All right, here we go. With Alec, Philip Johnston, StarCloud co-founder and CEO. I promise you what I'm about to say about this guy and this company is 100% true. They are launching data centers into outer space. And I remember when I first read that, I don't know where, it might have been one of your white papers, Philip, but I was absolutely unequivocally blown away, stoked, and I think I just reached out to you directly on LinkedIn and was like, 'Hey, can I interview you? Let's do something.' And you were kind enough to say yes, give me a little bit of time and we'll make it happen. And here we are. So without further delay, Philip, so good to have you here with the AI with Alec community. I know we're going to geek out. Give us something. Give us some origin stuff. Give us like, how did you come up with the idea? What problem are you solving? Like, let's go.
I mean, thanks so much for having me. I always like geeking out, so I appreciate you reaching out. Origin story is, I mean, my background, I'm more on the commercial side and also founded and sold another company in between. And then I'd spent time working at the space agencies of various governments in the Middle East with McKinsey & Company, a consulting firm. And so I'd seen from that period that the launch cost was coming down very rapidly, and I'm interested in space in general. So I'd been thinking, like, what would very low-cost launch enable? And in 2023, about May, just before the first launch, I went down to Starbase, Texas. I happened to be in the US and I was like, you know, am I allowed to swear on this thing? Can I say... I think I thought you were going to go there and I was like, yeah, go for it, see what happens. I had a week and I was like, I'm going to go down to see this thing. It was like the most insane, ambitious goal that they were building this fully reusable rocket. I think it's like three times the thrust of any other rocket before. Its payload bay is like five times the payload of a Falcon 9, which is the current workhorse. So it's just absolutely massive. So I was like, 'Okay, I'm going to go see this thing.' And what struck me about being there is not just that they're building this ambitious rocket, but the capacity they're planning for is absolutely insane. So they're building these two Starship Gigafactories, one in Florida, one in South Texas. They're saying within 18 months, they'll have three Starships a week coming off that production line. It's basically like a Tesla Gigafactory for Starships. And within 36 months, so in three years, they'll be having three Starships a day coming off that production line. And each Starship can fly three times a day potentially, you know, depending who you believe. Each one can put 100 to 200 tons of payload to orbit. I mean, the amount of capacity that we're going to have is launch cost could come down by X, capacity could go up by 10,000x, just because you can keep reflighting these things.
So that's my co-founder Ezra, we're from the same place in the UK. So I got in touch with him. I was like, you know, what would be... he's a space engineer. He spent the last decade designing satellites with defense and space and others, has a PhD in engineering. So I was like, what would be enabled by low-cost launch? And one of the first things we were looking at was space-based solar. So there's a sort of concept from the 70s where you have very large solar panels and then you use some kind of technology to get that power down. Usually people talk about microwave technology, so beaming the power down with microwaves. And that's an issue what we were looking at, and we were thinking, okay, what are we going to use this power for once we get it down? And the forecasts are, I think in 20 years' time, half of all terrestrial electricity consumption will go into data centers. So like, okay, we'll use it for data centers because that seems like a growing good market for it. The problem is you have like 95% energy loss or efficiency loss from space to Earth. For every 100 watts you're collecting in space, you're only getting five watts down usable on Earth. And so the next logical thing we were thinking of, well, if you can find a cheap way to get the data center to space, does it make sense to do that? And that's where the origin of the white paper came, was us looking at the calculations for what's the mass of a data center, what does the launch cost have to be for that to be economical. And it turns out that if your launch cost gets below around $400 a kilo, then it's a good deal. And if it goes much below that, then it's a complete inevitability. So I mean, the marginal launch cost that they're talking about with Starship is $10 a kilo. So that's like how much additional it will cost Starship for an additional launch. So it's potential. I mean, if we go down to $100 a kilo, we're a very good business. We go down to $20 a kilo, we're like an extremely profitable business. So that's really where we're playing towards, is building these extremely large data centers in space purely from an energy perspective.
It's amazing. So my audience, I know, is very into this dimension of thought, and just let's speak to this for a second. Right. So you just touched on the concerns about powering data centers if, quote unquote, nothing changes other than the trajectory of our consumption and use of artificial intelligence, for example. Can you paint that picture a little bit further and more? Because I think most people kind of like, they hear that and they're like, oh, that makes sense. But I think it'd be good to hear it from you a little bit further into how big is that problem and therefore how significant is this opportunity if and when it all works out the way that I'm sure it will for you guys.
Yeah, I mean, we're on this absolutely insane exponential growth curve now in terms of power consumption needed for running AI data centers, or data centers in general, but particularly for the latest inference data centers. Like, so I think let's say we have 100 GW of total data center capacity right now. You know, within five years that could be, yeah, 500 gigawatt, for example, and there just isn't that amount of... we just don't have the energy to do that right now. So we're going to need to build enormous new energy projects in order to be able to service these things. And like on a five-year time frame, that's doable. On a 10-year time frame, that's doable. But there comes a point where if we have, let's say, 100 times more data center need than we currently have, and that might be on a 20-year time frame, then there's just literally no way you can get 100 times the amount of power that we currently have from the grid. The amount of nuclear projects you'd need to build, the amount of enormous solar projects for that. And the real constraint is actually the lead time on the permitting, which I know sounds like a silly thing to be the constraint, but if you want to build a new nuclear project, for example, and by the way, nuclear isn't even that cheap. It's like 8 cents per kilowatt hour for a new nuclear project. But if you want to do that, the lead time on permitting is like 10 to 20 years currently. The best you can do is maybe get that down to five years, but it's extremely long lead time. Whereas with what we're doing, potentially within a month, you can launch a 40-megawatt data center. There's basically no limit to the amount of these things you could put in orbit, and you could do it very quickly. That's the real reason we're doing what we're doing.
It's amazing. And correct me if I'm wrong. I think I read in the Economist that recent article, and congrats on the coverage and all sorts of stuff, you know, of late. So I'm not surprised at all. But is it right that you're doing something in May, right? Coming up here with a refrigerator-sized, am I remembering this correctly? Data center, like this is a big deal, right?
Yeah. So we're launching our first demonstrator satellite. The launch window is May till August, and it's likely actually to be towards the back end of that. So that's not controlled by us, that's controlled by SpaceX. So yeah, certainly by the end of August we'll have an announcement out about the demonstrator satellite that's going up. So that will have the first terrestrial data center-grade GPUs in space. So about 100 times more powerful GPUs than are currently in orbit. So yeah, that will be a big milestone. We'll be the first to train a model in space, the first to run high-power inference, first to do fine-tuning of a model in space. Lots of firsts are coming up on that mission. And yeah, as you say, it's about the size of a fridge. It's like 50 kg or 65 kilograms with payload, small satellite bus. And then yeah, then next year we've got the second one going up. But just to finish on your point on total energy consumption, there comes a point, and we see that as around 100 times the current data center capacity, terrestrial data center capacity, where even if you solve the energy problem terrestrially, let's say nuclear fusion works, just the heat alone that these data centers are kicking off is enough to cause very significant problems in terms of global warming. And it's not global warming from carbon dioxide that's going out. It's global warming literally from trying to cool all of these chips. So basically most data centers are using fresh water to pump out steam. That's how they cool them. Well, number one, water vapor is actually a pretty good greenhouse gas itself. But number two, you're literally just pushing heat into the atmosphere. That will have a much worse warming effect than anything that is being caused by fossil fuels. So at some point, and that's only with 100x, we could be talking about a thousand-x or 10,000x increase over the next like 100 years. So at some point in the next 100 years, you do need to figure out how to dissipate all that heat into the vacuum of deep space. So you do need to have these things on satellites with large deployable radiators dissipating the heat and infrared into deep space. Otherwise, you're literally just making the planet uninhabitable by the heat process.
That's fascinating. And I think part of why when I first, like I said, I can't remember when or where I came across, I think the white paper, what have you, I was just blown away because maybe subconsciously it just makes perfect sense that with this level of change and movement in the direction of, you know, where we're all headed, this is a massive challenge and opportunity. And tell me, and you know, if there's any of this you can't share, totally get it, but tell us a little bit about what are you looking to test with that first launch? Like I think I read something about, well, I'm not going to steal your thunder, but go ahead and tell us a little bit about what you're comfortable sharing. What are you testing with that first launch coming up in that May to August period?
Yeah. So it's really just to say, can we operate data center-grade terrestrial GPUs in space? And that means, does our thermal management system allow us to do that? Does our radiation shielding techniques allow us to do that? Can the chips survive the vibration of launch? So we've done as much of this testing as we can on the ground. So we've put it on these huge vibration tables and we've shook the out of it, as my co-founder would say.
Nice. But you got to try, you got to test it for real, right? You got to get it over there and put it up in outer space.
