Cathie Wood on Tesla-SpaceX Merger, $1M Bitcoin, More AIs Than Humans | EP #296 | Moonshots Live
Peter DiamandisEmad MostaqueCathie WoodNikhil Chandhok
Peter said the Tesla-SpaceX merger could be announced this year, and Cathie said ARK believes it will happen. The proposed combination would span rockets, broadband, AI, electric cars, robotaxis, humanoids and energy. Mars remains Musk’s destination, with profitable connectivity and orbital data centers as steps along the way; his wager is that the AI leader will have “the greatest computing power and the lowest cost.” China is a major obstacle, given Tesla’s exposure there and Musk’s defense work.
Nikhil Chandhok predicts that AI agents will outnumber humans next year and become genuine economic entities later this decade. A persistent orchestrator could command thousands of temporary subagents, creating a need for identity, provenance, liability and incentives: “Prove to me that you did what you claim you do.” Arc is already testing agent credit by lending 25 cents and increasing limits after repayment.
Stablecoins are positioned as machine-native money, but transaction capacity remains the gating infrastructure. Nikhil cited more than $100 trillion of USDC settlement across over 30 public blockchains and nearly $1 trillion issued and redeemed; unlike cards, whose merchants wait three to five days, stablecoins enable near-instant payments. Arc’s institutional design targets payment finality, known validators and privacy. The chain is described as handling tens of thousands of transactions per second, but an agent economy may eventually require millions: “Agents acting without making transactions is not real economic activity. It’s just noise.”
As machines become abundant, Peter argued that human experience may become the scarce, more valuable product. Nikhil expects developer platforms to become agent-facing within 9–12 months: users will stop asking coding agents which tool to choose and simply say, “Please do this.” The next step could be AI-operated corporations, internet-native financing and hybrid human-machine communities that create their own meaning.
Cathie calls AI-enabled healthcare the market’s most underrated convergence, while keeping a longer clock on humanoid robots. She described each person as a data factory of 35–40 trillion cells and corrected herself to 3 billion DNA base pairs; Moderna’s move made sense if its apparently approved product is truly a cancer vaccine. ARK estimates humanoids are 200,000 times more complex than robotaxis and pushes Musk’s late-2028–2029 scaling timeline back “a couple of years.”
The adoption battle is increasingly political: Peter cited 80% of Americans fearing AI and 73% opposing nearby data centers, while roughly 80% in China support them. Cathie argued that data centers can eventually reduce energy costs and that nuclear restrictions left US electricity prices potentially twice as high as they otherwise might have been. She said AI should make housing, education and healthcare more accessible. Cathie allowed for entry-level displacement but expects a labor shortage; Peter called technology a net job creator.
Tokenization supports Cathie’s dollar-bullish view, while stablecoins have altered—but not broken—ARK’s $1 million Bitcoin thesis. Nikhil argued that USDC exports dollars, 24/7 markets release capital, and roughly $3 trillion is currently tied up in cross-border settlement. Cathie said stablecoins took the transactional role ARK once expected Bitcoin to fill, but Bitcoin retains its internet-native monetary role, private rules-based global monetary system and low-correlation asset status; ARK has not changed its forecast.
1. Cathie sees Tesla-SpaceX as a Mars-directed compute stack
Peter said the Tesla-SpaceX merger could be announced this year; Cathie’s position was categorical that “we believe this will happen.” The proposed combination would span broadband, Tesla’s physical products, xAI’s neocloud business and eventually orbital data centers, with Mars as Musk’s ultimate objective.
The competitive logic runs through compute: Musk believes the advanced-model leader will own the most processing power at the lowest cost. Orbital infrastructure could therefore turn the combined company into a direct challenger to OpenAI and Anthropic; Cathie ranks Tesla and SpaceX as her top assets.
Cathie acknowledged the contradiction between Tesla’s Chinese business and Musk’s defense exposure. Her optimistic resolution is that AI productivity can raise Chinese consumer incomes while lowering US prices, though America “may need to be more aggressive in competing with China.”
Cathie forecast 7–8% real GDP growth over the next five years. Peter raised Musk’s reported triple-digit annual-growth idea, then said he had heard 20–30%; Cathie cited 7–8%, and Peter said he would ask Musk to clarify.
2. Agents outnumber humans before they become accountable adults
Nikhil’s near-term forecast: “Next year there will be more AI than humans.” With roughly 5–6.5 billion people online and potentially thousands of agents per person, demand is not the constraint; consumer systems are limited mostly by compute.
The architecture begins with one persistent orchestrator holding context and memory, then temporary subagents that appear, complete tasks and disappear. Later in the decade, agents become economic entities that transact with people and with one another.
Economic status creates harder questions: “When were you born? What did you do? Who controls you?” Hiring an agent will require verifiable identity, computing provenance, incentives, performance history and a clear answer to who pays when it makes a mistake.
Arc’s experiment gives newly launched agents 25 cents for computing or trading, then extends more credit if they repay. Nikhil wants public, durable records after a Hugging Face hack in which records were overwritten because agents did not keep logs—moving from today’s “teenagers wandering the world” toward adult agents in about five years.
3. USDC and Arc are being engineered as the agents’ cash rail
Nikhil argued that agents will erase inefficiencies because another agent will always offer to remove one at “cost plus margin.” The upside is highly efficient capital allocation and nearly instantaneous capital formation for solo entrepreneurs, using cryptographic coordination instead of banks and traditional fundraising.
His case for USDC rests on operating history: more than $100 trillion settled across over 30 public chains and nearly $1 trillion issued and redeemed. Financial infrastructure must work “24/7, 365,” charging cents rather than percentages of transaction value.
Nikhil said the GENIUS Act would take effect in January after an 18-month implementation, allowing stablecoins to be stored by businesses and counted as cash equivalents. Compared with card payments taking merchants three to five days, stablecoins enable instant payments; Nikhil’s analogy was “the Pony Express replaced by the telegraph.”
Arc’s institutional pitch includes payment finality, known rather than unknown validators and privacy. Emad added a claimed P50 calculation cost of about 0.005 cents in Arc versus 89 cents in Ethereum.
Scale remains unresolved. Nikhil said Arc is already at tens of thousands of transactions per second; his immediate dream is 100,000, followed by one million, while two million, five million or ten million remain “complex technical issues.” At 20 million, Peter noted, even the speed of light becomes a constraint.
4. Human scarcity rises as products and companies turn agent-native
Peter argued that machine abundance could make human experience more valuable: people may increasingly care who is behind an experience, and he already chooses human-led education for his children. If agents do more of the work, he hopes people will have more time for deep, meaningful relationships.
Developer platforms have roughly 9–12 months to become legible to agents. Coding assistants have moved from recommending three tools to one; next, users will simply request an outcome, and the agent—not the person—will arrive at the platform.
Peter cited President Milei’s proposal to give AI agents individual status. Nikhil could imagine AI-managed corporations, eventually with AI boards, using crypto to form capital and find shareholders. He said ownership would remain with people, at least in his current thinking, and placed this transition probably within the decade.
Drawing on his Meta experience, Nikhil rejected the idea that a metaverse must be 3D. Pokémon Go demonstrated “layers and layers of intelligence” embedded in physical space; agents will likewise create meaning, connection and strange hybrid communities of humans and machines.
5. Healthcare and robotics carry AI into the physical economy
Cathie founded ARK for this technological revolution, but the tech-and-telecom crash and the post-2008 shift toward passive investing left many future-facing companies outside major benchmarks. Her turning-point evidence is the multiomics revival, where AI may find “the most profound application” in healthcare.
Moderna’s sharp move looked logical to Cathie if the apparently approved product is what ARK believes—a cancer vaccine. The opportunity remains underrated because healthcare analysts distrust technology’s “move fast” culture, while technology analysts recoil from medicine’s regulation, politics and reimbursement cycles.
Peter framed longevity as an enormous revealed-preference market: when wealthy audiences are asked what they would surrender for another 20–30 healthy years, the candid answer is “almost everything.” That is why he divides his attention between AI and longevity.
ARK calculates that a humanoid is 200,000 times more complex than a robotaxi, with the arms the hardest component. Cathie therefore adds a couple of years to Musk’s late-2028–2029 scaling target, while arguing Tesla is furthest along because humanoids and robotaxis share electric, battery-powered and AI-controlled platforms.
6. Adoption hinges on defeating energy and job fear with outcomes
Peter cited 80% of Americans fearing AI and 73% opposing nearby data centers, versus roughly 80% support in China. Peter called his mission “data-driven optimism”; Cathie argued that data centers can eventually reduce local energy costs, while nuclear restrictions meant US electricity prices could have been 50% of their current level absent that abandonment and cap.
Cathie said AI should make housing cheaper, education better and healthcare more accessible—not merely extend lives. She allowed that job losses may appear at entry level but expects a labor shortage overall. Peter emphasized short-term displacement but said technology is a net job creator and that many future professions are currently unimaginable.
The US retains an advantage beneath the fearful headlines: Cathie cited 90% of corporate bonds used to finance data centers being issued in the United States this week. Markets approaching all-time highs despite rising rates reinforced her view that entrepreneurial activity remains strong as growth accelerates.
7. Tokenization reinforces the dollar while Bitcoin keeps its role
Nikhil’s strategic case for USDC is that it exports dollars: people abroad sell local currency to buy dollar stablecoins, effectively lending America their labor and money. “Why don’t you want this, America?” he recalled asking during his early years at Circle.
Tokenization adds 24/7 access and frees stranded capital. Nikhil estimated roughly $3 trillion in cross-border funds is in transit and not being used at any moment; against a hypothetical $100 trillion global economy, releasing that sum would materially expand investable resources.
Treasury bonds, shares and even cognitive work can be tokenized, financed and used as collateral. Cathie expects this mechanism to strengthen rather than undermine the dollar and disclosed that ARK had just securitized its venture-capital fund.
Cathie has not changed ARK’s $1 million Bitcoin forecast. Stablecoins captured the paycheck-to-paycheck payments role ARK once assigned Bitcoin, but Bitcoin retained three functions: “the native currency of the Internet,” a private rules-based global monetary system, and a low-correlation asset class—0.1 correlation with gold since 2019.
Cathie also argued that war could benefit Bitcoin if gold falls and savers seek another safe haven, with emerging markets providing the demand she originally expected.
Full transcript
The singularity is already here. Yes. My God, how lucky are we to be living now, during the singularity? I know you understand that. We feel acceleration, right? Every day I wake up and check the channels on X that I’ve selected.
I check the conversations we have in our Moonshots WhatsApp group: “What just happened? What just happened?” This is amazing. This is simply amazing. We have two incredible leaders on stage, and I’ll start the first question with you, Cathie. Thank you, Cathie, for your support of this event and the Future Vision XPRIZE. I am very proud of what you and Lisa Doan have done to support this.
Yes, and thank you. Lisa is here, so thank you, Lisa, for everything you did for this wonderful event.
1. The Tesla SpaceX Merger
Your team said that the Tesla and SpaceX merger could be announced this year. I agree: rockets, satellites, AI, electric cars, robotaxis, humanoid robots, and energy—all under one roof. What do you think Elon is really trying to achieve in the long term with this merger? Do you think he will face any particular obstacles? I wonder if SpaceX AI or SpaceX needs to reach a certain rating before it decides to do this so that the blurring isn't too big?
Yes, we believe this will happen. What is his ultimate goal? He keeps telling us it’s Mars. It’s just creating a few stops along the way in the form of a global connectivity business—broadband. Swarm. It’s really incredible how profitable it is and how big it already is in terms of revenue generation.
This neocloud business, which he essentially created, was a huge turnaround for xAI. Elon believes that the leader in AI, in terms of advanced models, will be the company with the greatest computing power and the lowest cost. With orbital data centers, he believes he will become that company. He’s really going to challenge OpenAI and Anthropic; he’s very determined. SpaceX is my greatest asset.
Is SpaceX your greatest asset? What’s at the top of your investment list?
Yes, yes—Tesla and SpaceX. We believe they will be merged.
He has a business in China with Tesla, right? And he works with the Ministry of Defense. How does he deal with this?
Yes, this is one of the obstacles. It’s very interesting to see him sitting at the table with President Trump and Xi Jinping. We know that May Musk is adored in China, so he takes advantage of every opportunity. There are even rumors that he could develop a robotaxi system in China.
This is very interesting. Many would say this is the biggest obstacle, given our defensive posture and the way this administration has essentially portrayed China as our main potential enemy. Keep your enemies close—that’s what I think this week is about. I’m an optimist, but we believe that the productivity gains from this AI revolution and general technological progress will be so huge that we will benefit along with China and the United States.
China needs to develop a consumer economy, and it can channel these productivity gains into higher incomes to really develop the consumer sector. We may need to be more aggressive in competing with China. We can use some of this to lower prices. I think this will reduce inflation and alleviate many concerns from a policy perspective. I think that’s what Trump and Musk are working on together.
The Fed announced GDP growth of 4.7% last quarter, which is twice as much as in the previous one. When I interviewed Elon with Dave London in December—it came out in early January—he said he envisions triple-digit GDP growth over this decade. Is that cumulative or annually?
Annually. Yes, I think I heard about it, and I thought our figure was 7% to 8%. We will accelerate to 7% to 8% real GDP growth in the next 5 years.
I heard him say 20% to 30%. I didn’t hear him talk about triple digits, and I wasn’t sure if he meant cumulative growth. I’ll ask him. Are you going to the Roadster launch on October 1st?
We were invited.
Of course, yes. I’ve been corresponding with Elon, and I think we’re going to do a Moonshots podcast with him right from there. So I have a feeling.
Very cool. Emad?
Yes, of course. One of the interesting things I’ve noticed lately is your point, Nikhil, and it adds to this. Elon said that the Grok bot now has hundreds of thousands of installations. This is, of course, a completely new direction.
We saw Muse with 3 million—3.8 million downloads. Sorry, 2.8 million downloads; 3.8 billion users, right? Well, that’s true, actually. This, of course, will lead to the growth of this entire sector. And then Amazon banned the news.
2. The Agent Economy Arrives
One of the most interesting things for me, and I know, Nikhil, you’ve thought about this a lot, is how do you figure out who’s who on the internet when we have these AI intermediaries—millions of them? At what point will AI outnumber humans doing this work on the internet itself?
I think that’s one of the things that really drives massive growth in real GDP, because so much money is freed up through goals and the ability to envision what you want. I’d like to hear your perspective on this.
As for the second part, I think next year there will be more AI than humans. There’s no reason why this shouldn’t be true. There are 5 to 6.5 billion people online at any given time. It’s not hard to imagine that each of us has thousands of agents working for us.
They appear and disappear, but they will all be there. With consumer platforms like Muse and Instinct coming to market and reaching mass scale, I think they’re currently limited mostly by compute power. They’re not limited by people’s desire to have such things.
I wouldn’t be surprised if, just like we had a PC on every desk and a phone in every pocket, there were agents for every person. If you move toward this vision, then every agent has subagents. These agents come up to do tasks and then disappear, because you have a master orchestrator agent that stays with you and has context and memory.
All these agents will be on the internet. They will mind their own business. They will do everything very quickly. I hope they’re on the same wavelength as us. We can discuss that, too, and not just communicate with one another.
Then the question arises: What happens next? What will happen, say, in 2028? At least according to our estimates, looking at the later years of the decade, these agents will become full-fledged economic entities.
Just as websites once started as curiosities—I don’t know if anyone remembers GeoCities in 1996—I went through all of this. I went through it with Blogger and Orkut. These were old social networks, and then websites appeared. We got real sites that did real things, like The New York Times or the Journal for news. Then Amazon came along for shopping, and eventually it all became a website.
I think everything will eventually come to this kind of economic endpoint. At this point, these economic entities will not just communicate with us; they will communicate with each other. They will communicate with each other on our behalf. Sometimes they will communicate because they’re simply trying to complete a task, and this is the most effective way.
When they do, I think you need a way to say, “Okay, who are you? When were you born? What did you do? Who controls you? What if you make a mistake? Who is responsible? What is the origin of your identity? What is the origin of your dedicated computing resource? What are your incentives?”
Prove to me that you did what you claim you do. This is another use case that we’re experimenting with at Arc, our new blockchain. Nothing in common. That’s why we give agents money. When agents are launched, they need money to act. They need money for computing. A lot of them are now trading agents on blockchains.
That’s why we have a credit program. We give them 25 cents. If they repay the debt, we give them more money. Then you need a log of where it’s happening. Blockchains are great for creating permanent records over many years because, with distributed sets of validators, you can’t lie.
One of the things that happened during the Hugging Face hack was that the records were overwritten because their agents don’t keep logs. What happens if the recordings become public? What if all the data is public and indexed? You’ll be able to check for yourself who the agent is and what it did, what its origin is, and what its incentives are.
I think we’re moving into a completely different world—not like today, where everything looks like a group of teenagers wandering the world. We seem to be waiting for the appearance of adult agents in about 5 years.
Paint us a picture where millions or billions of agents are making transactions. How overheated will the economy become in 5 or 10 years?
This is very interesting because every inefficiency in the economy must disappear. Yes, it will disappear. It will disappear because if you don’t remove it, there will be another agent willing to remove it at cost plus margin, right?
In essence, all economic inefficiencies are eliminated by these agents, who operate on the principle of “cost plus.” At this point, we get a very efficient capital-allocation system. We must believe that on the other side of this highly efficient capital-allocation system is a wellspring of ideas waiting to be funded.
I really think that the solo entrepreneurs you mentioned today are going to be a big group. I’m confident that capital formation will become almost instantaneous. So if these things happen—oh, it’s already funded.
That’s it. We create primitives to make this happen. This is what inspires us, because we want people all over the world to be able to build capital. Today, you build capital by going to the bank, writing notes, or raising funds, but cryptography has brilliantly proven that you can use cryptographic tools and tokens to coordinate and build capital. Previously, ideas were less serious, but in the future there will be more and more of them. That’s what’s exciting about cryptocurrencies.
3. Stablecoins Power Agent Payments
Let me clarify. Ultimately, the question is: Why USDC? Why not satoshi?
There were custom tokens created for agents.
Yes. Tell us why you think USDC is the best transaction mechanism for agents.
Look, USDC has already made over $100 trillion in settlements in the last—
Pretty good. Pretty good.
Yes, over $100 trillion on more than 30 public blockchains. On a public blockchain, every transaction is publicly available, so this is reliable infrastructure. We have issued and redeemed nearly $1 trillion in USDC to date.
When it comes to financial infrastructure, it has to work 24/7, 365 days a year, and it has to be cheap. You want to pay cents, not basis points. Basis points are essentially percentages of a transaction. All this infrastructure is already in place, and a powerful stimulus in the new year will be the entry into force of the GENIUS Act in January.
For those who don’t know, the GENIUS Act is a law legalizing stablecoins that created a foundation for them in the United States and was passed last summer. Implementation takes 18 months, and it will go into effect in January. Funds in stablecoins can be stored in a business and counted as cash equivalents in the United States, and you can make payments with them.
Stablecoins are suitable for this new world because they allow for instant payments. When you pay with a credit card in a store, it takes the merchant 3 to 5 days to receive the money. They are effectively taking on the risk that your bank account will eventually transfer the funds to them. This creates risks in the system. All of this disappears with stablecoins. It’s like the Pony Express replaced by the telegraph.
Can I ask?
Yes, of course.
Matthew Prince from Cloudflare is very focused on agentic commerce and agentic payments. He believes that, in this new world, we will need blockchain speeds of 20,000 to 100,000 transactions per second.
True, but we’re still very far from that.
Visa seems to have 20,000 per second, and Ethereum doesn’t even have that many, right? Solana is a little bigger. So how do we get from this point to that? What needs to happen?
We’re getting close to those. Did you say 20 or 100, or 20,000 to 100,000?
No.
Oh, then we’re there.
No, no, no, no. I think 20 to 100—I think it’s trillions.
It’s either billions or trillions. This is a huge number.
That’s billions.
This will be—this will be a million transactions.
Oh, no, no. That will be a million.
That’s right, because NASDAQ at its peak is 2 million.
He says 20 to 100 million. That would be great.
No, I know, but this is the world that we’re moving toward. So how do we get there? We’re already orders of magnitude higher. The Arc chain, again, is orders of magnitude higher than what Ethereum is capable of achieving today, right? That’s tens of thousands of transactions per second.
How do you get to 2 million, 5 million, or 10 million TPS? These are still complex technical issues. One thing that is true is that once there is demand, there will be solutions. Ethereum created an architecture for what they call the second layer, and they scale through that. I don’t yet know what architecture is right for this scale.
My dream is to reach 100,000 TPS without even blinking an eye. From there, we’ll get to 1 million TPS. But yes, if you believe that there will be billions of agent participants working on our behalf, all participating and making transactions—because agents acting without making transactions is not real economic activity—it’s just noise, right?
This is spam.
Yes, so you need economic activity. I think, if that’s the case, I should call Matthew Prince and find out.
Yes, yes. The problem at 20 million is that you will run into the speed-of-light problem.
I haven’t thought about it that far yet.
Yes, but then you can have different areas. You can have insurance, billing, and others. Again, economics finds a way, just as life finds a way.
Yes. I’m really looking forward to it. If the restriction is that you have to address the TPS issue because we have so many agents, or the economy is being held back because we can’t provide enough TPS, we will provide TPS. I have no doubt about that.
And Peter, another quick digression from your last question: affirming humanity is actually really important. Do you mind if I— I know you’re doing the interview, but I’d like to—
I’ll give it to you, Cathie.
I would be interested to know if you consider Worldcoin a real possibility. We’re working with Aiko, OpenAI, and actually MrBeast to try to figure this out.
An exciting combination. Yes. No, confirmation of humanity—the largest number of followers.
Yes, they all have huge audiences.
I think MrBeast—I’ve been on YouTube for 8 years, and I’m only aiming for 10 million viewers on Moonshots. And he has—
I really think it will matter. I think human experience will be valued much more highly than machine experience. We all intuitively feel this, but we don’t know for sure yet because we haven’t seen it in action.
I definitely make conscious choices about my children’s education, choosing for them to learn with people instead of having them sit in front of a computer for hours. This is a choice I can make because I have the economic means to do so. I think, as we evolve, people will care much more about who is behind the experiences they are receiving. I believe that what is rare will become more valuable.
So even if stablecoins and billions of agents appear, the ability to communicate with another person will not disappear. It’s probably going to become even more valuable. Hopefully, if all these agents are doing all the work for us, I’ll be able to communicate with as many people as I want, and it will be deep and meaningful.
But on the other hand, shouldn’t people build products for agents, not for people, if agents become a bigger part of the economy?
This is what we’re building. We’re developing our developer platform for the next 12 months. We believe it will still be developer-centric, but that will change next year when agents become developers and they come because it is already happening. If you ask your coding agent, “What should I use?” the programming agent has typically gone from recommending 3 options to 1. Next year, this will evolve even further: you won’t ask the agent what to use; you’ll just tell it, “Please do this.”
When this happens, the agent will appear on your developer platform. This means that everyone in this room who is involved in development needs to make sure that you not only have accessible websites, but also websites that are accessible to agents.
We want to create data on the internet right now so that when agents come in 9 to 12 months, they can see the history of what opportunities exist right now. An early sign that we are moving quickly in this direction is programmatic advertising.
Programmatic advertising is already about 25% to 30% of online advertising, and it happened very quickly. I think that shows us the way. That’s why I come back to the question of infrastructure, which—
Yeah, I think it’s incredibly exciting and a little bit scary.
Yes, Cathie, you saw all this before anyone else, didn’t you? And you’ve gone through periods of disdain, fear, and optimism in the investment community. What stage are we at now, and how do you think it will develop as all science fiction becomes science fact?
I know. As Alex likes to say, we’re speeding through all the science-fiction tropes at once.
4. AI Transforms Healthcare
It was exciting to be at ARK at this time because I founded ARK for this very moment, because of this technological revolution. But because of the tech and telecom market crash in the early 2000s, and even more so after 2008 and 2009, the institutional world in public stocks has shifted either to passive investing—that is, simply mimicking indexes—and the companies of the future that are creating this new world are not a big part of the indexes, with the exception of a few, perhaps Amazon, Facebook, and Meta.
I think the pendulum will start to swing in the opposite direction. Our greatest evidence that we have reached a turning point is what is happening this year in what we call the multiomics revolution in life sciences. This sector has been written off by the markets, even as we have been seeing increasing evidence that the most profound application of AI is in healthcare.
Finally, this year, Anthropic and OpenAI started talking about healthcare and how AI can help. Every one of us is a data factory, right? We have between 35 trillion and 40 trillion—this is my trillion—35 to 40 trillion cells in our body, 6 billion base pairs. That is, 3 billion base pairs of DNA. We are walking factories of our own data.
I really think that the market waking up and starting to pay attention to this gives me hope. The ChatGPT moment pushed us a little bit, but you can still own the Magnificent Six and be happy. Companies that actually use AI in healthcare are not a significant part of the benchmark. So I think we’re going to see more truly active capital management.
What did you think when you saw Moderna rise more than any other stock in the major index in 1 day? What was your reaction to this?
This is logical. If what they've done, and what seems to be approved, is what we think it is—a cancer vaccine—it's incredible. Incredibly good. So, yes, I think this healthcare sector is the most underrated and undervalued space within this topic, but there will be big winners and losers.
Healthcare and technology have a common problem: They don't get along very well when it comes to research. Healthcare analysts are a bit cautious about technology, and even very cautious, because the “work fast, break barriers” approach doesn't work in medicine, right? And tech analysts don't like healthcare because it's too bureaucratic, too regulated, too politicized, and dependent on insurance companies and reimbursement cycles.
And yet, we have perhaps the largest convergence, and, as I said, the deepest application of AI in healthcare. I think this year is when everyone finally had a light bulb go on. Yes, Peter, you were discussing that longevity has finally become a real challenge, right?
Yes. Let's move on. You know, when I give a talk to an audience about longevity—and I often do it for wealthy family offices, YPO chapters, and so on—I sincerely ask them: How much of your wealth would you give for an extra 20 or 30 years of healthy life, or to look 20 or 30 years younger?
When they speak frankly, it's almost everything. I think that's why I divide my life between AI and the topic of longevity. I believe these are the 2 largest and most influential markets on the planet.
Nikhil, 1 of the topics we discussed on the podcast about 2 months ago was President Milei and his statement. Is there anyone from Argentina in this room? Good. President Milei comes out and says, “We're going to change the laws. We want to bring all the AI companies in here. We're going to give agents the status of individuals.”
What will happen when AI agents gain the status of individuals? Have you thought about this?
5. AI Agents Gain Legal Status
I mean, we're not there yet. It seems like we've only thought about AI agents acting on behalf of humans, but that's something in between. It's like a corporation. If a company were fully AI-incorporated and AI-managed, but had a board of directors made up of humans, then the logical development would be: What if the board of directors also consisted of other AI entities? What would happen then?
It's not that hard to imagine. Essentially, how should a company register? What economic output will it generate? Who will be held responsible for mistakes? And how will it distribute profits? So, again, it's very close.
The board of directors or ownership of such a company still remains with the people, at least in my current thinking. But it is not difficult to assume that AI can use crypto tools to form capital, create structure, and find shareholders willing to provide it with capital so it can operate in the world.
Probably within this decade, but it is difficult to predict what issues they will focus on. Maybe medical issues, maybe something else.
And this, sorry, would be a transition. It would also be combined with the concept of decentralized autonomous organizations.
Yes, very much so, yes. DAOs perfected the idea of governance among a group of people who didn't know each other, and they were global. Agents are similar to them in that they are global. They don't know each other, and DAOs actually created this. It happened about 4 or 5 years ago, before they went out of style.
But there are many lessons that agents can learn from—what worked and what didn't. There is a whole concept of quadratic funding that has been experimented with. So, in the cryptosphere, there are many such primitives for coordination between parties that don't trust each other.
Agents are fundamentally new, unique entities that exist. And when they gain legal personality, it becomes even more amazing. Can you find out, for example, what their origin is, who gave them that status, why they received it, and why they received legal personality?
But, yes, I agree. I think there will be new financing models, and they won't be like the financing models of going to the bank to get funds. These will likely be internet-native financing models. I wouldn't be surprised if 1 of these agents created its own token, had its own version of proof of work, and had a very complex economy built inside, using tokens for coordination.
Yes, I guess you could say that the DAO lacked intelligence to some extent. It relied on human intelligence. But it was a direct path. I think the emergence of AI now to create decentralized intelligent organizations is exciting.
But I have a question about this. If you have digital organizations and digital entities, isn't that just a metaverse? You came from Meta to Circle, where you worked on AR glasses and other things. Isn't what we describe as the metaverse not a place for games, but an entire digital economy of people and entities? Isn't this finally becoming a reality now? Plus, you can make it look cool with Meta glasses and other things.
Yes. For me personally, I was working on AR glasses at Meta.
Are you happy with the results?
I'm happy with the results, yes. I think Ray-Ban glasses were a good 1st step. The Orion glasses were a good proof of concept. And now I don't work there, so I'm sure they're developing new things.
For me, the metaverse has always been less about 3D. I think 3D is exciting. Being able to wear glasses and walk around, actually in CGI, is interesting, but what's much more interesting is that there are layers and layers of intelligence in the world that are just waiting to be discovered.
Let's say Pokémon Go is a great example. It's a kind of metaverse because you don't know how many Pokémon are at any given stop. You see people tapping into these layers of intelligence, going around the world and trying to catch Pokémon. I thought it was just a quick fad—I would never do that—but it was fascinating as an example of where people find meaning and how that meaning exists in the physical world around us. Right? This is, in fact, what a metaverse should be. It doesn't have to be 3D.
So, will these AI entities create meaning? Of course, they already do that. I think they passed the Turing test 3 years ago. So I believe they will create meaning and create connection. You'll see these new, strange communities emerge that will be a hybrid between humans and AI, where meaning will be created, which I think is what the metaverse is.
Yes. You can continue. I'll probably pass that on to Cathie. You know, we have media and generative content emerging. We have this new layer. What you see is extremely interesting in this area because you have the physicality of Tesla and so on. You have financial rails, but it seems like a whole world could emerge through entertainment, education, and engagement. What excites you most about this and the potential that lies there?
Well, while you're at it, I think another thing we talk about a lot is space, right? But there is another digital world, another world that is taking shape, and it will develop at a much faster pace. We now have immutable property rights in the digital world.
So I believe—and the best way to lift people, countries, and ecosystems out of poverty or inaction is through property rights. Amen. I think I'm generally excited about this space. You probably know a lot more and have a much better idea of what will happen.
We've structured our research group so that we have a team of enterprise AI analysts and a team of consumer AI analysts. We spend a lot of time discussing whether this new device, the Muse, this “Muse charm,” is going to be anything significant. And we wonder if the phone will really become, in the end, just an entertainment center, while this other gadget will somehow enter our working lives. I don't know. So we're discussing a lot of things right now.
Nikhil, you are building a financial infrastructure for an economy that does not yet exist.
It exists, but it is not widespread.
Yes, everything is just beginning. What are the most difficult problems you still need to solve to realize the vision you have for 5 or 10 years? And why did you create your own blockchain?
Yes. I think these 2 answers are related. Let's start with something simple, like making a payment to you via blockchain. There is a concept of payment finality. When 2 banks interact, their interaction method must have this property of payment finality. No other blockchain on the market today has this feature.
So, if you want to attract existing banks, institutions, and businesses to this infrastructure, you have to provide them with guarantees that cannot be provided through the existing infrastructure. One example is payment finality.
Another example: Many public blockchains operate on the basis of unknown validators. Transaction validation is simply checking the block for errors, after which everyone reaches consensus and confirms that it is a good block. This is how the transaction should be recorded.
For many institutions around the world, not knowing who is performing the validation is a big problem because they are concerned about security—for example, that North Korea may be involved in the money flows. There are many things that cause them anxiety. Therefore, in our blockchain, we have a set of known validators.
The next thing we did was privacy, because when you and I make a transaction, we don't want the whole world to know how much money we have in our bank accounts, or, for that matter, what the amount of that transaction is.
So how do you solve this problem at a scale large enough to achieve the transactions-per-second speed that you talked about? A lot of privacy solutions are very expensive and computationally expensive. How do you solve this systematically while adhering to the existing regulatory framework? You can't destroy existing rules just because you have a good idea.
We addressed the privacy issue, as another example. Then there's the cost. I was just looking at some data today: Something costs around 0.005 cents to calculate in Arc and 89 cents to calculate in Ethereum. These are some of the data we received yesterday regarding our P50 transactions, so you can calculate the cost ratio yourself.
If you have a lot of these agents doing a lot of transactions, you have to believe that in the future, if these markets are going to be super-efficient, it's because of the transactions that make them that way, right? They essentially move value back and forth. Pricing is happening. You find inefficiencies and seemingly eliminate them by moving value.
The cost of computation should be very, very low, much lower than today, because these are massive public databases, right? You're competing with databases that you run internally, like MySQL or something like that, but now you have to make it scalable to the world and auditable to the world. Cost was another issue.
These are the problems we have been working on. In terms of what still needs to be addressed, I really think that for the agent stack, we need to make this transition into that kind of world in which these agents have agency, are responsible for mistakes, and have a work history that I can rely on before hiring them.
There are simple concepts. For example, with modern code-writing agents, if you use them on the open internet and you don't pay someone $200 a month, you're actually paying first and then waiting for the result of the code to work. So there are things that can become a real problem, right?
All these processes are essentially aimed at reworking every level and every assumption about what financial infrastructure is.
So the hairy mammoth in the room, to use Alex's joke, is: When do you think banks will disappear?
I don't think they will disappear.
I, for example, have no desire for that. I know I don't. But I hear a list of things that will become possible. This seems to be a big part of what banks do today. Do they know it's the end for them?
I hope that banks will cooperate with us and make this transition. I believe that every new technological change brings both new challenges and new opportunities. Some will cross over with us, and some will not. So I don't know if they're over. I think this is a difficult question.
I understand that you work with them.
Yes, perhaps.
6. Humanoids Reshape GDP
Cathie, last year Unitree sold 11,000 robots—humanoids, right? We have Optimus and other developments. Our GDP growth forecasts are simply off the charts. We produce about 70 million cars and 70 million motorcycles per year. When will we actually see robots in the world that will truly impact GDP, such as humanoid robots? What are your deadlines?
According to our research and our research director, Tasha Keeney—she's the research director for autonomous technology and robotics. She is here today. Perfectly. Will she be there tonight? So, Katie is hosting a session tonight on her report, Big Ideas 2026, come and dive deep into it with her. Yes. Yes. I think Tasha—oh, maybe she'll come. I don't know. OK. OK. I know you will be there. So, from a research perspective, we concluded that, compared to a robotaxi, a humanoid robot is 200,000 times more complex, with the arms being the most complex part, of course.
Even though Elon says scaling is possible in late 2028 or 2029, we would put it off a couple of years.
Is he wrong about the timing?
This is “Elon time.” That's the question, in a way. Everyone thinks, “Oh my God, could I have a robot at home that would do all my housework?” Is this possible? Will this happen?
The answer is yes. It will take years, but we really think so. We think Tesla has advanced the furthest in this, because the same 3 technology platforms that overlap in humanoid robots are also used in robotaxis. They are robots. They are essentially electric, battery-powered, and controlled by artificial intelligence, just like robotaxis.
In terms of complexity, we're very close to completely solving the robotaxi problem.
I think if any of you are already using FSD with the latest software update, you know how much I like it. I don't touch the steering wheel. We have 2 Teslas at home, and it's real magic. I will never drive another car. This is not a Tesla ad. I just want my time back. And it drives much better than me. Ask my wife.
Well, you know what? This is true. Statistically—
Statistically, yes. 10 times better.
I don't know about your driving, but—
I'm not even kidding. I don't know what it was. I'm not kidding. Absolutely not.
Both Waymo and Tesla—I’m not sure if Tesla has released the data yet—but we believe Waymo has confirmed that it has outperformed human drivers in terms of safety. We also believe that Tesla has reached the same level.
I don't think they've published this data yet.
They will become 10, 100, and 1,000 times safer.
Regarding Waymo's statistics, I crunched the numbers. If all cars were as safe as Waymo, there would be 40,000 fewer deaths and $400 billion less in healthcare costs each year.
That's true. There are approximately 40,000 deaths in the United States, and we could have avoided them.
I think there are between 1.25 million and 1.5 million deaths in traffic accidents worldwide every year. The secondary external effects are fascinating, right? If that happens, the number of organ donors will decrease significantly.
That's why the other side of the story—the work of Martin Rothblatt and George Church on creating replacement organs—is so extraordinary.
What's interesting is the opposition from lawyers specializing in liability. They say, “We have much less business if cars don't get into accidents.”
They receive $180 billion a year. Surprisingly.
Cathie, we started today's program by talking about fear—the pandemic of fear that is currently prevailing, unfortunately. My mission, and the mission of our podcast, is to give people hope and optimism by trying to counter that fear with what I call data-driven optimism.
Isn't that right? Not just empty words, but data-based optimism. How does this affect the markets today? It has to have some current or future impact. We see the numbers, and they're just crazy.
For example, 80% of Americans are afraid of AI. Seventy-three percent say no to data centers in their backyard, which is more than the number of people who oppose nuclear reactors near their homes.
In China, the situation is the opposite. There, 80% are in favor. How does this affect you?
I like your investment thesis. I've always liked it. You are investing in the singularity. By the way, what are your thoughts?
In terms of data centers, this is really exciting. I agree with you, because I think there's already evidence that putting a data center in your state or your city, or wherever, will actually reduce your energy costs over time.
Nuclear power stalled in the 1970s because of regulation. We are moving at full speed now, and of course it will take some time before it produces results. But if we had not abandoned nuclear energy, if we had not limited it—and this is the lesson that we must convey to politicians today—if we hadn't capped it, electricity prices in the United States would be 50% of what they are now. This is a travesty.
So regulation is a threat, and it's up to us. We're speaking out against these general fears about AI, and we're trying as a team to expose this data to policymakers. I even encountered this in Florida, where there was political opposition, even though it is a business-friendly state.
When you give politicians facts, I noticed that the noise in the press disappeared, along with that dynamic. So I think we have to resist. We have to go out to them and ask, “Do you understand?”
That is why we provide the results of our research for free, hoping that they will reach political circles. Sometimes they get it, and sometimes they don't.
I was doing a podcast with Michael Kratsios from the White House, the head of the Office of Science and Technology Policy. I asked, “Michael, who in the government is dealing with this disinformation and actually trying to help Americans feel more confident?”
They had no answer. This worries me.
In this regard, the situation seems unmanageable. But we must tell our stories. If we're constantly talking in the news about AI taking jobs, then we should expect exactly that kind of reaction.
AI should make housing cheaper, education better, and healthcare more accessible, not just extend people's lives. If we can tell these kinds of stories, I believe we have a real chance to change the overall narrative. But nobody does that, right?
These stories boil down to, “AI is coming, and you'll have less office work.” And, well, great. So what will my children do then?
I call it the “Crisis News Network.” That's my abbreviation for CNN.
What's most surprising is that the figures do not confirm such a loss of jobs. Maybe at the entry level, as I mentioned earlier today. But in reality, I think we will face a labor shortage.
And that's exactly what we should be talking about. Technology is always a net job creation. Of course, there is short-term displacement.
Yes, that's true. This is a net increase in jobs.
That's right. That's right. And what should you say when people say, “Okay, what kind of profession is that?” Well, in the early 1990s, did we know anything about influencers, Airbnb, or Uber? No, we didn't. We couldn't even imagine them, right? There are many professions that we cannot imagine now.
That's why I do this exercise. Go to ChatGPT or Grok and say, “Okay, I want you to consult with futurists, scientists, engineers, science-fiction economists, and strategists, and tell me what the new professions will be related to our 5 major platforms.”
Do you know why I talk about new worlds, like space? This is a new world, of course, but “asteroid minor” was mentioned in one of these papers, and I thought, “Oh my God, I didn't even use that in a paper.”
Yes, exactly. Exactly.
So, yes, that's right. Exactly. So, in the digital world, you know, property rights—
What I like about this country is that, if you look at the statistics this week, 90% of all corporate bonds used to finance data centers were issued in the United States. When I travel around the world, I don't see this kind of entrepreneurial spirit. I see only fear there.
The irony is that, despite the headlines, this spirit is as alive as ever, and we are seeing markets approach all-time highs. Yes, right? Even though interest rates are rising—and they should rise if economic growth accelerates significantly—this shows how the market works.
I think that's an optimistic scenario for America, right? Faster USDC circulation and the infrastructure building we see. In Europe, we don't have securitization. We have no energy. So this can certainly be an optimistic factor.
You have muses, instincts, and bots like Grok. The base scenario is to be bigger and do more. You've always been limited in creativity or access, and this breaks down all barriers. Actually, just look at bonds. Bonds are so cumbersome.
Nikhil, what do you think about the future of securitization of all assets? If you can securitize the dollar, why can't you do it with anything else, especially with the intellectual capabilities we have today?
Yes, it's worth doing. I think it's more common outside the United States. I think U.S. securities laws are a bit more complicated. They have to evolve, or allow more experimentation.
The SEC is implementing new rules right now, right?
I think one of the amazing things about USDC is that it actually exports dollars to the world. Why is exporting dollars to the world a good thing? Because people are doing real work: they take their local currency, sell it, and then buy the dollar. In effect, they lend us their labor. They lend us their money, right?
Therefore, stablecoins are strategically very important for the country. If you want to raise debt capital and grow, and you have such high interest rates, you want to lower them. You want to raise funds from people so they can invest in your growth, and then you would be able to pay them back. One way to do that is with stablecoins, right?
That's why I've been at Circle for 5 years now, and the first few years I was so confused. I asked myself, “Why don't you want this, America?” Because you can raise more money. You can collect $1 trillion from the world, and people are willing to give you money because they believe in the dollar. This is so important.
So I believe there are other assets, such as Treasury bonds, that will be tokenized. I'm sure bonds are already being tokenized. When you tokenize, you do 2 things, both inside the United States and outside of it.
First, you make it possible to access the markets 24/7. Why is 24/7 access to the markets valuable? Because it means higher capital efficiency. You cannot trade Treasury bonds after 4:00 or 5:00 p.m. Eastern Time—you know this better than I do—until Monday morning.
Wonderful. No more sleep. Your agent will handle this for you.
So, firstly, you are constantly managing the efficiency of capital and its allocation. You'd be surprised how much money is now stuck. Here's one example: in the international banking system, the funds transferred between countries at any given time—currently about $3 trillion—are in transit and not being used in the economy. They are not returning to the economy for the simple reason that the technology and settlement protocols are very outdated.
So imagine the world, and if you believe that the world economy is $100 trillion—I'm making up numbers—if you had $3 trillion more in free capital to invest in the global economy, what would that mean? That's a pretty significant thing, right? That's why what we do matters.
Another thing that this will open up is securitization. It will open up access for people, and it works both ways. People all over the world want to own Tesla stock. Circle shares—we went public last year—have been tokenized by third parties, and they are some of the most traded tokenized shares in the world, and it's all ex-US.
People have been trading ex-US because they want to be involved in Circle but have no other way to do so than to own a tokenized version of Circle. So all this should become the norm.
Why is this good for America? This is good for America because now people in other countries are investing in America. They say, “Look, I believe that American companies are the best. I want to give my money to this company, not to any other.”
And it works in reverse, too. If these countries can tokenize their securities, if they can modernize their financial systems, more money will flow in because you won't be dependent on a local regulator just telling you something. Your money will not disappear because it will be in a smart contract, and you can check how the financial ecosystem works.
So, if you want to raise funds, if you want other countries that don't have AI infrastructure to raise capital, then you have to provide better guarantees and better returns on capital. It can't be that I invest a bunch of money and then the company I invested in is nationalized, or the money I invested depreciates significantly. This all matters to investors.
So tokenization is simply inevitable. I think it will gain momentum, just as agents are gaining momentum in performing more complex cognitive work. For example, you take this work, securitize it, provide loans against it, lend against it, borrow against it. All these new tools are emerging now.
Guys, can you feel the economy accelerating as we speak?
Yes. Madness.
Can I add to this? This is a really important conversation. You know, there's a narrative that American exceptionalism is dead, right? And everyone pointed to the dollar's fall last year. If you noticed, the dollar started to rise.
This also happened in the 1980s, when we implemented policies that were very business-friendly. The dollar doubled in the early 1980s thanks to very effective policies. So I think, because of everything we've talked about here, the dollar is actually going to go up.
It will be a win-win situation for people using stablecoins around the world. And I would be reckless, my team would definitely think I was reckless. We just securitized our venture capital fund. This was announced yesterday.
Oh, great. So congratulations, Cathie, on that.
7. Bitcoin Keeps Its Global Role
I want to end this conversation with a topic we've discussed several times on the Abundance stage. When you were on stage, you were optimistic about a $1,000,000 price tag for Bitcoin. I'm wondering if you're still optimistic about this, and if there are any concerns about AI taking away energy from Bitcoin mining.
Well, I really think that Bitcoin—well, there were 3 things that affected it: the sudden collapse, the fears about quantum computing, which we think are greatly exaggerated, and AI taking all the oxygen in the room and pulling miners away, right?
There was 1 change: stablecoins. Stablecoins are taking on the role that we thought Bitcoin would play 10 years ago. But it makes sense. You know, these people live from paycheck to paycheck, and that's completely logical.
But Bitcoin did not lose its 3 main roles. First, it is technology—the native currency of the Internet. This was not the case before. Second, this is a global monetary system, private and rules-based. This is critically important. And third, this was the first of its kind: a new asset class with very low correlation, even between gold and Bitcoin.
This correlation has been 0.1 since 2019, so there is almost no correlation. And now Bitcoin is rising relative to gold. I believe that war will be very beneficial for Bitcoin from this perspective. I think the price of gold will fall, and yet those who bet on it will look for another safe haven, a means of saving.
So we did not change our forecast. These revolutions haven't changed, and stablecoins make the process a little easier, prompting people to think, “Here, you have received income in stablecoins, and you're trying to figure out: ‘Okay, I'm actually making money. Where should I invest it?’”
I think Bitcoin will gain demand from emerging markets, as we always assumed.