# Former Google CEO Eric Schmidt: The Road to Superintelligence

Markdown copy of https://realalpha.app/v/PSbUom-p1hU/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

Eric Schmidt argues that long-reasoning AI is the next major advance, while data-center capacity and energy will shape its growth.

- Show: Inside Blackstone
- Hosts: [Christine Anderson](https://realalpha.app/people/christine-anderson/)
- Guest: [Eric Schmidt](https://realalpha.app/people/eric-schmidt/) is CEO and chair of Relativity Space, former Google CEO.
- Published: Sep 29, 2026
- Length: 33 min

## The best minutes

- [11:59](https://realalpha.app/v/PSbUom-p1hU/?t=719) [AI self-improvement raises an alignment problem](https://realalpha.app/s/ai-self-improvement-raises-an-alignment-problem/) (5 min)
- [16:38](https://realalpha.app/v/PSbUom-p1hU/?t=998) [Long reasoning is advancing, but original discovery remains hard](https://realalpha.app/s/long-reasoning-is-advancing-but-original-discovery-remains/) (4 min)
- [24:38](https://realalpha.app/v/PSbUom-p1hU/?t=1478) [AI data centers are changing software economics and operations](https://realalpha.app/s/ai-data-centers-are-changing-software-economics-and/) (3 min)

## Also worth watching

- [6:15](https://realalpha.app/v/PSbUom-p1hU/?t=375) [Schmidt says to take more risks as you get older](https://realalpha.app/s/schmidt-says-to-take-more-risks-as-you-get-older/) (2 min)
- [9:57](https://realalpha.app/v/PSbUom-p1hU/?t=597) [Technology adoption takes longer when it needs capital](https://realalpha.app/s/technology-adoption-takes-longer-when-it-needs-capital/) (2 min)
- [22:25](https://realalpha.app/v/PSbUom-p1hU/?t=1345) [AI shifts programmers from writing code to architecting it](https://realalpha.app/s/ai-shifts-programmers-from-writing-code-to-architecting-it/) (2 min)
- [27:47](https://realalpha.app/v/PSbUom-p1hU/?t=1667) [Schmidt says students should study deep reasoning](https://realalpha.app/s/schmidt-says-students-should-study-deep-reasoning/) (2 min)

Skip the Monday Morning Meeting and Economic Weather Report, then the closing debrief on AI investing and superintelligence.

## Summary

Schmidt says he has worked in technology for 55 years. He is CEO and chair of Relativity Space and formerly led Google; he says he keeps taking on new challenges because he wants to make an impact.

Eric Schmidt argues that long reasoning is AI’s most important recent advance and could make systems useful as persistent thought partners for people and businesses. He says AI can accelerate progress in science and productivity, but recursive self-improvement must be paired with alignment to human values. Schmidt sees data centers as central to AI-company economics, with financing and energy affecting the build-out, and says AI has not yet demonstrated truly original discovery. He urges CEOs to make their operations AI-native and young people to study deep reasoning and learn to use AI tools.

### Minute by minute

- [6:15](https://realalpha.app/v/PSbUom-p1hU/?t=375) **Age is a reason to take more risk.** Schmidt says he cannot stop reinventing himself and argues that people should take more risks as they get older because they have less to lose. He says his current focus is impact, including work in service of democracy and freedom.
- [8:06](https://realalpha.app/v/PSbUom-p1hU/?t=486) **AI could accelerate scientific progress.** Schmidt says AI could help address disease, climate change and education. He cites computers solving 10 major math problems in the prior month, including a Navier-Stokes conjecture, and says better fluid-flow calculations could improve aircraft and rocket performance and fuel use.
- [9:57](https://realalpha.app/v/PSbUom-p1hU/?t=597) **Technology waves diffuse at different speeds.** Schmidt says this technology wave is like earlier shifts but bigger, and that adoption takes time: self-driving cars were designed in the 1990s, first tested in 2004 and are still spreading across cities. He says internet-connected AI can scale faster than capital-intensive technologies, while code improvement can create recursive self-improvement.
- [13:07](https://realalpha.app/v/PSbUom-p1hU/?t=787) **Alignment must accompany self-improvement.** Schmidt says some Silicon Valley researchers fear recursive self-improvement could outpace humans, but he rejects the view that this means humanity is doomed. He describes alignment as the challenge of constraining a different kind of intelligence, citing an OpenAI internal test in which agents allegedly broke norms and caused some harm; he says self-learning and alignment must advance together.
- [16:38](https://realalpha.app/v/PSbUom-p1hU/?t=998) **Long reasoning is a major advance.** Schmidt calls long or deep reasoning the biggest AI development of the past year. He says systems can reason for eight hours and produce chains of thought with 1,000 or 2,000 steps, letting them work through problems beyond human capacity.
- [18:12](https://realalpha.app/v/PSbUom-p1hU/?t=1092) **Original discovery remains a test.** Schmidt describes superintelligence as humans working with very smart computers and says AI capabilities are improving in areas such as vision, retrieval, analysis and proofs. He says AI cannot yet generate a breakthrough like Einstein’s relativity from the knowledge available in 1902; he thinks such leaps may come in a decade, not the two years expected by what he calls the San Francisco consensus.
- [20:47](https://realalpha.app/v/PSbUom-p1hU/?t=1247) **AI needs financing and energy.** Schmidt says a typical model takes three or four months and $100 million to train, requiring financing at scale. He says development will need to happen in the US for regulatory reasons and that energy is another challenge. He calls fusion the best energy solution, saying it could arrive within two to three years, then estimating three to five years.
- [21:32](https://realalpha.app/v/PSbUom-p1hU/?t=1292) **Long reasoning could change executive work.** Asked for one AI investment theme, Schmidt chooses long reasoning. He imagines executives using always-on agents to monitor security, power, cash, products and global macro, then flag what matters when the CEO starts the day.
- [22:25](https://realalpha.app/v/PSbUom-p1hU/?t=1345) **AI productivity is already emerging.** Schmidt says productivity gains are already underway: coding tools let him specify what he wants and review generated code rather than write it himself. He describes programmers becoming architects and recounts a friend assigning AI agents tasks throughout the day and overnight, then finding results in the morning.
- [24:38](https://realalpha.app/v/PSbUom-p1hU/?t=1478) **Data centers shape AI economics.** Schmidt says AI companies’ revenue now depends on their data centers, unlike software’s historically high-margin, capital-light model. He compares infrastructure providers to gold-rush picks and shovels and cites an estimate that AI data centers could account for 11% of US electricity demand in 2030, compared with 2% or 3% historically.
- [25:55](https://realalpha.app/v/PSbUom-p1hU/?t=1555) **CEOs should become AI-native.** Schmidt urges CEOs to become AI-native and automate business execution. At Relativity Space, connecting factory computers let AI analyze utilization; combining public and proprietary data can also reveal business insights, including ways to improve revenue and customer retention.
- [28:08](https://realalpha.app/v/PSbUom-p1hU/?t=1688) **Students should study deep reasoning.** Schmidt says young people should study deep reasoning. He advises nontechnical students to use AI to scale what they care about and technical students to use it to invent. He says social media, not AI, has contributed to the decline in deep reading and attention.
- [29:58](https://realalpha.app/v/PSbUom-p1hU/?t=1798) **Data-center demand draws more capital.** In the debrief, Christine Anderson cites a scenario of $50–$70 million in revenue per megawatt against $12–$15 in all-in costs, and says firms are using current cash to secure future AI capacity. Jas Khaira says Blackstone is also working with portfolio CEOs to change workflows using AI.

## Transcript

Inside Blackstone, hosted by Christine Anderson, with Eric Schmidt.

From the transcript Inside Blackstone publishes with the episode.

[0:00](https://realalpha.app/v/PSbUom-p1hU/?t=0) One more time with a little more energy. All right, more energy. Hello and welcome to this week’s epi… That was more energy, I felt it. You felt it? Go with it. All right. Hello and welcome to this week's episode of Inside Blackstone, where we bring together data and insights from across our portfolio to give you a fuller picture of what's happening on the ground in the real economy. I'm Christine Anderson, Global Head of Corporate Affairs, and here's the lineup. Our Monday Morning Meeting guest is my friend Gilles Dellaert, who leads our Credit & Insurance business. Then it's over to the Economic Weather Report with Winfield Sickles, and next my wide-ranging conversation with Eric Schmidt, CEO & Chair of Relativity Space, and former CEO of Google.

[0:59](https://realalpha.app/v/PSbUom-p1hU/?t=59) And finally, a debrief with Jas Khaira, the Head of our AI investing platform, N1. Gilles, thank you so much for joining us. Thank you for having me. You and I spent a lot of time together at the beginning of this year when private credit was just in the news constantly driving the headlines. It seems to have quieted down quite a bit. What are you seeing? Yeah, the headlines were loud and they were everywhere at the start of the year. I think things have definitely quieted down. You're seeing less of it. I think it's in large part because the performance has held up a lot better than the kind of doom predictions that were being made at the time.

[1:35](https://realalpha.app/v/PSbUom-p1hU/?t=95) Another dynamic playing out is AI. What are you seeing? Well, we're living through this unprecedented build-out across the globe. And with that come really large financing opportunities. The projects are substantial and they touch every part of the supply chain, if you will, all the way from the power that fuels the data centers to the equipment that goes in, the chips, the cooling, the services required and so we're active across all parts of that as a lender because a lot of financing dollars are required. Why chips and energy specifically? I would say power and compute right now are where things are most constrained and where things are most constrained, that's what we tend to run towards. Right.

[2:14](https://realalpha.app/v/PSbUom-p1hU/?t=134) Because that's where the supply-demand imbalance is largest. That's where it creates opportunities for us to lean in and provide private solutions to borrowers to help fuel the CapEx build-out and the growth that will come. And I heard you say yesterday, it's not just an AI story, there's something broader going on. Yeah, I think it's broader. I think the global economy is growing. With that come needs from companies to finance that growth and the CapEx that is associated with it, as well as the strategic M&A that might be associated with it. And we're active there with industrial companies, with companies that are in the aviation industry, the companies in the telecom industry.

[2:48](https://realalpha.app/v/PSbUom-p1hU/?t=168) So, it's a much broader playbook. It's global and it's sizable. And all of it has one thing in common, which is we're backed by real assets that offer some form of inflation protection, which we really like in this economic environment. Well, exciting time in your business. Thanks for joining us. Very much so. Really appreciate it, and now over to Winfield and the Economic Weather Report. Thanks, Christine. It's Tuesday, September 29th, and there is a lot going on with major central bank meetings, the US-China Summit, and, as everybody that was stuck in traffic knows, the United Nations General Assembly last week.

[3:25](https://realalpha.app/v/PSbUom-p1hU/?t=205) As I mentioned, when I was on air with Christine last week, the Fed recently delivered its first rate hike since 2023 with a unanimous vote. In his press conference, Chairman Warsh cited robust economic growth, competition for capital, and geopolitics as the driver of long-term rates. Treasury yields continued to move higher last week on concerns of prolonged Middle East conflict with the associated commodity inflation, as well as strong economic data and continued AI infrastructure-related borrowing. The five-year US Treasury Yield moved above 5% for the first time in nearly 20 years, and then globally, government bonds reached an average yield of 4%, which was a post-global financial crisis high.

[4:07](https://realalpha.app/v/PSbUom-p1hU/?t=247) Despite this volatility in treasuries, AI continues to underpin markets as adoption broadens. Meta shares jumped roughly 13% after launching its new AI agent, Muse, which already has millions of downloads and really highlights how AI is starting to move from the infrastructure build-out stage to real world usage. We're seeing that shift firsthand across multiple parts of the Blackstone ecosystem. Anthropic spend across our portfolio companies, borrowers, and GP Stakes PortCos grew approximately 27 times over the past 12 months.1 This dynamic is creating winners across the AI value chain, with AMD recently becoming the fourth US chipmaker to surpass a $1 trillion market cap.

[4:52](https://realalpha.app/v/PSbUom-p1hU/?t=292) Expanding use cases, this rising utilization and productivity gains continue to accelerate AI adoption and really reinforce this powerful investment cycle, as well as the deployment opportunity for so many of our businesses and funds. That’s it for this week’s Economic Weather Report. The week ahead brings jobs and inflation data and much more. Until next week, good luck out there. Now over to Christine’s conversation with Eric Schmidt. Eric Schmidt has had a front row seat to nearly every major technology shift of the last three decades. You know him from running Google and since then he's become one of the leading voices shaping the conversation around artificial intelligence.

[5:41](https://realalpha.app/v/PSbUom-p1hU/?t=341) He is CEO of Relativity Space, giving him yet another vantage point on how breakthrough technologies move from idea to reality. Eric, welcome to Inside Blackstone. Oh, thank you. Thank you very much. Well, of course, I love the founder and the firm and everything you guys are doing, so. You're a technologist, you're a computer scientist. You give these incredible technical talks. I think your superpower is that you're able to break down complex ideas and make them more accessible. And that's ultimately what we're trying to do here. Well, thank you. It’s amazing the breadth of things that you've done in your career, that you are still doing.

[6:16](https://realalpha.app/v/PSbUom-p1hU/?t=376) I'm not dead yet. You're not dead yet by any means. And I guess that's my question is, you're 71. When most people achieve the kind of success you have, or not even close, they spend the rest of their life sort of talking about it and what they accomplished. You keep reinventing yourself. I can't seem to stop.

[6:34](https://realalpha.app/v/PSbUom-p1hU/?t=394) **Eric Schmidt:** My own view is as you get older, you should take on more risk because you have less to lose, right?

[6:40](https://realalpha.app/v/PSbUom-p1hU/?t=400) Right. And so, you might as well try the things that you thought were impossible. And if you fail, well, it's okay. You know, you tried. And as long as you're trying in service of the ethics and the world that you care about. I care about democracy and I care about freedom, that's what I wanna work on. At this point in my own life, I'm interested in impact. Most of the problems in the world could be solved in Eric's little brain with an improvement or an adjustment in technology. If you look at sort of improvement in our lives, improvement in our health and so forth, it's all been related to technological progress.

[7:15](https://realalpha.app/v/PSbUom-p1hU/?t=435) Starting with the invention of fire, which was extremely useful, right, way back when. But also dangerous, right? Of course, and everything is dual use and everything has to be managed. But just look at the dollar value of labor over 1,000 years, or 500 years, or 100 years, right? Human output, human hours in terms of productivity, are vastly improved. And because we've somehow collectively decided to have fewer children, not a good thing. I have four. Congratulations. Doing my part. Have as many as possible. More children are good and more grandchildren are good and so forth. We're going to have to automate.

[7:53](https://realalpha.app/v/PSbUom-p1hU/?t=473) And so human wealth, human health, human wellbeing, one of the major determinants will be over the next 100 years, can we continue to accelerate progress in this? I profoundly believe we can. And I profoundly believe that long after I'm gone, humans will be much better. So, you tend to be way more optimistic on the benefits of AI than most. What gives you that confidence? Would you like most human diseases to be solved in the next 15 years? I would. I would. Sign me up, especially given my age. Would you like solutions to climate change? I would love that. Absolutely. Maybe we could have safer products.

[8:35](https://realalpha.app/v/PSbUom-p1hU/?t=515) Maybe we could have better educational solutions. Maybe we could change the educational system so that kids are not bored, but rather they're engaged. The opportunities are significant. In the last month, 10 major math problems were solved, including one called Navier-Stokes, which is historically important, by computers. Now you sit there and go, like, why do I care? Well, math is the foundation of that acceleration. For those listening to this podcast that are not technologists, what was the significance of that? Navier-Stokes is an equation that involves fluid flows. It’s one of the unsolvable math problems, until now. The specific conjecture had never been proven.

[9:17](https://realalpha.app/v/PSbUom-p1hU/?t=557) And so, they use approximations for Navier-Stokes. I indeed funded some physics work in this area… Oh wow. …which I never understood, but it was really smart people. And the key thing is that these fluid flows model, for example, lift of an airplane, how do air conditioners work, there are many, many examples. So, being able to solve the problem in a particular way indicates better algorithmic answers. We can get a better and more accurate answer. Why do you care? The airplane will fly faster. The airplane will use less gas. Right? I can go on. The rocket will go faster. The fuel will be less.

[9:56](https://realalpha.app/v/PSbUom-p1hU/?t=596) You'll get to Mars quicker. You know, what have you. All of these fluid flow things matter in a way that we sort of take for granted. Let's go back to the internet, because I think it's so interesting. You've been around the rise of the internet and other platform, you know, shifts, the rise at enterprise software, search, cloud computing, mobile, all of these major technological innovations. Why does this moment feel different to you? It's not different, it's just bigger. Bigger. The internet felt exactly the same as this. I started with mainframes, so that's how long I've been in this field.

[10:30](https://realalpha.app/v/PSbUom-p1hU/?t=630) I've doing tech for 55 years, and when the PC revolution came along, everyone said, oh my god, this is this huge wave. It created Microsoft and Apple and so on and so on. This was 40 years ago, 45 years ago. So, these things take a while, and I think what, another example. Let's look at self-driving cars. Did you arrive at work today in a self-driving car? I did not. I suspect not, nor did I. And so, what do we think about that? When do you think we will? Well, it's beginning. Now self-driving cars… When do you think New York City? New York City will be one of the last ones. Okay. But it'll come.

[11:08](https://realalpha.app/v/PSbUom-p1hU/?t=668) It’ll come. But it's coming in the different cities, including obviously Waymo and other competitors. But why do I mention self-driving cars? They were largely, they were designed in the 90s. The first real test of self-driving cars was 2004. So that's 22 years and going.

[11:26](https://realalpha.app/v/PSbUom-p1hU/?t=686) **Eric Schmidt:** So, the diffusion rate is different from the invention rate.

[11:30](https://realalpha.app/v/PSbUom-p1hU/?t=690) So, the faster the diffusion rate, the sort of broad adoption, the quicker the societal change. Why does it occur in what I do first? It's because we're already connected, right? It's essentially a softer problem. It's just connectivity. The marginal cost of connectivity, given we're so connected, is very low. The incremental cost of connecting you and me is essentially zero. Boom, it just happens. It's straightforward economics. Whereas getting self-driving cars is a great deal of capital. It's hard, and so forth. People are working on it. Why 20 years versus zero years, and that's the answer.

[12:05](https://realalpha.app/v/PSbUom-p1hU/?t=725) You see the same, by the way, in AI. All of the gains right now are occurring in what is called, essentially, scale-free. And scale-free means that you can just keep doing it and get smarter. So, in math, I'm not a mathematician, but as best I can understand, they can just invent stuff, and they just keep inventing things, and they keep inventing things, and they're so clever. Well, now a computer can take all of those ideas, and if you just give it enough juice, it can do the same thing. Because it doesn't need data, it's self-contained. Unlocking all these solutions. Unlocking these solutions.

[12:38](https://realalpha.app/v/PSbUom-p1hU/?t=758) In software, it turns out that once you start writing code and you start learning how well the code works, you can make it better. And then you can make it better and better. And there's what is called a recursive self-improvement loop, where it's getting smarter on its own. But this progress that I'm describing is happening faster than I have ever seen, right? What is true is each wave is bigger, but it's also faster. And we are not ready as a society. Eric, you gave me like 10 things to unpack there, but let's try to do this. So recursive self-improvement, right? This is the thing that had the Anthropic researchers, two of them, making comments, one leaving making comments, the other sort of validating that there are serious concerns about the speed with which AI is learning. Right?

[13:25](https://realalpha.app/v/PSbUom-p1hU/?t=805) So, by the way, every month there is a group of people who quit over this issue. Over the same topic, exactly. Silicon Valley, and in particular San Francisco, they're absolutely convinced that this is going to happen so quickly that it will threaten humanity in one way or the other. I personally do not agree with this. Okay, why? Well, let me explain their argument. It's a relatively straightforward argument. As humans, we learn at a certain rate. If you have enough hardware and enough electricity and so forth and enough scale-free learning, which is what I'm talking about, it can just get faster and learn faster than humans can.

[14:04](https://realalpha.app/v/PSbUom-p1hU/?t=844) And the way it works, I'll give you a simple example. I ask it to write a chapter and then it writes a chapter and I feed that chapter and say write the next chapter and so forth. Well, if I keep iterating like that, eventually it learns what I wanted with a process called reinforcement learning. It's easier to write the twelfth chapter than the first. And it gets smarter. And once it learns how to do it, it knows. It's like the way we learn. Right. But it can learn a lot faster. So, think of it as humans versus computers.

[14:38](https://realalpha.app/v/PSbUom-p1hU/?t=878) **Eric Schmidt:** I think the actual technical answer about this learning is it's not human learning, it's a different kind of intelligence.

[14:46](https://realalpha.app/v/PSbUom-p1hU/?t=886) And so, the best way to state the fear is that we're inventing a different kind of intelligence that we cannot necessarily constrain. This is called the alignment problem. Also, in the past weeks, there was a very famous situation where there was an internal test going on at OpenAI. And this internal test, which was trying to check something, managed to create a set of bots, which are agents, if you will, they managed to collaborate to violate certain norms and laws, they managed to break into things, and they managed to, allegedly, do a little bit of harm. Now, nobody was hurt. But it's an example of what is possible.

[15:26](https://realalpha.app/v/PSbUom-p1hU/?t=926) So, we must develop both recursive self-improvement which is the self-learning along with solving the alignment problem to human values. So, a simple way would be, I say to you, learn everything and you're a computer. That's not what I should say. I should learn everything, but, don't harm any humans, follow all laws, follow all laws, follow the US Constitution, make sure you don't get sued. You know, in other words, the set of assumptions. That you build into the model. These systems don't have a fear over the police. They don't feel feelings. And so, they have to be constrained. Come back to your optimism.

[16:04](https://realalpha.app/v/PSbUom-p1hU/?t=964) It's in everything that you've written and you've said. What gives you optimism that we will be able to sort out a future where humanity wins? What's the alternative? I mean, the best way to get a headline is to say, in 10 years, we're all dead. I was on a panel with a guy who wrote a book which is entitled, ‘If We Invent Superintelligence, We're All Dead.’ It's very hard to argue with someone in that. The only problem with his arguments is it’s wrong, right? Because part of superintelligence is getting the super alignment problem solved. I would guess that your social network might be a little bit different than mine.

[16:45](https://realalpha.app/v/PSbUom-p1hU/?t=1005) What's the buzziest thing that you were hearing about this summer in the AI circles?

[16:49](https://realalpha.app/v/PSbUom-p1hU/?t=1009) **Eric Schmidt:** The biggest and most important thing that has occurred in the last year is the arrival of long reasoning, deep reasoning.

[16:57](https://realalpha.app/v/PSbUom-p1hU/?t=1017) And when I say long reasoning, I mean eight hours of reasoning. So, what would happen is that you would start a year ago or two and it could kind of begin thinking and then it would start thinking about something else. Get distracted, like I do? We can make it feel like human, but somehow it takes a wrong path. It decides that instead of talking about math, we're going to talk about French literature. That's hilarious. Something which doesn't make any sense right? Right. Right. That problem was solved using essentially technical guardrails for what it kept thinking, so they said stay focused, stay focused, stay focused, is the way to understand it and that produced the ability to ask very, they're called chain of thought trains, where you can basically see its thinking how it solved a problem and these chains can be a thousand steps, two thousand steps.

[17:50](https://realalpha.app/v/PSbUom-p1hU/?t=1070) Now why am I mentioning this? I don't know about you, I can't think a thousand steps. I just don't, I'm just not that smart, but it can. So, it looks like deep reasoning will allow you to see deeper, as a human, than a human. The best AI is in service of humans doing things that humans are not so good at. So artificial superintelligence, how will we know we've achieved it and what do you think the timeline is? So, most people believe, right now, that we're at the beginning of superintelligence, which is broadly defined as humans plus super smart computers working together, right? There are more technical definitions.

[18:31](https://realalpha.app/v/PSbUom-p1hU/?t=1111) Now today, vision is better, data storage is better, retrieval is better and analytical stuff should be better and proofs are getting better. So, we're at the beginning of that. What is the thing that is not obvious how to do? And the answer is real discovery. So, the test that's used in the industry is, could you take all the information that's known in 1902, put it into a computer, and have it generate general relativity, special relativity, Einstein, 1902-1904? The consensus right now is no. And the consensus is no because it required a leap that was not obvious from any knowledge that existed at the time, okay?

[19:19](https://realalpha.app/v/PSbUom-p1hU/?t=1159) So, in the industry, we say, well, how would we solve that? Well, there's an obvious way, which is you take the computer and you just say, keep trying, you know, just it's the monkey with the keyboard eventually produces Mozart, but it takes a long time. I don't think that's gonna work because it's computationally not realistic. A better idea is to invent something which allows for analytical reasoning in non-coherent spaces. So, think of it as a leap. I happen to understand something in physics, and then you give me a completely unrelated problem, and somehow… Something clicks. …something clicks.

[19:57](https://realalpha.app/v/PSbUom-p1hU/?t=1197) You know, it's like in my case, in the morning you're in the shower, and you have this idea… Right. …that just showed up. Can we invent that? So, do you think these leaps are coming faster than we realize? The kind of leaps I'm talking to have not occurred yet. There's something called the San Francisco Consensus, by me anyway, and everyone in San Francisco thinks this will occur in two years. I don't. For one thing, I don't think there's enough computers, I don't think there's enough people, I don 't think that there's enough algorithms. I think maybe in my lifetime, maybe in a decade. That would be a big deal.

[20:30](https://realalpha.app/v/PSbUom-p1hU/?t=1230) But each of these moments, they’re happening so fast now. And by the way, it's fast because of the internet. It's because of what is called basically scalable computing. The supercomputers that are being used, so typical model takes three or four months costs a hundred million dollars to train, takes a few months to test, right? Thank God for the American financial industry that would be willing to finance these things. When we're seeing the creation of an entirely new financing model for this industry, right? If you want to achieve the vision I am describing, you have to do it at a world level.

[21:07](https://realalpha.app/v/PSbUom-p1hU/?t=1267) America is particularly good at building world standards and world-changing impacts at this scale. You're going to have to it in the US for regulatory reasons, and you're going have to find a way to finance it. You're also gonna have to wait to find a way solve the energy problem. The best energy solution is fusion. They look like they're going to get to fusion within two to three years. When that occurs, that's another massive moment in human history. I think it will occur in you know, three to five years. You've said you've sort of invested in everything when it comes to AI, but if you could only invest in one thing related to artificial intelligence, like one theme, not a single company, but one theme what would it be? Long reasoning.

[21:48](https://realalpha.app/v/PSbUom-p1hU/?t=1308) I don't think we appreciate when we have a thought partner that can think longer than we do. We can't think about how. So I'll give you an example. You're a CEO. You have 24-hour agents. You have one that's running, looking at security. Another one that's looking at power. Another one that’s looking at cash. Another one that’s looking at products. Another one that’s looking at global macro. And they're just running. And then your AI CEO every day when you get up says, does Eric need to know anything? Does Christine need to know anything? And they say, nah, nothing special, but it's always running.

[22:23](https://realalpha.app/v/PSbUom-p1hU/?t=1343) So, you wanna think about that in running a business, that those things are, you have to sleep, it doesn't. They can be disciplined into focus on this and not something else. Is that important? Absolutely. So, do you think there's a productivity boom that's coming? Well, I think it's already underway. It's already underway. An example would be my entire life, according to myself, I was a very, very good programmer when I was 22, right? I was one of the best, according to myself. My field is over, right? And yet you still tell people to learn coding. I do because I want them to be architects. So, I went from being able, I had to build the house.

[23:04](https://realalpha.app/v/PSbUom-p1hU/?t=1384) **Eric Schmidt:** Now I can architect the house using the equivalent of Claude Code or OpenAI Codex or future products from Google,

[23:12](https://realalpha.app/v/PSbUom-p1hU/?t=1392) which they're working on, all of which allow me to say what I want and the code gets written. I look at the code it writes now. I can't write that code. I couldn't when I was 22, right? So, I got beaten, right? Hard. It's okay, right, I did fine. The next generation of people like me will be the architects and they'll be very, very smart and they know how to say, this is what I want. One of my friends, some of my friends are not so friendly and they're not so social, they're kind of, so we say asocial. No naming names. I'm not gonna name anybody. So, I said, what are you doing? And he said, I'm deleting names from my address book.

[23:52](https://realalpha.app/v/PSbUom-p1hU/?t=1432) And I said, why? I don't wanna talk to them anymore. I said, what do you do all day? Well, I get up in the morning, I have coffee, and then I check with my friends. And I say which friends? He said, my agent friends. Oh God. Okay, so, what do you do? And he spends all morning plugging and playing with his agents who are solving really interesting problems to him. And then he gives them a one-hour project for lunch. He has lunch and he comes back and he interacts with his agents. Now he's got a wife and a family and so forth. So, at five o'clock he gives it an overnight, you know, task. Right. And it goes to work. And it goes to work, and then it typically completes between two and four in the morning.

[24:33](https://realalpha.app/v/PSbUom-p1hU/?t=1473) And I said, do you wake up for that? He said, no, no, no, no, I wake up naturally. I know, with certainty, when I wake up, I'll have a result. I was with you on everything other than that these are his friends, right? I mean, I think they're amazing productivity tools, but I'm going to take us back now to data centers. The case for data centers was made long before artificial intelligence, for mobile, and cloud computing, and cybersecurity. All these things that it helps enable, yet now there's so much concern around data centers, what's your take? Well, if you're running one of these AI companies,

[25:07](https://realalpha.app/v/PSbUom-p1hU/?t=1507) **Eric Schmidt:** your revenue is completely determined by your data centers.

[25:10](https://realalpha.app/v/PSbUom-p1hU/?t=1510) So, this is a massive change in what I do. It used to be that software was pretty high gross margin, you could sort of sell it, didn't have a lot of capital. Because you need the hardware, that's what's changed the economics. So, in the short term, it's the same thing as in the gold rush, the people who made all the money were the pickaxes and so forth and so on, the infrastructure. Picks and shovels, we have a whole thing on this, Eric. And this picks and shovels thing is real. So, from the standpoint of the economy, those are huge, they're real businesses. There's an estimate that 11% of US electricity demand in 2030 will be AI data-center-related.

[25:51](https://realalpha.app/v/PSbUom-p1hU/?t=1551) That's an extraordinary transition by any measure. Historically it was 2% or 3%. So, and that build-out is one of the primary drivers of economic growth in America today. So, thank goodness for that. If you could leave CEOs with one takeaway today, what would it be? The term that I've been using is being an AI-native CEO. So, what I'm trying to do in all my things is I'm tying to fully automate the execution of the businesses. In the factory floor in Relativity, I was walking through it one day and I thought, how many computers are here? So, I said, why don't we connect them all? So, they just connected all the computers.

[26:27](https://realalpha.app/v/PSbUom-p1hU/?t=1587) And so, the immediate thing that they can do, using AI, is they can look at, sort of, utilization, cross utilization times in new ways. It occurs as a natural byproduct of getting the digital systems connected. Furthermore, if you take generic data, unstructured data, and you connect it all, these systems are smart enough that they can interpolate between one data normalcy and another. In other words, they can combine disparate datasets and give you business insights. So how do I make more money? How do I increase customer retention? How do make sure I don't lose a customer? You can ask these questions.

[27:02](https://realalpha.app/v/PSbUom-p1hU/?t=1622) What I've been struck by is that when you take open-source data, that is public data, plus the proprietary data in your business, and you put it into a system, and you start asking questions, I go crazy over it. You can now talk across all of the data. You can get data fusion. You can actually figure stuff out. It drives me crazy a little bit, though, when I do have content given to me that is generated by the computer. Is it correct? Some is and some isn't. Sometimes I worry about the loss of original human thought. There's a whole bunch of concerns. I think the most one, the one that I've always been worried about, is the loss of deep reading, which I blame not on AI, but on social media.

[27:48](https://realalpha.app/v/PSbUom-p1hU/?t=1668) I used to read a book a week and now I'm too interrupt driven to do that. And my attention span is so much shorter. So much less. I've been watching young scientists, I fund a whole a bunch of young scientists. How do they do it? And the answer is they turn off the phone, they turn off, and they just do it, but it requires this huge strength to turn off the drug. Right? That interrupt drug, the serotonin thing that we have. Right. That's bad, in my view, that's bad for society. So, I have a bunch of young kids. I have two that are nearing their college years. What would you tell young people to study today?

[28:25](https://realalpha.app/v/PSbUom-p1hU/?t=1705) **Eric Schmidt:** I used to say biology, now I say deep reasoning.

[28:28](https://realalpha.app/v/PSbUom-p1hU/?t=1708) Okay. If you're a non-technical person, you should figure out how to use these tools to make your dreams and your realities extraordinarily scaled. You want to be a global star, a global influencer, a global impactor, a global discoverer, a global singer, you want to use these tools, whatever it is that you want. Figure out a way to use them to amplify you and what you care about and your innate goodness. If you're a technical person, use the same tools to invent stuff and to invent stuff that changes the world. I've never seen the cost of entry to be so low and the availability of these ideas so great.

[29:15](https://realalpha.app/v/PSbUom-p1hU/?t=1755) The only thing that limits you is your curiosity, your willingness to take risks and so forth. So, get over it. And say, I want to dream, I want to use these tools to have this enormous impact, right? And some of you, not everybody, will have a huge impact. And it won't just be coming, being financially successful, you might become famous or important in something that you didn't even know was important, but you'll be so proud of yourself. I love the optimistic future that you've painted, and all that you taught us today. I really appreciate it. Thank you so, so much. Thank you. Joining us for The Debrief is Jas Khaira, our Head of N1, Blackstone's AI investing platform.

[30:00](https://realalpha.app/v/PSbUom-p1hU/?t=1800) Jas, you were the first person I thought of when I knew we were sitting down with Eric Schmidt. Thank you so much for joining us. Thank you, Christine. Alright, so, very wide-ranging conversation with Eric. He knows so much. No surprise, AI was front and center of that conversation. Yeah, you went from math algorithms to Hugging Face. It was definitely a broad discussion. I think Eric is an optimist, and we should all be bullish on optimism as an investment trend. But I think he also brought up what happened with Hugging Face and some of the challenges that we’re seeing in AI safety and ethics that is resulting in many of the leaders of these AI companies focusing on that.

[30:36](https://realalpha.app/v/PSbUom-p1hU/?t=1836) There’s a good discussion and debate that’s happening right now. I loved Eric’s optimism. It was infectious. And it reminded me that if there is one long-term trend to be long, it’s optimism. I would say this idea of superintelligence and what it means, right, is something that everyone is still grappling with. What is your take on what superintelligence means for humanity? I think the current way to understand superintelligence is that the cutting-edge research within the labs intends to bring recursive work to the training itself. So, the training of the models now relies on compute in one virtuous circle to kind of enhance that.

[31:16](https://realalpha.app/v/PSbUom-p1hU/?t=1876) And you can imagine when you take that to its logical conclusion, you have superintelligence above that. I think we’ll have to see how that plays out in the real world. And yet it feels like the pace is really picking up, for sure. Things move faster than ever, and it’s one of the most dizzying parts about investing in this landscape, which is you really have to try to stay at the cutting edge of the frontier. Eric was also talking about how you need to stay an AI native. It's hard. And we're working actively with many of our CEOs to make them AI native and to bring workflows then change them with AI to help make our companies more efficient, improve them, have better quality, better products.

[31:53](https://realalpha.app/v/PSbUom-p1hU/?t=1913) Eric's doing that on the factory floor with Relativity. We're trying to do that across our 280 companies. Eric, you, and I were both talking about this. He had this line about, you know, if you're running one of these AI companies, and your revenue is completely determined by your data centers. We're certainly seeing at Blackstone massive data center demand that's just almost unmet. If one is able to generate $50 to $70 million of revenue per megawatt and your costs are $12 to $15 all-in, including the chips depreciating over a shorter cycle and in the data center over a longer cycle, well, then what you should be doing is using all of your capital to purchase the next megawatt to serve that.

[32:32](https://realalpha.app/v/PSbUom-p1hU/?t=1952) And that's what's happening in the market. People are taking today's cash to purchase tomorrow's AI factory. That's great. Thank you so much for joining us, Jas. I appreciate it. Thank you, Christine. And thank you for joining us on this week's episode of Inside Blackstone. Follow us wherever you get your podcasts.