683. In the New Space Race, Who Makes the Rules?

Sponsor Introduction

0:01Freakonomics Radio is sponsored by Hotels.com. Hotels.com gives you access to hundreds of thousands of hotels worldwide. And when you join for free, you can save up to 20% instantly with member prices. Plus, every stay earns rewards. So when you book one trip, you're already lowering the cost of the next one. Hotels.com is simple, rewarding, and designed to help you get more out of every stay. Sign up for free and book today. Hotels.com. It's all in the name.

0:30Freakonomics Radio is sponsored by Amica. You know what they say. If you want to go fast, go alone. If you want to go far, go together. So go with Amica and get coverage from a mutual insurer that's built for their customers. One that looks after what's important to you together. Auto, home, life, and more coverage that fits your unique needs. Amica helps you protect what matters most. Visit Amica.com and get a quote today.

1:00Freakonomics Radio is sponsored by Mint Mobile. Ditch overpriced wireless with Mint Mobile. Cut your wireless bill to 15 bucks a month at mintmobile.com slash freak. Upfront payment of $45 for three months, $90 for six months or $180 for 12 months. Plan required. $15 per month equivalent. Taxes and fees extra. Initial plan term only. Greater than 50 gigabytes may slow when network is busy.

1:30Includes up to 20 gigabytes hotspot. Capable device required. Availability, speed, and coverage varies. See mintmobile.com.

Elon Musk and SpaceX

1:38A couple months ago, Elon Musk became the world's first trillionaire after his rocket company SpaceX had the largest IPO in history. Two weeks later, Musk lost his trillionaire status, at least temporarily. Why? Part of the answer is AI. Because SpaceX is also now an AI company, having gobbled up XAI, another Musk firm.

2:13And AI investments are, as you probably know, pretty volatile at this moment. But SpaceX, along with several other big players in the new space race, are still betting huge on AI, specifically putting AI in space. This would mean fewer power plants and data centers on our planet. It could also mean hundreds of thousands of new objects in the skies, increasing the risk of collision, environmental degradation, and perhaps much more.

Google's Space Ambitions

2:41In last week's episode, my Freakonomics friend and co-author Steve Levitt talked to some of the people involved in Google's attempt to move AI to space. I fully expect that by the year 2100, the great majority of energy used in the solar system will be going toward AI and will be both harvested and consumed off Earth. These AI in space projects will be complicated and very expensive. At the moment, most of the investment is coming not from governments, but from wealthy individuals and firms like Google and Musk and Amazon founder Jeff Bezos.

3:19You may think this is a modern phenomenon, but it is not. Wealthy investors have been financing space ambitions for a long time. He initially thinks about building a large pyramid for himself that he can be buried underneath like the pharaohs. Eventually, some astronomers get on to him and say, well, you know, sir, you could actually build the first ever mountaintop observatory.

Space Exploration History

3:41So today on Freakonomics Radio, where is this new space race headed and who's going to regulate it? Space is extremely dangerous. And you really don't know what the consequences of your actions in space will be. Steve Levitt is back in the host chair for part two of our series on relocating AI to space. And it starts now.

4:03Last episode, I spoke with people working on Project Suncatcher.

4:34Google's attempt to build AI data centers in space. That team is thinking about the future, how much processing power we'll need and how cheaply we'll be able to launch satellites. So this week, I wanted to ground the conversation in the present and the past.

MacDonald's Background

4:49So I started by reaching out to an economist. My name is Alex McDonald. I was the first chief economist at NASA, and now I'm a senior associate at the Center for Strategic and International Studies in Washington, D.C. So you were the first chief economist at NASA, and it wasn't like you just stumbled into that job. You were not a typical economist. You're a Ph.D. economist, and your research had been focused very squarely on space exploration.

5:19Now, I've never met another economist who thought about space before. Was space travel something that has been captivating you since you were a kid or just something that caught your attention as promising research area when you were in grad school? Well, it was a combination of both for me. I remember very distinctly the moment when I knew that I couldn't be an astronaut, and that was in grade four when I got glasses. Because back then, if you had suboptimal eyesight, you could not be an astronaut. I'd always had a passion for space.

5:49I was a big fan of science fiction growing up. I grew up in Canada, and I really always enjoyed just going out on the winter nights and looking up at the stars. When I was doing my master's degree in economics, a couple of things happened. One was the flight of Spaceship One, which was the first privately funded human spaceflight vehicle to go above the von Karman line, which is the internationally recognized definition of space, 100 kilometers up. And NASA, George W. Bush, had announced the plan to return to the moon under what was then called the Vision for Space Exploration.

6:22I remember very distinctly thinking, okay, if we're going to build a moon base, that is an economic development problem. It's a very distinct type of economic development problem, right? There isn't really farmable land, and you have to essentially import or produce at significant costs all of your air. But it's still an economic development problem. And so economists are going to be needed to make that happen. That was in 2005, and I basically switched my Ph.D. subject. So in your analysis, Alex, of the economics of space, you don't just focus on the last 60 or 70 years studying with Sputnik or the Apollo missions.

Historical Space Observatories

7:03You take a long view. So in particular, you've looked at the construction of astronomical observatories over the history of the U.S., and you've come to what I would say are really interesting conclusions. One thing that I found very surprising is how expensive some of these observatories were. They were massive investments. Yeah. As you know, there's different ways that you can do these cost adjustments from historical time periods. You can adjust by the share of the economy that these projects represented, and you can also just adjust for the cost of the primary input, which in space observatories in the 19th century and today is still primarily skilled labor.

7:39And so when you do those calculations, we find that projects like the Palomar Telescope, which was essentially funded by John D. Rockefeller's wealth, or the Mount Wilson Observatory, which was essentially funded by Carnegie's wealth, these are in the hundreds of millions to low billions of dollars, depending how you do that adjustment. My favorite single example is actually the Lick Observatory, and James Lick isn't very well known today, but in the 1870s, he was the richest man in California because he had bought up a lot of property prior to the San Francisco gold rush, and he's coming to the end of his life, and he decides that he wants to leave a legacy for himself.

8:22He initially thinks about building a large pyramid for himself that he can be buried underneath, like the pharaohs, and people convince him, hey, well, maybe not. Eventually, some astronomers get onto him and say, well, you know, sir, you could actually build the first ever mountaintop observatory. This would be a much more impactful astronomical observatory, to which Lick reportedly said, well, okay, but can I be buried underneath it? And the astronomers, being very clever, said, yes, of course, sir, if you pay for it, you can be buried wherever you want. And so to this day, James Lick's final resting place is underneath the main plinth of the primary telescope of the Lick Observatory.

8:58And that was another one of these billion-dollar-scale observatories because they had to build a whole road up the mountain. This was in the 1870s, and they had to import these lenses from Europe often. These were very complicated technologies. In fact, there's these great stories about the giant reflector lens that had been built by Corning out on the East Coast, largest ever built for the Palomar Observatory. And they had to then take it across the country by train. And people would line up to watch it because it was such a technological marvel of the day.

9:32And we kind of think that the Apollo program is such a high point of modernism and technology and public wonder and all these things. But actually, space has been doing that for hundreds of years. These observatories were always these large spectacles as well as institutions of science. And when you say they cost a billion dollars in modern terms, that's not too different than the cost of the kinds of things that NASA does. Now, not necessarily going to the moon, but other projects they do aren't really categorically more expensive than these observatories were.

10:06Exactly. That was, for me, the core insight for a modern, let's say, probe to Mars or the outer planets like Jupiter or Saturn. These are projects that will come in regularly in the half a billion dollars to a few billion dollars. So that means that there were these precedents of privately funded missions equivalent to the costs of modern space missions.

NASA's Economic Strategy

10:26Now, you were appointed as the first chief economist at NASA in 2019, and you served in that role for five years. It says a lot about how unimportant economists are to the world that NASA didn't think they needed a chief economist until 2019, don't you think? Well, there'd been a lot of economics work done, of course, at NASA. There was the famous economic impact studies that had been done during the Apollo program. And, of course, the space shuttle involved a pretty extensive economic analysis as part of the cost estimation process.

10:58That was the main way in which economics intersected with NASA's portfolio. It was predicting how much things would cost. But what started to change in the mid-2000s and the 2010s was that we started to see a lot of private investment. And once you start to see private investment on the scales of billions of dollars, you start to see that, well, you're going to need to intersect NASA's policies with economic strategy.

Economic Wins for NASA

11:25Now, are there specific situations you can remember where, by thinking like an economist, you're able to score huge wins for NASA? Yeah, I'd love to hear about a few of those. The human landing system was probably the single biggest change that I was involved with. In the Apollo program, we had essentially a government design for a whole system to land humans on the moon. These were built by contractors, right? The majority of the money still went out to contractors, but NASA was paying for it and was responsible for the system at the end of the day.

11:59That was true for the Artemis II mission, where NASA did the same thing for the space launch system and the Orion vehicle. But that will not be true for our actual landing of astronauts on the lunar surface because we took a fundamentally different approach with the human landing systems. And the difference was basically we said we were going to buy astronaut delivery to the lunar surface from commercial companies. We were going to put essentially fixed price milestone payments for the development of those systems and then the purchase of services afterwards and that we were going to compete that.

12:36And of course, now we have two of the richest people in the country whose space companies, SpaceX and Blue Origin, are competing to win those NASA contracts. That means that there is investor money now covering some of the upfront costs that NASA is not having to provide because they're looking to compete for those contracts in the future. So it's given here that space exploration and government funding of it is a sensible thing to do. But I've never quite understood what the government thinks it's optimizing in the space program.

13:10What do you think motivates politicians to fund these programs? Well, I love that your assumption is that there's an optimization function going on in government. I'm not quite sure that I would say that that's exactly what the process of government is. I think when you look at the origins of spaceflight, there's really two core, I'll say, state demands. The original and leading one was essentially national security. The ability to launch something around the Earth has foundational importance for military systems.

13:40So things like satellites have turned out to be super important. That was true in the mid-20th century. That continues to be true today. The very first maneuvers as part of the Iran war were in space. They were electronic interference and jamming activities that the U.S. undertook against Iranian satellite assets. So that really is a core and deeply enduring reason that started it all and continues. But shortly thereafter, there was this second function, which I tend to think of as a signaling function. And of course, Steve, you're well aware of signaling theory.

14:12I'm sure many of the listeners are too. But it's this idea that a costly action can credibly transmit information. We talk about this a lot in education. My favorite example is always if you see someone driving a Lamborghini, you don't necessarily know anything about them. But because you see them driving a Lamborghini, you know one of three things, right? They're either wealthy themselves, they have wealthy friends who get access to credit, or they're good at stealing things. And you know this because access to Lamborghinis is expensive. If you think about being someone in the world in 1957, when Sputnik launches into space, you're looking at the Soviet Union, you're looking at the U.S.

14:47If you're in Thailand or Ghana or somewhere in the world that these countries are trying to convince you to join the alliance, so to speak, you don't really have necessarily a lot of information related to what's the quality of life. This is the pre-jet age and a lot of people traveled to these countries and seen what it's really like. But if you know one thing, which is that one of these countries has launched something into space and the other has not, and you know what that takes, and you can verify it independently with radio telescopes and radio communications, then you know something meaningful about the technological and economic power of those countries.

15:19And the same thing, of course, applies to landing on the moon. And so space has become and continues to be a really important part of how leading nations signal to the world leadership and technology and economic organization, which, of course, is a proxy for many other things. I was surprised to hear that early on in the space race, John F. Kennedy offered to Khrushchev, the leader of the Soviet Union, the chance to work collaboratively to reach the moon. And if Kennedy hadn't been assassinated and Khrushchev wasn't so suspicious of Lyndon Johnson, maybe getting to the moon would have been a joint U.S.-Soviet mission.

15:58This seems at odds with a signaling story, doesn't it? I would argue that basically Kennedy is trying to signal he wanted a more collaborative future. He was very worried about a future nuclear conflict, and he said, hey, we need to figure out a way to collaborate here. So he tried to use space as the signaling device to encourage cooperation. At the same time, importantly enough, Congress actually passed a law that essentially said, thou shalt not use any of the money in this bill appropriated for NASA for any collaborative programs with the Soviet Union.

16:31And so Congress was well aware of the signaling good that they wanted the United States to put out there and achieve. So Congress was trying to signal to the developing world that the U.S. was the country to jump behind. And Kennedy was trying to avoid World War III by signaling to Khrushchev that he was a friendly type. Yeah, that's my interpretation.

16:53At the same time the U.S. and the USSR were racing to the moon, they were also negotiating rules to govern human activity in space. Even in space, the world is run by lawyers. Like this one. I'm Rosanna Hoffman. I'm the head of space law and policy and now also sustainability at the United Nations Office for Outer Space Affairs, or UNOSA, as we like to call it. We're really the UN's home for space. We're also the convener for space negotiations.

17:25So we help states negotiate everything around space, from space law to space technology to really everything from disaster management and on and on. I advise states, but also more and more industry on how to really implement space law, how to ensure that their space activities are done sustainably and safely. And I help states negotiate space law and the use of space. And that's what I've actually been doing the last two weeks. The Committee on the Peaceful Uses of Outer Space, which meets six weeks a year in Vienna, where I'm based, just concluded its meeting a few hours ago.

18:03It's so interesting to me to hear you talk in such a familiar way about these organizations that I've literally never heard of. It's the kind of world that I think the regular person can walk through and never know that what you do exists. That's true. We're a small office, but we exist since 1957. Some listeners might know the year. It's the first time a satellite was launched into outer space with Sputnik the year after. Of course, we had more satellites being launched into outer space.

18:34And it was that year, 1957, that states in the UN and the General Assembly in New York said, OK, space activities are becoming a reality.

18:46These matters, ensuring that space is used peacefully, safely, sustainably. We need to make sure that the wars currently ongoing here on Earth, the geopolitical tensions, this is in the middle of the Cold War, right? That we don't move that on over to space. It's surprising right in the middle of the Cold War that these countries were sitting down and talking in such a civil way. I agree, but I just come from two full weeks of negotiations between member states. And just yesterday, I sat in a room with the U.S., with the Russian Federation, with the Chinese, with the Ukrainians, with the Iranians, all in a room, civilly discussing space activities.

19:29Space is actually one of those domains where states continue to negotiate and discuss. They've always done so. We're so dependent on space, from navigation to Earth observation to our Internet connection, that if one state becomes a bad actor, so to say, we will all suffer the consequences, including the bad actor themselves. So it is that inherent need for space that brings us together to the table and allows for these discussions to continue.

20:03Although I have to say, I don't want to sound too optimistic here. There was some self-preservation behind the negotiations in the 50s and 60s. It was all about ensuring, because we weren't sure who would get to the Moon first, ensuring that whoever did would not suddenly have the rights to the Moon. And also ensuring that should we place weapons of mass destruction or nuclear weapons in outer space, we really didn't know what the consequences for Earth could be. Could that be the end of Earth as we know it?

20:34So it was really this protective spirit and self-preservation at the end of the day.

20:41Out of that era came five treaties. The foundational one is the Outer Space Treaty of 1967. It states that space belongs to everyone and that no country will place nuclear weapons in orbit. 118 countries ratified it, including every major space-faring nation. And then the treaty making slowed down. The last big treaty was ratified in 1984, even though there are a lot more people making use of space now than there were then, including private actors.

21:13I asked Hoffman why the law stopped keeping up. It's such a good question, and I get it so often. And every time I think of a bit of a different answer, because there is no one answer that fits to that question. The last treaty that was negotiated was the Moon Agreement. And even that treaty, some consider being a failed treaty, because it only has 17 ratifications. Compare the Outer Space Treaty with 118. So already there, even in the 80s, we already saw a steep decline in states willing to enter into treaties that would bind them with international obligations.

21:51Why did this happen? Generally, when we look at it historically, we see that states beginning in the 80s are less willing to enter into international binding treaties. There was a big push for it after the Second World War. You had the establishment of the United Nations. You had a lot of international agreements being negotiated. There was a strong willingness between states to enter into this multilateral discourse and bind themselves internationally. That completely ebbed out in the 80s and 90s.

22:24And then another point, which is specific to space, is that a lot of the topics that they needed discussing were very technical. I'm talking about space debris mitigation. I'm talking about space traffic coordination or space resources, for example. They're so technical in nature. The way we do those types of activities will change so much in a short time frame that a treaty isn't the right way of dealing with it. A treaty takes a lot of years to negotiate and is not easy to amend.

22:59So what happened after the 80s is that states decided to go into the direction of something called non-legally binding instruments. Guidelines, resolutions, principles, standards that dealt with more of the technical aspects of space and that could be rewritten and revised more quickly, more easily going forward. But, of course, it's not legally binding. So some would argue it does not have the same effect. Others would say it can if states implement it into national law.

23:33So I'm curious. So let's say China's got a satellite and for some reason it's maybe going to crash into SpaceX's satellite. Does someone in China call up someone at SpaceX and say, hey, we've got to do something about these orbits? Or how does that actually work? That's a really good example. And it happens more often than you might think or know. Currently, there is no global space traffic coordination mechanism. It's what the UN and its member states have been asked to work on since 2025, June.

24:04And we just finished a session of the committee where this was the main topic of negotiation. So what is happening for now is that the Chinese operator, through their governmental entity, would contact, for example, the State Department in the U.S. And that often takes quite a lot of time because we're talking about SpaceX. It's a commercial entity. When will that information of that close collision finally get to the operator's technician's room?

24:37And when will they be able to know whether they should move or not? I was on the way back from a mission I did in Ghana last June, helping Ghana draft their national space law. I had a stopover in Brussels, and my phone was being called by the Malaysian space regulator, saying that a satellite in low Earth orbit, one of their only satellites, an Earth observation satellite, a large piece of an object, is about to crash into a satellite owned by North Korea.

25:10And they don't have diplomatic ties with North Korea. They don't have a mechanism of talking with them, of informing them that there is this very near chance of collision. On top of it, the Malaysian satellite was non-maneuverable, so they couldn't move. So what we were asked to do, and we really only had a few hours to do it, was reaching out to the North Koreans, and luckily, really in the 11th hour, they were able to move the satellite, and collision was avoided. And this would have created amounts of debris that would have impacted space activities for a very long time.

25:45So we used this example, we shared it with member states, and we're talking, you know, UN member states, it's a lot of them, and said, you need to come together and come up with a mechanism. Everyone needs to have at least their 24-7 point of contact within the system, and not going through my phone, or right above me, my director's phone, and then that's it. You know, that's who we are. It's a small team.

Orbital Data Centers

26:12After the break, orbital data centers may sound like science fiction, but the history of spaceflight has been influenced by fiction from the beginning. He wrote in his diary about, after having read those stories, climbing a cherry tree one day, trying to trim some of its branches, and having a vision of a vehicle taking off from the valley below. And from that moment, he basically dedicates his life to the development of the technologies to make that possible. I'm Steve Levitt. You're listening to Freakonomics Radio, and we'll be right back.

27:10Healock Lender in the U.S. Helps tap equity you've built in your home, so you can do things like tackle a renovation or pay down debt. Healock is a line of credit secured by your home. You can draw what you need, repay, and draw again. Figure offers a 100% online application. Get approved in five minutes and funding in as few as five days. It's faster than a bank loan and cheaper than a personal loan. Check your rate today without impacting your credit at figure.com slash Freakonomics. This post is sponsored by Figure, NMLS ID 1717824.

27:46Figure is an equal housing lender. For more information about their licensing, visit nmlsconsumeraccess.org. Figure's Healock is subject to credit approval and terms apply. Back to my conversation with Alex MacDonald. He's a former NASA economist who studied the history of private spaceflight. I was shocked to learn reading your research that Robert Goddard, the father of modern rocketry,

28:16he got about half of his funding from private sources, especially from the Guggenheims. And that's despite the fact that what he was doing with rockets had such obvious military applications that I would have thought he would have been just swimming in government money devoted to defense. He did get a fair amount of money from defense. He was essentially developing his rockets in the 1930s at the exact same time that Wernher von Braun is developing his rockets in Germany. And so those two people are actually competing to be the first people to launch something into space.

28:47Goddard's primary funder is the Guggenheim family. Wernher von Braun's primary funder is the Wehrmacht, the German army. What happens is, interestingly enough, in the early 30s, they basically achieve an equivalent level of spaceflight, which is maybe a kilometer or a couple kilometers up. This is obviously a significant achievement for a liquid fuel rocket. But the responses of their funders are very different. The Guggenheim family says, that's great, keep going, but we're keeping the funding the same. Whereas, essentially, the German army commits to a billion-dollar development of an entire

29:20research facility in Penemunde, in north of Germany, and begins the development program of the V2. One of the reasons was actually because these were not ICBMs, right? They're really able to bombard only a few hundred kilometers away. For the U.S., you won't be able to do launches across the Atlantic. But for Germany, of course, there's a lot more targets nearby. And so the incentives for the two different militaries were fundamentally different. Goddard, however, is convinced that he needs military money. And so he does leave his Guggenheim perch, which he had a nice life out in New Mexico.

29:52And he ultimately goes and works for the U.S. military to develop jet assist to take off rockets and move to Annapolis. It was actually Charles Lindbergh who personally convinced the Guggenheims that this was the future of flight. There's this part of the story of Robert Goddard, this famous professor of rocketry and developing the first rockets, where his first flight in a plane is literally being flown back from the DuPont facilities by Charles Lindbergh himself. It's interesting hearing you talk because you don't talk like the typical economists. You like stories.

30:23You seem like a little bit of a romantic, which fits in with the idea that you wanted to be an astronaut. Do you think that the rich history of science fiction stories about people traveling to the moon or to Mars, do you think those have served an important role or do you see them as kind of secondary to the technology? I think they're actually quite foundational. You know, spaceflight is a very interesting economic outcome, let's say. As I think folks know, it's pretty hard to live out in space.

30:56That's the understatement of the year. Yeah. You know, as they say, Mars ain't the kind of place to raise a kid. What that means is that a huge amount of effort, a huge amount of investment, a huge amount of blood, sweat, and tears goes into creating the systems and surviving out there. What motivates us for that? There's a little bit of the achievement and public acclaim. That's a type of incentive. But actually, most people who do spaceflight work, including most astronauts, they're not really the public figures that they would have been in the Mercury program, the Apollo

31:27program, right? And so I don't know if that's really the main incentive. I think the real core incentive is an internal psychological one, which is that the story of humanity figuring out how to become sophisticated enough, advanced enough, maybe even benevolent enough to manage to go out into space together and extend the story of humanity throughout the solar system and in the far future, potentially even to other stars. I think that story is very motivating to people.

31:57Not everyone in the space industry necessarily believes in that story. There are good technical and scientific reasons to believe that that may not ultimately happen. But if you don't believe in that story, then you're not going to self-select into the community of people who works on it. What is so fascinating to me is that spaceflight is largely, in many ways, the result of enough people believing in that story that they dedicate their labor to making it happen, and it has worked. And for me, that's something that fundamentally, I'm not sure we've really fully accounted for

32:29within economics. I remember early on during my PhD, I became very convinced that this idea of an inelastic supply of labor was really important, that actually people would be dedicating their labor irrespective of the economic incentives, right? Because they might believe in a particular story sufficiently strongly that other incentives don't really matter. So you're essentially saying Robert Goddard didn't really care if he got paid for these rockets. He was just going to do it because he was born to do it. Whereas somebody like me, well, you know, I'm kind of willing to do whatever people

33:02pay me to do. And it is something that's more or less outside of our economic models, the idea that people just believe so deeply in something. Belief is a hard thing for us to model. Exactly. But it clearly motivates a large amount of labor, right? Goddard, he didn't just believe in the story randomly. He received it as a teenager. He read about it. He read The War of the Worlds, and he read the unauthorized sequel to The War of the Worlds called Edison's Conquest of Mars. He read that as a teenager. And he wrote in his diary about, after having read those stories, climbing a cherry tree

33:35one day, trying to trim some of its branches, and having a vision of a vehicle taking off from the valley below. And from that moment, he basically dedicates his life to the development of the technologies to make that possible. I mean, that's a powerful belief, right? And when I was adding it up, his belief, the similar belief and motivation that the original German pioneer of liquid-free rocketry, Hermann Oberth, had, and the same kind of figure in Russia, Konstantin Tsiolkovsky, they're basically dedicating, you know, on the order of 50 to 100

34:06years of high technical labor to a project independent of any economic incentives. And they lay the foundation for good-free rocketry. So, yeah, you know, if there's an appeal to economists out there, think about how to incorporate inelastic supplies of labor and the role of belief in the labor market and the economic systems, and we might start to see some new economic theories emerge.

34:32Belief may have been a big part of what got us into space, but space today has become big business, and the incentives are more purely economic. Satellites today are an enormous economic activity. Do you have a rough idea what the revenues are right now of satellite companies and other private companies operating in space relative to NASA's annual budget of about $25 billion? Yeah, that's a great point. The overall estimate of the size of the global space economy is roughly on the order of $500

35:06billion to $650 billion a year. So, 20 times, 25 times bigger than NASA's budget. That's right. And the vast majority of that, 75, 80% of it all, is simply telecommunications. Historically, this has been satellites in geosynchronous orbit around the Earth, so 36,000 kilometers away. These are the satellites that would be providing your satellite television. But increasingly, it's now these low-Earth orbit constellations, constellations like Starlink. They're bringing in revenues on the orders of tens of billions of dollars already, just

35:37for things like Starlink. So, there's almost two space industries. One is all the kind of stuff that seems interesting, like space exploration and human spaceflight and even space defense. But that's actually a relatively small portion of the economy relative to telecommunications, because moving data around in space is actually by far the largest space industry. I've been talking to some folks at Google who foresee a world in which a huge share of all the computation done by humanity would be done in space in the form of swarms of these solar-powered

36:11satellites, what they call Project Suncatcher, and they'd be acting as data centers for a future AI-driven world. What do you think the chances are that they or someone else will actually pull this off? Yeah, so orbital data centers is the hottest new topic for economic engineering debate within the space community. There are currently zero revenues, or maybe single-digit million-dollar research revenues at best, related to the use of orbital data centers. But the projections that some people are putting out there are very, very significant.

36:44And of course, this is tied to the projections overall of increases use of AI and the potential to put these data centers in space, not just because of potential benefits of solar power, but also, frankly, because you have very different regulatory regimes up there. You don't need to take up anyone's land to put it up there. I have no doubt that we're going to see a number of attempts to do so. The cost of trying to put one of those up there is really only in the orders of hundreds of millions to low billions of dollars to really see how long can these GPUs last, what are really the efficiencies, what are the costs of, and frankly, the scales of these data centers.

37:18You talked about that in a very matter-of-fact way. But most of the way you talk about space is more romantic and more intuitive, I would say. What's your intuition? I mean, you know, two economists making predictions about these things. I'm not making any predictions. I'm only asking you for one. Yeah, touche, touche. You're putting all the pressure on me. Look, I think we're going to see some of them, but I don't think the night sky is going to be radically transformed. I am moderately bullish on AI in general. I think the ultimate demand for orbital data centers will be limited, and there are going

37:52to be a lot of practical challenges with it. I do think we're going to see a number of them fielded, because I think there are going to be edge cases for data use, even just in space, where you're going to want to be storing and processing your data on orbit. I do think like many tech hype cycles, though, there are a lot of people eager to sell you axes and overalls for this particular hypothesized gold rush. And so you may want to be on the overalls and pickaxe version of this rather than the searching for gold one.

38:19Ultimately, there's a practical dimension to this question. AI space data centers are objects launched up into space that need to be regulated just like everything else. I asked Rosanna Hoffman from the UN how she would think about regulating AI data centers in space. The good thing is a data center is nothing else than a bunch of space objects, right? A lot of them, of course. So they, too, will have to adhere to the Outer Space Treaty. You'll have to register them. You will be liable for damages caused.

38:52They need to be authorized and supervised. They need to meet sustainability expectations, etc., etc. However, the number of satellites required for data centers will bring that question from emerging spacefaring nations, as we call them, or the global south or developing countries that say, sorry, but if you're planning hundreds of thousands of satellites in low Earth orbit, will there even be place for us?

39:23Where will we be able to put our satellites? So that's one question. And another question is space situational awareness or space traffic management data centers, because they would consist of so many satellites, there has to be some global mechanism on space traffic coordination. But then going a step further, it's just the broader notion of space sustainability. Even though we think space is infinite, it really isn't. And especially low Earth orbit, it's not, right?

39:54So we need to make sure that that area of space remains usable for future generations. We need to make sure that if we're planning hundreds of thousands of satellites, that they are able to deorbit when it's their end of life. Do we require, for example, attacks for every satellite launch that goes into a pot that then would be used for removing satellites or removing debris? Spoken like a true economist. That's exactly the sort of thing that economists would say we not only should have,

40:28but really need if you're dealing with this negative externality of space tank. Exactly. And these are discussions that are happening at the UN and member states are discussing it. But the question is, I ask myself this, not as a UN official, but as, you know, just Rosanna. Do we need a major disaster happening first until we see the need for something like this? We haven't really seen major damage happen on Earth. We haven't seen a cascading event of damage happening in outer space.

41:01So do we have to wait for that to happen? Or can states and commercial entities already foresee this need today?

Regulating AI in Space

41:14After the break, if AI goes to space, what will it learn? We train AI on all the beautiful deltas of the Earth and the forests of the Earth and the river systems and so on. It will care more about the Earth, the humans, the rest of life on the Earth. I'm Steve Levitt. You're listening to Freakonomics Radio. And we'll be right back. In this episode, we've heard some reasons that space might be a good place to put data centers.

41:58But there's another reason AI and space belong together, and it has nothing to do with electricity or real estate. It was explained to me by Will Marshall, who I spoke with on the last episode. He's the co-founder and CEO of Planet, the data company whose satellites photograph the entire Earth every day. Before NASA, I had done a PhD in physics. It was incredible to work with some geniuses and humbling, which helped set my career more into space, which I thought I could handle because these physicists are just out there, next level people.

42:29There's a lot of public uneasiness around AI and its growing importance in society. You've got a vision of a future version of AI that looks very, very different from our current models. Could you describe where you think AI is going to take us? I think there's a new era of machine intelligence that I call planetary intelligence. So let me do a little backdrop first. Everyone's familiar with large language models now, incredibly powerful. They've read all the books in the world and the text on the internet, can write an essay, can hallucinate, obviously, sometimes, but they can do code.

43:07I was testing some of those elements early on. I was inspired to find just how capable they were at advanced physics. I was like, I think I'm smarter on this specific area of physics that I did my PhD in. Let's see. And I was like, damn, that's not messing around. They have a lot of late knowledge and they can pull all that together from all the disparate fields like no human can. But for all this capability, LLMs, the chat GPTs and Geminis and what have you, are essentially blind. And what I mean by that is that they have no understanding of what's going on in the real world day to day.

43:41So I'll give you an example. A farmer starts asking about farming for their field in chat GPT. It's going to tell them all about the theory of agricultural science, about crop science and all this sort of stuff. But it's going to know nothing about that person's field today, what the conditions are and what they could do better. Well, there's all these physical world data sets. But perhaps one of the core things is Earth data. Landsat in 1972 started imaging the whole planet every month and has done for the last 50 years.

44:13Planet in the last 10 years has been imaging the whole Earth at three meter resolution every single day. Both those systems together have about 5,000 images of the whole landmass of the Earth documenting change over time. Imagine adding all of that to these LLMs so that they have all that knowledge of agronomy and glaciology, but they also have literal understanding of what's going on in space and time day to day. So then the farmer asks that question and they can say, well, your field's doing this. It's got blight in this corner.

44:44The next door neighbor farmer did this and they did better yield last year after they had done this intervention. Why don't you do that? And interestingly, all the AI leaders have been talking recently about physical models, trying to build AI to have understanding of the physical world. And of course, AI is only as good as its training data. And here we sit in the space community with a whole ton of data about the physical world, you know. And it's not just satellite data. It's sensors around the place. But Earth data is, if you like, the Wikipedia for this next phase.

45:15It's the foundation data set that will enable us to then answer questions about the physical world, about those fires, that farming, that insurance, that so on. And that is going to open up a huge new set of domains of applications. Why hasn't it happened already? I mean, interestingly, before the LLMs, I think people would have said that what we've done with computer vision was actually remarkable and had gotten way ahead of what the LLMs have done in the meantime. What's the holdup?

45:46Is it the models? It's not the data? Good question. I've been impatient with this for a while. In fact, I did a TED Talk in 2018 called Queryable Earth, where I was just talking about how we should, with computer vision and all this Earth data, be able to index everything on the Earth and make it searchable. But it was harder than I thought, because each different thing you'd want to monitor, trees, ships, planes, you'd have to build a bespoke model. It was a lot of training. And then it only worked in this area and didn't work in that area and all these things.

46:16Because LLMs have enabled them, it's more generic. And in fact, excitingly, this is not just hypothetical anymore. We're just about to launch a public beta app, which should be out by the time this podcast comes out, that enables anyone to go in, look at our images, and chat with all those images. So you can search the whole Earth for an object by text or by finding one and say, search for more, look like this, write me a report of all the agriculture in this region, or land use change over time over here, or shipping activity over there.

46:49And it can have a pretty sophisticated analysis of a whole set of imagery. I really hope that will unlock a lot of the latent value in all this Earth imagery data sets. Now, you've talked about this in a very practical way, but I've also heard you talk in a very wistful, pro-humanity way. Could you talk about the Fermi paradox and how you think this kind of planetary AI will fit into the future of humanity?

47:21We want to align AI with human interests and the rest of life on Earth. And AI alignment is a big challenge that we don't know how to solve yet. But I think a couple of things. Firstly, AI is going to be hard for it to learn well without sensors. A baby learns through interacting with the physical world by having sensors, eyes and ears and so on, and actuators, arms and legs and touches and so on. And interacting with the world, it learns, it becomes intelligent and ultimately self-aware and conscious.

47:53I think the same of AI. We want it to care about life on the Earth. Well, people tend to care more about things they know more about. My partner's into ornithology and as we've learned more about birds, their calls, their marks, their different types. You don't just look at, oh, there's some birds over there. I start going, oh, there's two robins, two what have you. And you care more about them intrinsically. So I hope that as we train AI on all the beautiful deltas of the Earth and the forests of the Earth and the river systems and so on, that it will care more about the Earth, the humans, the rest of life on the Earth.

48:32Now, that's very speculative. But I haven't seen a good thesis for how we ensure AI alignment. And I think that's part of the puzzle, potentially. That sounds like a really great science fiction story where you train the AI to value the beauty of Earth and humanness in order to protect our long term. You asked about the Fermi paradox. Look, fundamentally, the Fermi paradox is the universe is really large. There's lots of opportunity for life. Why aren't we seeing the aliens yet? And roughly speaking, I'll spare you the math, there's about a thousand billion billion Earth-like planets.

49:06That is, planets that look like the Earth in terms of geology, liquid water and all that. That's about a hundred billion Earth-like planets for every human on the Earth. It's just a vast, vast number. We've looked at loads and found lots of planets around nearby star systems. And none of them look as good as the Earth. None of them have found any signs of life yet. At the minute, we either know we're either very rare or alone. And either way, the incredible, interconnected, complex web of life that we have on the Earth is incredibly precious.

49:43One of the answers to the Fermi paradox of why haven't we seen life is simply, do we get technological and then blow ourselves up? Or any species, when it becomes technological, it blows itself up. And all you have to believe for that is that when a species becomes technological in our universe, typically they get faster at building the technology than the social smarts to figure out how to control that technology and accidents happen. Now, I won't speculate too much on how good we are at social systems, but we're certainly very good at technology. I mean, look how far we've come from a horse and cart to man on the moon in a century or what have you.

50:16And it's just crazy how fast we're going. And now soon we're on the precipice of artificial general intelligence or super intelligence, which might be just a couple months to a couple of years max, a couple of decades away. So in geological times, really, very soon, we are going to have general intelligence or super intelligence, super intelligence being smarter than humans on everything. This is not a minor moment. This is arguably the most important things humans have ever done. It's not like any other technology because it's a technology that can build anything else.

50:47And how we do that could actually relate to Fermi paradox. Are we going to steer ourselves smartly through there? We sort of bumbled our way through nuclear weapons. We didn't blow ourselves up, but we came close a couple of times. What are we going to do here? This is a big test. And I happen to think that we haven't got a good path yet. But maybe, just maybe, training it on understanding all life on Earth will help it to appreciate that. I don't think we should rely on that. I think we need more robust mechanisms. But I think that could be part of the answer.

51:19I wanted to end where this whole series began. You talk about this in terms of a moonshot, but also you're describing it as if this is going to happen for sure. So I'm curious, in 40 years, say, what do you think the chances are that this will have come to fruition and we will be doing a lot of our AI computing in space? Like a number. What's your number on that?

51:49My number is 90%. That doesn't sound like a moonshot. No. In that sense, it's different from a moonshot. And the reason is that when we started to work on the Apollo project, we really did not know if it was possible. It was unknown whether it would work, but also it was unknown whether there was really a compelling reason to go to space. And indeed, it turned out that there wasn't at the time. The reason to do it was to show off and to look big alongside the Soviets after the Sputnik moment.

52:21For really doing space at scale, the motivation was just not really there. Whereas here, I think we're in a different situation for two reasons. One of them is that the physics and the engineering are already proven. Like we know how to do this. And it really is just a matter of whether there is a point in doing it. And we also know that there is a point. The physics of energy and the demand curves of AI are things that I think we've already understood. So when I think about what the 10% looks like where this doesn't happen, it is either my expectations about AI demand are completely violated, which would really surprise me, or we have some much larger scale setback and civilization really slows down for a while.

53:09I mean, those are the scenarios in which I think this will not happen. Let's hope that the great majority of that 10% is your expectations about AI, not nuclear holocaust or whatever incredible societal disaster that would lead us to be no longer able to launch satellites. I am an optimist, but there's no question that things like nuclear proliferation is getting more serious now. There are real risks, right? And I think we do have to acknowledge that. I'm not talking about AI apocalypses, et cetera.

53:40I'm just talking about the fact that we are an advanced and in some ways fragile civilization, and we have many, many means to harm ourselves given our great powers today. But you think that AI apocalypse is not among the leading worries of the next 40 years? I wouldn't say that it's not among my concerns, although I guess my versions of that probably look quite different from those of many AI doomers. What's your version of AI apocalypse? What do you think if something's going to go really wrong, what do you think would be?

54:12Well, AI can definitely be used to do really large-scale cyber attacks, bioengineering, et cetera. There was just a story in the New York Times about huge numbers of synthetic drugs that have been killing a lot of people in the Cook County prison. So that's a classic dual-use sort of scenario. The fact that we can use AI to make really large new numbers of drugs is great news in terms of drug development, but it has a dark side, too. And that's true of nearly every great power that we acquire.

54:43And to be clear, the reason that I'm fundamentally more optimistic than some people I encounter is because I have come to the conclusion that intelligence has this very fundamentally social character. Intelligence is pro-social. That's how intelligence is made. And so in that sense, the sort of fantasy of the evil supervillain, like I always used to wonder, why aren't there more evil supervillains? Why is that only a thing in cartoons? And I think the reason is that there is actually an inherent pro-sociality in intelligence.

55:16And I believe that has already saved us many, many times over. So that's why I'm optimistic. But it doesn't mean that we can't screw up.

55:27A lot of people talk about using energy for AI in a way that implies that energy is a zero-sum sort of thing, that if it's going to AI, then it's not going to something else that is better for humans or something. And there are two things that I think are important to keep in mind about this. One of them is that as entities become more intelligent, more sophisticated, they spend more and more of their energy on thinking. It's why so much of our energy, of our bodies, goes to our brain. That's only true of the brainiest recently evolved species.

55:59Fully a quarter of our energy goes into the operation of our brains. And as cities grow and become more intelligent, meaning that they develop a larger technological base and they become capable of more things, more and more of the energy in the city goes toward thinking as opposed to manual labor. So these are really large-scale trends that have to do with the whole evolution of intelligence on Earth. And from that perspective, you know, I fully expect that by the year 2100, the great majority of energy used in the solar system will be going toward AI and will be both harvested and consumed off Earth.

56:39So when I think about what 2100 could look like, it could look like this Earth as a beautiful biological paradise with lots of happy humans and animals and plants and so on, and a kind of computational mesh that extends much farther out in the solar system. And the amount of computing happening in that mesh will dwarf what is happening on the ground along with the amount of energy. So that's where I see things going over the longer term. And I think that it's a continuation of a process that has already been happening for three and a half billion years.

57:16Big thanks to Steve Levitt for hosting these two episodes on Space Data Centers. And thanks to all of our guests, Blaise Aguera-Yarkis, Travis Beals, Will Marshall, Alex McDonald, and Rosanna Hoffman. Coming up next time on the show, former SEC chair Gary Gensler has had a front row seat for every financial shock in recent history. So how is he thinking about the AI boom? I think that we have a stock market that is highly valued by any measure.

57:46We have a boom in the capital expenditures that will plateau and maybe even decline in the next few years. So when that comes, that's a reversal that you see will happen for all of these chip manufacturers, construction, and so forth. And something has to give. That's next time on the show. Until then, take care of yourself. And if you can, someone else too. Freakonomics Radio is produced by Renbud Radio. You can find our entire archive on any podcast app.

58:17It is also at Freakonomics.com, where we publish transcripts and show notes. This episode was produced by Augusta Chapman and edited by Gabriel Roth. It was mixed by Jake Loomis with help from Jeremy Johnston. The Freakonomics Radio network staff also includes Dalvin Abouaji, Eleanor Osborne, Ellen Frankman, Elsa Hernandez, Elaria Montenacourt, Pete Madden, and Theo Jacobs. Our theme song is Mr. Fortune by The Hitchhikers, and our composer is Luis Guerra. As always, thank you for listening.

58:52I've gone over time. I apologize. I was so engaged in our conversation. I didn't look at the clock. I don't even have the clock here because my phone's charging, so.

59:04The Freakonomics Radio network. The hidden side of everything.