Let's get started. Um, Philip, why don't we start at a very high level? Why put data centers in space?
That is a good question. So, we are building data centers in space primarily because of the constraints that we're facing on energy for building data centers terrestrially. So, if you want to build a new energy project terrestrially to power a data center in North America, you're looking at a sort of 5 to 10 year lead time just on the permitting for that. And so, we're really very constrained now on where we can put new energy projects. And in space, we can actually deploy these things incredibly quickly. Once the launch cost hits a certain point, it's really a no-brainer to move almost all data centers to space because we will have both a lower energy and infrastructure cost and a much faster time to deployment. So, that at a high level is the reason we're doing it.
And so, to introduce you and the company, this is Philip, who's the founder and CEO of Starcloud. And I'm Chetan. I'm on the board of Starcloud. And we led their Series A. Talk a little bit about how you formed the company, and the team, and what it takes to actually build a data center in space because you're the only company in the world today that has a GPU in space that's capable of doing inference and training.
Yeah. So, very quick background on myself. I started my career as an engineer and studied math and physics undergrad and masters. And then actually moved to the more commercial product side. So, I was with McKinsey working with the space agencies of different governments on different missions. My co-founder, Ezra, who's our CTO, has a decade of experience building satellites, has a PhD in engineering from Imperial. And then my third co-founder, Adi, who's our chief engineer, was previously at SpaceX as a principal software engineer. And he was responsible for a feature on Starlink called tracking beams, which enables Starlink via the spacecraft. And then before that, 20 years on the data center side with Microsoft. And the way that we came to this is essentially in 2023, I took a trip down to Starbase, Texas, kind of randomly on a weekend because I was interested in what was going on down there, even before the first Starship launch. And what struck me about being there is the scale at which they're building for. So, they're building these two Starship gigafactories that are designed to produce something like three Starships per day. So, it's absolutely mind-blowing new capacity that's going to come online. And I started speaking with Ezra soon after that about, 'Okay, so all of these concepts from sci-fi like space-based solar, where you have huge solar panels in space, all of that's going to come, is going to be possible at some point. It's just really a matter of time when the launch cost hits the point where that's practical.' And so, we actually started with that concept, and it's not a new concept, the idea where you have these huge solar panels in space, and then you beam the power down somehow using microwave or infrared technology.
The problem with that is you lose most of the energy in transmission from space to Earth. So, we started for about a month, we were just looking at the math. We wanted to know, what is the break-even launch cost where that business model breaks even with terrestrial data centers or terrestrial energy projects on just pure energy cost. And we came to a number of about $50 a kilo, which is in range of what's possible with Starship. It's a very long way off from where we are today, but we kind of thought, 'Okay, that's good enough. Let's get going on that.' And so, we started with that. And then we were like, 'Okay, once we get the power down, what are we going to be using it for?' And even a few years ago, most new energy projects were being built to power data centers. And so, we came to the conclusion that either directly or indirectly, this power is going to be going into data centers. And so, then we were like, 'Okay, let's rerun the numbers assuming we have a cheap way to get the data centers to space, and we don't lose all that energy in transmission.' So, we reran these calculations, and we came to a new figure of around $500 a kilo as the break-even launch cost where it makes sense to build data centers in space. And then that became the basis of a white paper that we put out in mid-2024, and then that essentially became the basis of the company.
That's amazing. And then talk about your first launch that you did last year and you got it working.
Yeah, so in only 21 months, so from day one founding to having something on orbit, it was 21 months. We launched our first spacecraft, what we call Starcloud 1. For context, the fastest any space company has done that before us is 4 years. So from day one founding to having something on orbit. So we moved very quickly. And that had five GPUs on there, two from Nvidia, three from ARM. The most interesting one is the Nvidia H100 chip, which is about 100 times more powerful GPU compute than has been in space before. So with that we were the first to train a model in space. We were the first to do high-powered inference on satellite imagery and SAR, synthetic aperture radar data. We were the first to run a version of Gemini, first to do fine-tuning of a model. And so that really proved the concept that we can run these very power-dense GPUs, terrestrial, so data-center-grade GPUs in space. And before that people really thought you couldn't actually run these high-powered GPUs for two reasons. One is they thought the thermal, it was too power-dense and you couldn't get rid of that heat in a vacuum. Number two, they thought that the chips wouldn't survive the high radiation environment of space. And so it was a real significant proof point showing that we can actually do this. And yeah, that was essentially the successful Starcloud 1.
And then, you know, I think the two common things that people ask you about a lot is, you know, cooling data centers in space is going to be impossible. The radiation, and then, you know, the chips are going to break and burn out on the way up. And then third, you know, who's going to take you up? And so, talk about how you solved all three problems. And then also talk about, you know, what impressed us the most is, you know, you're making everything. And the amazing thing about Starlink that I think is not well appreciated outside of Starlink is that the number of machines that they used to make the stuff, they've made themselves. And so, it's like truly vertically integrated manufacturing and it's super impressive in your factory in Seattle. So, talk about all of those things where you're not only solving the science problems, but you're doing it at extreme low cost such that the only cost of these space data centers is launch and the cost of the GPU.
Yeah, I mean, that is the core unlock for making this work is to make everything extremely low cost and manufacturable at scale. And so, yeah, as you mentioned, the two big engineering problems that we're solving are number one, how do we get rid of all this heat in a vacuum? And number two, how do we make these chips work in a high radiation environment? And so, for the first, it's really just building a very large, low cost, and low mass deployable radiator. And so, actually we know that radiators work in space because the International Space Station has been using this radiator for the last 20 years. The core challenge is actually just making it cheap and light. And that's not a new physics problem, that's just a new kind of manufacturing and engineering challenge. And so, for that one, about 70% of our team is actually working on building this radiator. And so, some of the things you mentioned and when you came, you know, we were working on is we basically built from scratch this... I don't want to go into too much detail, but like a very novel CFC machine...
That's some of the proprietary tech that you have. People don't know about it.
But I will tell you, like watching the team manufacture these components on the machines that you've built is truly impressive. And you know, I think that's like the part that is not simply like, can we put GPUs, can we put it in a box, and then can we throw it up into space and see what happens? But how do you think about launch partners? And what is the sort of manifest reality of launch partners and how excited people are to put Starcloud data centers into space?
Yeah. So, we are going to require an absolutely enormous amount of launch capacity. And, you know, right now we have a great relationship with SpaceX, and so we got three launches booked coming up for the next year. They launched our first spacecraft, Starcloud 1. You know, actually they're looking for users of Starship capacity. So, because they're building sort of on the order of 10,000 times the current capacity that Earth has to get things to space, they are very keen to find launch customers of that capacity. So, we're certainly excited to work with SpaceX. There are a bunch of other launch vehicles coming online very soon. So, Blue Origin has their New Glenn vehicle. They had a quite large explosion recently, but it looks like they're going to be back on track by the end of this year. And so, hopefully, you know, we've been speaking with Dave Limp. It looks like they're very keen, keen to actually get that up and running by the end of the year. Then there's Stoke Space. They're the only other one that's building a reusable upper stage. And so, yeah, we're excited to partner with them. We're booking launches with them. And then Rocket Lab also has their Neutron rocket, which is going to launch for the first time later this year. And then also Relativity Space. And we're actually booking launches with all of the above mentioned. So yeah, we are keen consumers of launch.
