Today we have a very special episode of the Giant Ideas podcast. Something a little bit different.
We are in Menlo Park recording live from Aether Flux's offices with none other than Baiju Bhatt, co-founder and CEO.
Hey man, we're in San Carlos actually, but right next to Menlo Park.
Close. Yeah, close. What we got here on the walls here with the murals?
Yes. So the concept behind this mural was George Washington crossing the Delaware. But I wanted to make it a little bit weirder. So it's George Washington crossing the Van Allen belts on a satellite. And the creative direction that I gave was I wanted it to be like a who's who of American presidents.
To avoid any politics. We go back to the ones that are no longer with us. The legends. Not the living legends. And the prompt was I wanted all the leaders to be shirtless and ripped. And I was like, I wanted it to be aggressively American. And we kind of had to settle on like, no, no guns. Like actually let's make it aggressively American like hamburgers and weightlifting.
And so we've got George Washington over there with Lincoln and Nikola Tesla. George Washington is holding a hairless cat.
Is that a reference to Dr. Evil?
No. I think that was just, I really love cats. I don't know about the hairless one.
Okay. It strikes me as like a pantheon of the American gods with a psychedelic twist in space.
Little bit. Yeah, a little bit. The robotic American Eagle. And we've got Teddy Roosevelt over there riding a T-Rex in the arena.
Yeah, exactly. And a robotic shirtless Reagan just sort of looking down.
It's inspiring. Well, today there was some leaked news that you may be raising a large financing round. Let's give our audience an overview of what your giant idea is.
Yeah, would love to. So, Aether Flux, the vision of the company is to build an energy grid in space. So, we'll start out with kind of the idea behind it and then what the exact thing is. So I started the company with the idea of exactly this, that low earth orbit was an emerging platform for commercial activity and the sort of scope of commercial activity on this platform I felt was pretty limited. Prior to this wave of industrialization it's been earth imaging, it's been defense stuff, and it's been telecommunications. That's kind of what's going on in space. And I want to live in a world where the space environment is not dominated by governments and nation states but it should be a commercial environment because if we want a lot of people to be a part of the space economy we need to have great jobs for people to do and we need young people that are coming out of school to be like, yeah, that's what I want to do. I want to be in space. I want to get a job working at a space company and to do that we need to create opportunities that are purely commercial in space. And my thought process was that energy is the next industry that's going to exist in space and triangulated that from a variety of reasons. One, that the component technology is reaching the right level of maturity. Two, some applications that I originally had but this is kind of getting into where the vision of the company is going, new ones that are emerging. And also for any economy energy is like one of the core ingredients.
The core input, right. Yeah.
And so the vision of the company started out at first with pursuing these DoD applications, delivering power from space where it's either a contested environment, there's no existing power grid, or you're trying to send power somewhere that there's no real alternative to getting power. So think things like forward deployed bases, military locations, but also things like power to things in the atmosphere or power to other satellites. So that's how we started out. We're actually going to be doing space-based solar power, which is putting solar panels on satellites to generate energy in space and then using lasers to transmit it. And actually, there's the very first one that's right over there.
So you got these satellites, they got solar panels on them, and then you've got a laser that's going to beam that energy back to parts of Earth and potentially to the atmosphere or assets in space. Some space GPUs which would power them up.
Exactly. So we did that and about a year ago I was thinking very critically about the technoeconomics of energy in space and the light bulb went off that actually two things. Number one, the primary driver of energy use was going to be artificial intelligence. I think when I first started the company that was a thought, as time went on it became clear that this was kind of the driver. Like second to none. The second one is that the energy economics of doing it in space, of harvesting power in space, were very very interesting and I thought they unlocked an opportunity there that would be purely commercial in nature. So about a year ago started working on the concept of solar powered data center satellites and now I can say that that's one of the major focuses of what we're doing.
Okay. Like that's kind of the direction that we're taking the commercial side.
Yeah it is. It is. There's a lot more activity which frankly is awesome. It's much better that people coalesce around a set of ideas and pursue it and we obviously think that we have a very good path towards being a major player here. We think there's going to be other people that are also going to be successful doing this.
It's definitely not a winner takes all.
No, I mean I really don't think so.
So I want to get into the technoeconomics because that's a really interesting topic. But before we do that, let's assume that things go as well as we could possibly imagine. What's the most outrageous vision of the space future you could see? Because this literally will sound like sci-fi to a lot of people. And I think the first articulation of this idea, solar panels, energy in space, actually was in an Isaac Asimov novel, right? In the 1940s, 50s.
It's about like robots in space, too. It's kind of all converging on that idea.
It's all converging. It's like what's the most wild wild future we can imagine.
So I think it's literally rings around the planet. I think if a civilization were to look at the Earth, say in 500 or a thousand years, it looks a lot like Saturn. And why do I think that is? It's because I think a lot of this data centers in space and energy and space applications are specifically going to exist in these dawn dusk sun-synchronous low Earth orbits. Those particular orbits are kind of magical because they're still within low Earth orbit and so you're not in the extremely harsh radiation environment sort of in geostationary or beyond. Still a relatively harsh radiation environment, but you can have nearly continuous illumination by the sun. And so it's kind of like the best of many worlds. And so I think you're literally going to see rings around the planet because I think these orbits are going to be very densely populated and I think as humans we'll look up and we'll be like, 'Oh, that's where the computers are.'
Amazing. Yeah. Amazing. Is that audacious enough?
And I guess people going up there regularly to do whatever we can do in space potentially.
Okay. So the technoeconomics, there are a lot of people who are very bullish on this. Elon's obviously kind of very vocal about it. There are some people who are pretty skeptical who think that it's hard to kind of square the math that it's very hard to service GPUs. This particular data center in space, I think it's very hard to service GPUs in space. Cooling is a huge issue in space. We are assuming that the cost of launch is going to continue to come down but nonetheless it's going to be very hard to just beat the economics of data centers on Earth. What do you think about all of that?
Number one, I think there is a path to having competitive unit economics. There's going to be a bunch of different approaches that we take, that different people take to it. We have one approach that we'll be talking about in the not too distant future that kind of shows what our approach to this is going to be. I think the fundamental thing that we're looking for at least is the cost per GPU hour and benchmarking that against what people pay on Earth. And can you get that cost competitive? And the real value proposition of this I think over the next five to seven years is going to be the demand for AI compute is, there's no world in which I see this slowing down.
I have a great story on that. Someone who works very closely with the CEO of DeepMind told me that if they could get 10x the data center capacity tomorrow they would have demand for it immediately.
I don't disbelieve that at all.
I thought that was pretty amazing from Google. I mean, it's again, there's a lot of different ways history can play out and this is one viewpoint, but when you just look at how much stuff we could be doing with that and kind of when you think about AI at scale serving the whole population of the planet, you're like, yeah, you can see a world where every one of us has multiple agents or AIs that are running in the background nearly continuously.
Sure. And just soaking up GPUs and using more and more energy and higher and higher performance GPUs to get more and more sophisticated with things that we're doing and right now you just can't do that, right? And you have a very exciting announcement with Nvidia, right?
Yes, that's absolutely right. We are excited to be partnering with them on this and we've talked about the announcement there but I think taking a step back, the cool thing about this kind of connects to the point that you made earlier that there is competition here. Which is that people care about this, right? There's a world in which we have to do this where we are sort of like pioneering making these chipsets work in space which is frankly a lot harder than having the company that makes them being like, yeah, no, we're committing to actually supporting this. We want to grow this as a business model. And so it actually fundamentally makes this more applicable to a broader set of potential buyers in the future. And it kind of shows that as the GPUs on Earth are getting better, there's going to be work that's happening in the background to get a parallel path ideally of space rated ones that are getting better too.
Do you think that's the end state kind of in 10, 20 years or do you think that's a reality that we're going to see in the next kind of two to three years as some of the most bullish?
Yeah, I think the two to five year time frame.
Two to five year time frame. Okay. And on the kind of pure, the kind of more...
I don't think we have to wait 10 years just to...
Not 10 years. Okay. Not no, I don't think so. And what do you think of some of the competition in the industry? What's your take on it? We've got StarCloud, I think they announced the round yesterday. Obviously SpaceX is doing this thing.
Awesome. Really proud that they were able to do that.
Yeah. We've got Google, we got Eric Schmidt, I think doing stuff. Jeff Bezos. What's your kind of take on the whole space?
Man, there's a lot of very successful people that built huge companies that are working on this. I was going to say it seems to be the rite of passage. If you're one of the most successful entrepreneurs in the world, you must build a space data center company.
No, it's, you know, specifically talking about Bezos and Elon and those folks. It's really awesome to see the sort of biggest space companies in the world committing to this. The thought that comes to my mind here is kind of comes back to this commercialization thing, which is if you look at the total commercial envelope of space, it's not a ton. The total dollars being spent for commercial use is not that big. And when you put that next to the dollars that are going to be spent on AI and infrastructure build over the next decade, it's trillions and trillions of dollars. The reality in which some part of that then flows to developing the space economy. Let me be really emphatic about this. This is the first time this has happened at this scale. There's a reason that everybody that's working on space is like, oh yeah, that's on an unseen scale. Because the thing that has limited the space industry has been that it's really expensive and it's super capital intensive and it's on a bit of an island of how capital and resource intensive it is to do it. Now, if you can have even a small percentage of that spend on AI and the buildout of AI drive the commercialization and buildout of space, having a vibrant space ecosystem is going to be a byproduct of this. It's a really really big deal.
And again, I can't stress this enough, the commercial opportunities in space before this have been few and far between. People looking for what are the commercial uses for heavy launch vehicles and stuff like that. I mean this is like, it doesn't get bigger than the projected dollars that are going to be spent on AI.
And I guess one of the blockers here is just how hard it is actually to get data centers connected to the grid. Yeah. Just kind of getting permits to build. That's all kind of a huge blocker here to acceleration and the time frame on that stuff is ballooning, right?
Yeah. Exponentially. And I think that is actually the sort of real concrete problem that space data centers at scale will solve is you can go from like if you have two options, right? One is you got a truck full of GPUs, you can drive them to a construction site where you're building a data center or you can drive them to our site where we integrate them into a space vehicle that's going to then put them in orbit. On one path it's going to be, we know how long it takes, it's pushing five years in many cases domestically, less in other ones, more sometimes abroad. Versus you take those same chips you can integrate them into a vehicle, into satellites, get them to the launchpad, have them up in space ideally in a matter of weeks. And the reason is because you're taking the process by which they're made which is industrial manufacturing and you're solving the congestion problem by having the process by which they come online not be a real estate and permitting problem or like a real estate construction energy grid problem.
It becomes a manufacturing problem.
Manufacturing problem. Okay. Okay. And I guess the limiting factor here could be the rate of and cost of launch. Yes. So how do you think about that? Obviously there's kind of one game in town at the moment which is SpaceX and Elon. I know you've already, I think, booked your initial launch or you...
We've got a couple of things in there that are going to be going up.
There we go. But yeah, how do you think about that dynamic?
I think it's going to be really interesting to see how it plays out. I think there's going to be a lot of new companies that are entering the launch capability game. I think the ones that have been doing this for a long time are building truly spectacular vehicles that if we can see those things sort of live up to their potential, there's going to be a ton of mass going to orbit, every week, month, etc. So, I think this is actually one of the really important limiting factors for this space reality becoming a thing.
Yeah. Clearly. Yeah. Because I think that maybe the negative view could be that he, you know, they could keep the cost at launch at a place where it makes sense for them to do their activities but not for others, right?
Yeah, that would be, that's certainly possible.
Possible. Okay. Well, listen, one of the things that initially attracted us so much to your idea was this idea of abundant clean energy, right? For...
