, or ExO, is performing 10× compared to your competition. Salim and I have written an extraordinary book on the subject. Check it out. But more importantly, go sign up for Salim's workshop. We'll put it in the notes below, but where do people go to?

Just go to openexo.com. You'll see a link to it, but we'll put a link in the notes.

Dave, what are you excited about for the coming weeks and months?

Well, regarding that comment in last week’s podcast about the “geeky version of All-In,” I really appreciate people commenting on this podcast, too. Entertaining you, sure, that’s fine. But we actually want to do things and say things that help you map out your life and change your direction.

If you’re suggesting guests or topics, keep it actionable. Really be honest with yourself: Did anything I heard here today impact the way I’m going to spend the next week? That way, we can really focus on those things, because this is the only podcast I know of where we cover not just the technical aspects, but also the societal and macroeconomic aspects. I don’t know of any other place that does that well.

I think the umbrella premise is so simple, right? Technology is a major driver of progress in the world. As Ray Kurzweil says, it may be the only major driver of progress in the world. These breakthroughs that are happening at an increased frequency become more and more material to how our lives are going to be lived.

And please know that between Salim, Dave, myself, and our entire research teams, we’re scouting everything that’s going on in humanoid robotics, in space, in AI, in Bitcoin, in 3D printing, and we’re doing the work. It’s why I’m bald but beautiful. We have a lot of resources, too. If there’s a complicated topic that you really want to understand, we can either figure it out or farm it out and then bring it back to you. Anyway, give us your feedback. We love doing this. Guys, I love having you as my moonshot mates. See you guys soon. Take care, guys. Everybody, thanks for listening to Moonshots. You know, this is the content I love sharing with the world. Every week I put out two blogs, a lot of it from the content here, but these are my personal journals, the things that I'm learning, the conversations I'm having about AI, about longevity, about the important technology transforming all of our worlds. If you're interested, again, please join me and subscribe at dmad.com/subscribe. That's dmadness.com/subscribe. See you next week on Moonshots. [Music]