What will the next three years look like?
I have a hunch that it will be something architecturally different.
He is Steve Jurvetson, one of the earliest investors in SpaceX when almost no one believed in the private space sector. He backed Tesla before electric cars became popular. For over 30 years, he has been betting on the future, and history keeps proving him right.
When someone comes to you with just an idea, what would be the best thing they could do?
A single person with an idea—I would try to find a co-founder. Rarely does an individual... Jobs and Wozniak, Batman and Robin.
You work a lot with Elon. What are the three main lessons everyone should learn from him?
I do try to observe leaders in action, even with deliberate effort. It's not always obvious, but there are a few things. One is an incredible ability to...
Thank you very much. This is going to be very exciting, Steve. I'm so excited to have you on this stage. What an exciting time, right? SpaceX, right? Public offering, and you were there from the very beginning. What did you see that most investors didn't see back then?
The simple answer to the question is that there were almost no investors considering the space sector. It wasn't a category anywhere. So the slightly different question is, why would we invest in a sector that simply isn't a venture capital sector?
The same could be said about the automotive industry like Tesla, or energy with nuclear fusion. There were, you know, a handful of investments, but very few. So in short, obviously, he's an incredible entrepreneur, someone we had already worked with before. I've known him for, gosh, 29 years now, and I invested in all of his companies of the century and also in those of his cousins, so to speak, I'm all in.
The uniqueness of the opportunity—so what we've come to appreciate in a somewhat fuzzy way back then, but now more clearly, is the way a software-centric systems engineering approach can transform a sleepy industry that hasn't changed in decades. It can really generate incredible value and opportunities. You can see it in the aerospace industry, now you can see it in the automotive industry. It was like a long-term bet when we first invested, but now we can see in retrospect how that will play out across almost every industry over time, how they become information-based businesses.
Obviously, you are very good at predicting the future, and as part of this podcast, I really want to understand how you think about the future. You have this incredible chart, 130 years of computing, and basically it grows exponentially. What does that mean for all of us? What will the next three years look like because of what's happening with computing?
Now, I'm just curious: how many people have seen this version or this abstraction of Moore's Law? It was originally presented by Ray Kurzweil in his 1999 book, *The Age of Spiritual Machines*. It looks like about 25% of the room. Good. I always ask because I'm curious to know how much this has entered the zeitgeist, because I think it's the single most important thing that has been incorporated, and I give Kurzweil credit for spotting this pattern when nobody knew they fit on a curve. So, just for those who don't know, this spans like five different technological substrates, from mechanical devices to relay-based computers, discrete transistors, and integrated circuits. And only in the most recent era, what Gordon Moore called Moore's Law would be almost a refraction of a much longer-term trend that transcends, you know, all kinds of company dramas that came and went. It's almost cosmological. Why has humanity's computing capacity multiplied for 130 years? And to get a sense of scale, that's an exponential scale, right? Logarithmic scale, so a straight line is exponential. This chart shows a 10-trillion-fold improvement in the computing power that can be purchased.
Nobody buys transistors when they buy integrated circuits. They don't say, 'How many transistors does that have? I'm going to buy the one with the most.' No. They buy computing capacity or memory. And both have followed a steady trajectory. And so, answering your question, the first and most important thing would simply be to predict that this will continue for three more years. I mean, why would it suddenly stop and hit a brick wall, like Intel has been saying would happen? And when companies say that, like Intel, it's usually a sign that they are losing their business to someone new, like Nvidia 15 years ago. So in the next three years, I think you'll see analog chips continue to carry the banner of Moore's Law, and custom AI silicon doing discrete tasks—matrix multiplication and addition very efficiently. And this is what is going to drive the juggernaut that we all take for granted keeps advancing.
In fact, I would say that without this kind of exponential shift in technology, startups wouldn't exist. You wouldn't have this opportunity for disruptive innovation like what we mentioned at SpaceX or at many other companies, because if the business is predictable, if there is no disruptive technological shift, the big guys just get bigger. You know, they have all kinds of ways to prevent new competitors from entering the market. Usually someone reinvents an industry, and it's almost always based on something computational. I think artificial intelligence and everything we're talking about at today's conference is the epitome of this. It's like the most intense crucible of computation-centric innovation, economic growth, and in a way, economic innovation—the transformation of what were previously low-margin, unattractive industrial businesses into information-centric, information-age businesses. Over the next three years, that means expanding into, I think, energy, agriculture, construction—three industries that are enormous, growing as a percentage of GDP, and are the least digitized industries on the planet, not to mention healthcare, right, which comes right behind that.
Do you think those are the industries where we will see the most change? And what will drive that change? Will it be more advanced LLMs, or do you think there's something else? I don't know, people are building world models. People are very into robotics. What will be the technological engine that will drive the most change over the next three years?
That's a great question because it's very hard to answer with certainty. I have a hunch that it will be something architecturally different. It could encompass the models we know today. You could think of something like a mixture of experts that encompasses other architectures, or the diffusion model we heard about earlier today, which ultimately translates into a transformer, but it's a different way of thinking about the transformer—a massively parallel form of the diffusion model.
And in the back of my mind—we haven't invested in this, so what I'm about to share, we haven't invested in this—I've met with some companies, and I've been intrigued, and my intuition tells me they are probably going to achieve a major breakthrough. And this is a whole new generation of neo-labs focused on reinforcement learning, because we're almost returning to DeepMind's original premise, which they then kind of set aside for a while when the whole LLM thing took off. So if you could imagine what it would be, you could express it in agentic terms and say, what is the agentic process that lasts several decades—not minutes or hours, not driven by some outsider pulling strings like a puppeteer, but something that is almost like the evolutionary drive for creatures, or for humanity, for whatever we consider the mission statement of our lives or of humanity in general? What will that thing be? Is it understanding the universe, as Grok and xAI say? That becomes an engine for an artist like that.
It's something like a novelty-seeking algorithm that says, 'I'm going to keep learning about the world and use novelty as my filter to say, I just discovered something new.' How do I know if I'm making progress? What is, if you will, the selection pressure in an evolutionary algorithm? What is success? It's not just reproductive fitness in the biological sense; it's something bigger. And I know some of these groups are working on what that is. Is there a single reinforcement learning algorithm with continuous learning, unleashed in the real world with all the internet datasets, that can generate intelligence on its own in that sense and in the way we think we're seeing it in large language models today? But to a large extent, I think we attribute consciousness to other beings. We attribute meaning to other things, and we see patterns where there are none. And that's why I think a lot of this is a fun interaction, but it's not quite the same, right? We simply know there is nothing in there; there is no light turned on inside, so to speak.
So what you're describing, I think, is superintelligence—when it is learning by itself, setting its own goals. Are we going to see some version of that in the coming years? I know Jack Clark, co-founder of Anthropic, gives it a 30% chance of happening next year, which I think is...
Yes, yes, absolutely. So I thought, well, that's kind of fun. At least there's one person staking out a position. I don't know.
I think they have a lot of strong opinions about what they do, but they also think they're on the right track. And there's a big debate about whether this recursive self-improvement thing—which they wrote about today and which Jack has been talking about for a few weeks, and I spoke with him about that last month or two months ago—will there be some leap that we currently don't see in how these systems acquire purpose and meaning, as I mentioned a moment ago? Because right now, everything they do is human-directed. It's like the AI self-improvement loop they are already witnessing—these huge improvements come from a series of steps that are still human-directed. There is automated verification, improvement cycles in the training process itself, tuning hyperparameters from one training run to the next. There are many ways you could imagine high-throughput experimentation being mediated by AIs, but what is the objective? The goals are still set by the human, and that may only be a small part of the activity it doesn't yet perform, but somehow it's the most important one, right?
And they will admit they aren't sure how that happens, right? What triggers that transition? And I don't know if it will need to reproduce some of the functional specialization in our own brains. For example, we evolved to where we are today with a history of reactive limbic systems and other emotional centers, which the cortex—and more and more cortex—then layered on top of. That whole construct may be, as we heard in an earlier talk, the starting point for consciousness as an awareness of what we perceive. Do we need to have those same things in our robotic or artificial intelligence systems, right? It could be. So it's a philosophical argument. The primary answer to your question would be: I don't know. And truthfully, I wouldn't even know what probability to assign to it. I put it in the category of an uncertain future. Yes, that could happen, but simply saying so is more convenient as an intellectual shortcut than truly thinking about it as a serious, difficult problem—it's deferring it. Three years feels far enough in the future that it's difficult to predict almost anything.
So we are seeing all the robot demonstrations, and current technology, I think, is more advanced than the actual implementation. We are still adopting, we are still adjusting. What is this gap? How big is it between what the technology is actually capable of doing and how we are using it?
All right. Yes, that's a very good point, and there will inherently be very different domains of adoption. Here's a great example, very easy to understand: if it involves the world of atoms, it takes time. So even though it's obvious today that fully autonomous vehicles are the inevitable future—that every car will be autonomous, every train, every plane, everything that moves on Earth will be fully autonomous in the future—how could it not be? It's crazy to think now, or argue, that it isn't so, even though we've been saying it for decades.
The pace of change is going to feel glacial in certain parts of the world, right? People keep cars on average for 11 to 12 years, so you just have the physical replacement cycle of cars. Mobility shifts don't happen overnight. That's an obvious one. Physical robotics might be the same. How long does it take to manufacture a billion robots? That takes time, even with recursive manufacturing techniques.
And so where I think this spreads like wildfire could be in areas that, curiously, we sometimes considered exclusively human: the creative arts—movie making, imagery, whatever we've already seen. In a way, it's surprising that that came first, and then office jobs, as mentioned, because the capacity of an office job, for example call centers, right? That's about 1% of US GDP. That happens just like that, right? I mean, you don't have to wait decades for that to change almost completely. And curiously, people will increasingly prefer these interactions over human ones when they are better, show more emotional understanding, and better understand the situation. And you see that in everything from the way doctors treat patients to chatbots or customer support agents, where AIs do a better job of emotionally connecting than humans.
Yes, it's crazy how in some industries this is happening super fast, especially in software engineering. Some of my friends were editing 70% of AI-written code a year ago; now it's only 30%. I wonder what it will look like in a year. So, you've worked with some of the most incredible entrepreneurs. You work a lot with Elon. I know many people in the audience are builders. Are there, perhaps, three main principles that everyone should learn from him?
It's funny. People have been sending me these books that I assume they asked an AI to write about AI, you know, about Elon, how he thinks, Elon's secrets. I've actually been stacking them on my nightstand, but I'm not sure a human wrote any of them. A lot of people have asked Elon's mom, you know, 'Hey, Maye, how did you raise Elon? How did you get him to be the way he is?' And that's a tough question; she hasn't been able to answer it either. So I'll take it with a bit of humility, because even as a close observer—by the way, I try to observe leaders in action. I worked with Steve Jobs briefly, briefly, and I put all my energy into trying to understand how that guy works. But even with focused effort, it's not always obvious; people are complex.
But there are a few things. One is this incredible ability to focus, which might seem ironic given how many companies he runs simultaneously, setting new records in doing so in a way that, you know, when Steve Jobs was CEO of two companies, that seemed strange, and now it's fashionable. But one thing it allows you to do is use the fact that you have clearly competing demands fighting for your attention as a way to focus, prioritize, and not go to meetings. So if a normal company CEO didn't go to the end-of-year holiday party, it might be seen as strange and like, 'Wow,' but nobody questions it if Elon has other things to do, has other companies. So whether it's an excuse or simply works that way, he very effectively rejects distractions that aren't critical to the mission right now.
For example, years ago I was trying to put him in touch with Craig Venter to brainstorm ideas about how we could terraform Mars more easily and bring back life samples from Mars with genetic sequencers and recreate microbes on Earth. It was a fascinating topic for me. I thought, 'Wow, this is so fascinating.' But he said, 'No, it doesn't matter until we have Starship flying. None of this about Mars matters. I have to make that thing work first before thinking about what we'll do when we get there.'
I think there is something perhaps more important than what I just said. Even more important is this maniacal focus on what I would generalize as the innovation cycle velocity, which is how fast we can run experiments or iterate on our learning cycle. What is the primary learning loop, whether it's launch cadence or data collected from all the Teslas before fully autonomous vehicles arrived that could be used to train the models? How can we ensure we have an advantage over anyone else in the speed with which we learn from customer interaction, product features, and technology overall? And as an example of how powerful that is when you do it right and have that data stream, you can accomplish a lot for artificial intelligence. Just as an example, Tesla cars today with their cameras collect more data for their AI training set every four days than Waymo has gathered in its entire history. And what was brilliant was enabling every vehicle, regardless of whether the customer paid for full self-driving or not, to be a data-gathering vehicle.
So focus, learning cycles, and a whole set of well-honed skills in identifying talent that I wish I could replicate—I simply can't. It's like sometimes there is pattern recognition, and he shares snippets of this, like not relying on credentials or specific backgrounds or experience. In fact, that's often a burden. Instead, it's having people explain in detail major engineering crises or how they solved problems, and then digging deeper and deeper to see if they really mastered and have the depth of understanding of what it took to make something successful. So broadly speaking, being a talent magnet, finding a way to present and hone a vision that makes people want to join you. One of the brilliant things he does at Tesla, SpaceX, everywhere, isn't just saying, 'Yeah, we're building rockets, we're building cars,' but really thinking about something much bigger, right? Catalyzing the transition to sustainable energy, making humanity multi-planetary, understanding the universe now that xAI has been integrated into that. These are the kinds of lofty goals that motivate some of the best and brightest to want to work with you. And that's a compounding benefit that ripples across the entire organization, because great people want to work with other great people.
Right. I'm talking to many entrepreneurs, and especially these days when everything moves so fast, there's some shiny new thing every week. How do you stay true to your mission when the rest of the world, 99% of the world, tells you it's too early? For instance, let's talk about space: 'We have so many problems here on Earth.'
That's an interesting question, and I realize I have a bit of sample selection bias on that. I've done the best I can. I've worked in venture capital for 30 years only with people who have a true, sincere, almost messianic mission in mind, which is what drives them. And they aren't arbitrage-seeking opportunists who just look at the next shiny object or, gosh, where should I go next? And one of the ways—and I'll answer your question in a moment—one of the ways I filter for that in meetings is, say I'm getting really excited about a company, I often ask, 'Okay, what does your business look like in 50 years?' And usually I get two reactions most of the time. One is usually a laugh, like, 'Well, what a ridiculous question.' You know, the arbitrage-seeking opportunist will say, 'By then I'll be on my third startup.' Like, 'What? How could I know what my startup will look like in 50 years?' They just laugh at the question. And then we rule those people out.
And the best is when the person feels so relieved, like, 'Thank God, now I can tell you what I've wanted to tell you all day, which is: this is what motivates me. It's something that is many steps ahead of what you'd probably want to invest in today.' Like, colonizing Mars is an uninvestable proposition. Go back to the founding days: when you start a business on day one and say, 'I'm going to colonize Mars,' do you know what follows? No. For most investors, that's not a door-opener. So most entrepreneurs who have that sincere, true vision have found a way to subjugate and postpone their real dreams and talk about something much more prosaic and short-term. So I think the answer would be, as an entrepreneur, that happens naturally, and you must try to find investors, partners, and of course employees who are with you for that long haul and who have a plausible path to get there.
So, you know, this is kind of the joint tension, I think, in the best startups that is difficult to satisfy simultaneously: having an audacious 50-to-500-year vision—this is what this company is going to do to the economy or the universe—coupled with, 'And by the way, over the next three years, we're going to iterate with real customers, learn from that, and be able to chart the path from where we are now to that shifting future.' Sometimes they chain backwards from the future to the present: to get there, what do I have to build now to then advance along that path? But it's not like going into a research lab, emerging in 20 years, and solving all the world's problems.
Yes, that's a really fascinating trait that I see in many great entrepreneurs. It's as if they are reverse engineering from 50 years out. Is there anything about these incredible entrepreneurs that still surprises you?
Well, I suppose it's a bit surprising in a way every time it goes incredibly well, and curiously, this might sound weird, I don't think I've ever thought about that question before or been asked it before. So the perpetual surprise for me is like, wow, in year 8, 9, 12, some new opportunity opens up and develops out of, in a sense, the expanding option value of pushing into some new frontier of the unknown.
So what I mean by this is that at our firm, Future Ventures, we try to invest in things that look like nothing we've ever seen before, but are related to what we already know. So ideally, it's a company that is literally one of a kind based on things we're accustomed to, whether it's artificial intelligence, synthetic biology, or whatever, but they are taking it in a new direction. So the window, as long as you have an agile mind and you're watching it, you think, 'Wow, nobody thought of that.' When we started, for example, when we first invested in Tesla, there was no concept of autonomous driving. It wasn't in the business plan, it wasn't talked about, it wasn't on anyone's mind. The way the electric powertrain uniquely enables that, along with control and precision, was fascinating. Or at SpaceX, the Starlink opportunity, you know. Yes, of course, when you lower the cost of launches so much, you can have...
...megaconstellations, but what would be the novelty that would make sense to do that we didn't do before? We didn't just invest in Planet Labs for Earth observation—yes, a constellation of telescopes—but this whole idea of building a network, you know, a backbone for the internet in the sky was, and then direct to cell, like each one of these things is a development, and then orbital data centers, right? They weren't even on the radar five years ago, so that continues to amaze me. In a way it's not easy, but it seems much more powerful as a business vector than intentional design, so to speak. This is almost like exploring the option space or the light cone, so to speak, of possibilities in an economy rather than planning something 10 years in advance and having it go according to plan, so to speak.
It's so fascinating how you've succeeded in so many different bets you made in the past, and they are so distinct from each other and in different industries. What are you betting on now? What should we be paying attention to?
