Computing in the Cosmos: Sophia Space's Mission to Move Data Centers Into Orbit

Hey everyone. Welcome back to over the Air AI Connected Devices and the Journey. My name is Ryan Prosser. Today I'm joined by my very good friend Rob d' Amelio from Sofia Space. So Rob, for folks that don't know, give us a little overview of Sophia. Thanks Ryan. First of all, happy to be here. Appreciate the opportunity to talk. So Sophia Space is a space tech startup company. It was spin out of a firm called Mandela Space Ventures, an idea that's been incubating for about three years by Dr. Leon Alkali. We started the company in earnest late December, early January this last year. And its purpose is to get computing off the planet and closer into orbit. Very cool. So you say off the planet into orbit. You know, people are dialing in today. They maybe have Starlink. They're becoming familiar with what's happening in lower Earth. Like where do you guys fall in? Not on this planet. So the way that we're structuring Sofia going forward, we are initially selling our in orbit servers, which we're calling Tiles. For the first few years of our existence. This will be to existing customers that produce satellites, stations, anyone that wants to have edge computing closer to the source of the data collection. However, our long term goal, and by long term we're talking about 2030, is to start putting up orbital data centers. And these are data centers that would be in orbit in space, chained together into configuration that we call string of pearls. And they would supplement the data centers that are occurring on the planet right now. So for someone looking up at the sky, are you guys the light in the sky that never moves? Are you the Starlink style, It zooms across my horizon. We would be. The Starlink style zooms across your horizon. But when we put up the orbital data centers, there'll be in a sun synchronous orbit. So that's an orbit that goes pole to pole. So it'll look a little different and it's a little higher than Starlink. Starlink is very low, low earth orbit. Sofia would be around the 800 kilometer level. Fascinating. And when you're thinking about the market that you're serving, what would be the type of use case that this serves best? So there's a couple different types of use cases. So when we're selling the tiles specifically to our clients, these are people that want edge computing at the source. So these are folks that have large data collection issues. They're collecting from sensors on satellites that may be terabytes or petabytes of information. What happens now is because compute is so Low powered for a lot of these satellites. Even though they have these amazing sensors that collect petabytes and petabytes of information, they throw about 95 to 98% of it out because they can't do computing locally. And it's too hard to get that kind of data back down to the planet to do processing and put it back up in orbit. Our early clients are going to be people that have a lot of data coming in that they want to process in near real time and make use of that data in orbit. So think of things like missile tracking, air traffic control, maritime tracking, environmental catastrophe tracking, weather alerts for the planet, that sort of thing. One of the things that we've heard a lot about in the last 10 years is space is becoming more crowded in the near Earth orbit world. Does Sophia have a kind of a position or a thought on what that might look like in 10 years? Getting crowded. The thing to remember is that when you start talking about orbits that are in the 300, 400, 800, 2000 kilometer level, your surface area grows. And so there's a lot of computer graphics that you see of like debris floating around the orbit. It is absolutely a problem, but those graphics make it look a little bit more dense than it actually feels when you're up there. Now, having said that, yeah, there is a whole push for something called a LEO economy. There's a lot of companies right now and government agencies all around the world that are trying to place things into orbit to take advantage of that space for a number of reasons. So it is getting crowded. But the bigger problem, of course, is all the debris from the last several decades of pushing things into orbit and just disposing of garbage in space. Right. So in a lot of those layers, there's small bolts, chips of paint, that sort of thing that are floating around. There are several companies that are coming up with schemes to remove it or at least mitigate it. Our position on it is it's there. We'll assist with computing. Of course, for anyone that wants to use this to help target debris and pull a debris out of the way. But when it comes to satellites bumping into each other, you're not going to see that happen for a while. Okay, dumb question. Like a typical size of space garbage, if that's the word, is like a bolt. Yeah, I don't think there is really a typical size. It can go down from particles the size of a grain of sand up to sheet metal, basically, but less and less of the larger pieces. So, yeah, it is mostly around the Size of, you know, think of small screws and bolts and things like that. I think people are often surprised when they look into the Pacific Garbage Patch. I mean, look into as in research, and it's not this mass of giant, like porta potties and these tires. It's a huge, like everything is the size of a skittle. You know, a lot of tiny pieces that have all come together. It seems like. Talk about the garbage in the Pacific Ocean. Yes, exactly. The cheeriomagnetism effect. Yes. I think I've also. Cheeriomagnetism. Yeah. I wish I could claim that term, but I can't. Yeah. In orbit you don't get that effect, that skin effect of things like, you know, kind of adhering to each other is just sort of free range, dead roaming. But it's all in, you know, since. Since spacecraft and satellite and other vessels get launched, they get launched into a particular direction for each orbit that they're in. So all of these particles are kind of moving at the same in the same location, the same rate, but they're up there. Yeah, it is a problem. Hang on, please tell me we've agreed as a planet, we're using kilometers when measuring things in space. Yeah, yeah, we absolutely are. Yeah. So. So we, you know, we talk about things in terms of kilometers and, you know, you talk about velocity in terms of metric units. And there's a lot of. Most of. Most of the language of space is metric. So as a planet, we've come together that when talking about things not of this planet, like, if we're talking about a spaceship, we're going to talk about how many meters long it is. Sure. Yeah, we talk about meters. We talk about kilometers. Yeah. We talk about kilometers per second. Yeah. All right, America, let's get our act together on this. Can we please come to. I think it's this. And it's daylight savings time. You know, as a people, like, let's get there. I have strong opinions about daylight savings. Okay, let's take it back to Sophia's here. So I was curious about, you know, just looking at you guys product going through some of the materials, which are totally fascinating. There's obvious advantages with solar. So energy harvesting in space is going to be tremendous. But the thermal piece I thought was really interesting. So we hear a lot about data centers generating a lot of heat. Space is both famously hot and famously cold. What's the play to take advantage? How do you, in a vacuum actually make that work? So the way that we're doing this, we have invented an object called a Tile. And you can think of a tile as our atomic unit for compute. So IT is a 1 meter by 1 meter by 1 centimeter object that consists of four servers that are the equivalent of four earthbound data center servers. The way it works is we collect solar energy in from the top through a solar cell. Brings in about 1.3 kilowatts of energy. Once you do the math, that computes out to about 400 to 450 watts of power to power the compute layer. And then because the compute layer is spread so thin, there's a proprietary metal heat sink on the back that even at max CPU and GPU use of those servers, the heat dissipates back out into the cold of space. And at the altitudes we're talking about, space is about 4 degree Kelvin. So it's extremely cold. And so the heat transfer process is very efficient. So tiles are self contained, they are modular, they are solid state. There's no moving parts. As a matter of fact, they wouldn't even work on Earth. If you turn them on down on Earth, they would heat up to the point where the circuits would stop operating within seconds. So in orbit, however, these things will function really, really well. So we're dealing with three solutions at once. So terrestrial data centers, they consume a lot of power. Something like 40% of the power of a terrestrial data center goes to cooling it, and nobody wants them around. The NIMBY thing is very true. So we have significantly fewer terrestrial data centers than we need right now. Projections are that by 2030 there'll be 3x the amount of data centers there currently are. There's a lot of debates as to where to put these things, how to keep them cool, et cetera, et cetera. But the big bugaboo that people who build out data centers on Earth don't like to talk about is heat. You're right. So that is actually the number one topic, because the laws of thermodynamics are real. You can't really get rid of heat on the planet. You can just move it from one place to the other. So you're left with the land, air or water. There's debates going on as to whether or not terrestrial data centers are going to contribute to the overall global heat footprint. But what they absolutely do and cannot be debated is that they provide a localized heat source wherever they're placed. And if you look at all the studies on urban heat sinks and heat sources, it's a problem. So, you know, we've already got problems with cities built out of concrete and that they produce a lot of heat because all the people live there. So now we're adding to it a data center that, you know, takes as much power as a large city and then 40% of that power gets translated to heat in a localized area. So it is a problem. And the reason, you know, here at terrestrial data center people talk about this is because there's no solution. Right? You can just move the heat from place to place, but that's it. So in orbit, you don't have that problem. Your heat sink is the entire known universe. So it's fairly easy to dissipate heat. So when you're in 4km of temperature and those tiles are producing heat, it just gets passed off into space through radiation. Fascinating when you're thinking about your business, competitive advantages, et cetera, without revealing secret sauce, of course. But like, what are some of the really difficult problems, problem or problems on this show, we call it the wrong side of impossible, but that you guys are working to solve that, like, hey, we've got to get this right. This is core to our business, is it? This heat problem, is it? We actually think we have the heat problem solved. So I don't want to make too light of it, but it is a problem that came out of the Mandela and JPL studies that started this whole thing. So we believe that that problem is something that we can address. One of the issues that keeps me up at night, if we're still using those trite expressions, is this has to be self managed and self maintained. You're not going to get the terrestrial data center IT person up into orbit to fix a broken tile. And so we're putting a lot of time and effort into that interface between the hardware and the operating system to make sure that the operating system can handle issues that are occurring within the tiles, including the tile being taken out of service by debris, for instance. Right. So the operating system will become aware of how many tiles are attached in a collection of tiles. When a tile is acting strange or seems to be overheating or has too much processing going on, the operating system will move those processing the processing tasks around all of the other modules. Same is true for heat. Right. So the secret sauce here is all the work that we're putting into making sure that the tiles continue to function and that the operating system does the job of the guys in the white suits that go into data centers. When you're thinking about satellites and communication technology in space, do these things ever come back home? Or like, once launched, is it. It's going to Work until we decommission. It'll work until it gets decommissioned, it'll be deorbited and it'll burn up in orbit. Do you think we'll ever see a day when the thing in space, the satellite, whatever is valuable? What I'm getting at here is you said we're not going to have an IT guy in space. And my qu. You know, what I'm getting at is like, I wonder if that's not true. You know, I wonder, are we going to see a space IT guy? Because the asset in space is valuable enough that it merits putting Fred up there with a wrench to do the thing. Do you think we'll ever see that? Or it's like, well, I mean, some of that was already started as far back as Skylab, right Where you had people in a space station doing work on the space station. And as the space shuttle program took off, one of the big selling points in the components of the space shuttle was that there were astronauts on board. You could send up mission specialists and that's literally what they would do. They would put shuttles in orbit and they use the Canada arm to grab something. This is how the Hubble was repaired. Grab something that is in orbit, bring it into the shuttle bay. They would work on it and they put it back. So it's something that has already happened, but it is and was expensive to do that. Will there be a time where space travel is routine and people will be able to do repairs in orbit? Probably, yeah. I would say that was be probably true. But that's not going to be anytime in the near future. We're relying a lot on robotics. We're relying a lot on autonomous vehicles. And there's a reason for that. We had a well known space agency on the show a few episodes ago and ended up with a lot of redacted content. I'm not going to put you on the spot. Bob's Space Agency, is that what it was? A different one? The other one. I'm curious, what's your final question? But would love, you know, you guys are going to space. Space is cool. We're big fans of space here on the show. What's your take on Mars? When are we going to see the first guy. Person, I should say, when are we going to have the first person on Mars, do you think? Like boots on red? Hard to tell. It's sooner than people think. It's not as close as certain people want you to think. Right. So definitely, probably within the next couple of decades. I think there'll be something along those lines. But you're going to see a lot of. I subscribe to the Carl Sagan model, right. Where you're going to see a lot of robotic activity on Mars long before you see people get to Mars. Mostly because unless there is a. There is a compelling need to do so to have human presence on the surface of Mars. It's just cheaper and easier to send robots. It's hard to keep these ugly bags of mostly water alive as you get them to the surface of Mars and then bring them back safely or leave them there to populate the planet as some people would like to do. It will absolutely happen because it always happens. Expansion, human expansion is a thing that just actually happens. But I don't think you're going to see it in the next 10 years. You'll probably see it in the next 30. Interesting. I think people have been pointing at, I believe it's 2028 for some time because the proximity of Earth and Mars. Yeah. Folks at home, Rob checking his watch. Yeah, 2028 feels like it's going to come and go. But you feel like, hey, this is not a question of three years versus five. This is a question of. Is it two decades versus three? Yeah, yeah. It's cool that Mars is going to have its closest approach to Earth in a while. It's an artificial window, I think. And I can't imagine it happening in the next three years. I know this is the actual last question. Do you. Let's pick a time horizon of 30 years. So it's, you know, we're sitting here, it's 2025. So between now and 2055, what do you think is the most significant space oriented milestone that humans will have achieved in that time horizon? Is it this Mars thing? Is it. It's business activity. Right. It's commercial and consumer activity. As crass as it is to sound, to say, and I am aware of what it sounds like, business does drive this sort of thing. I'm all for the Star Trek ethic. We're exploring to explore and that of course will continue. But there needs to be sort of a fiduciary benefit for people to sink the literal billions or trillions of dollars into moving us off the planet. I think that you're going to see a lot of activity around the asteroid belts before you're going to see getting into and out of the gravity wells that are planets. Right. So it's hard, it's hard to get down to Mars and get back up. It's hard to get down to Titan and get back up. So, you know, from that point of view, I think that is got to have a lot of reason to, you know, to get us there. Asteroids, however, are a different story, right? Asteroids and the asteroid belt in general. There's almost no gravity cost to getting on and off of an asteroid. People at home will correct me, but you know, you know what I'm talking about. Yes, yeah, I take your point. Absolutely. Yeah. They're small, they're small objects and they come to you and they come to you and they're filled with things. They're filled with, you know, water and gold and you know, there's a lot of, there's a lot of material that are in sort of the raw beginnings of the solar system that are floating around. And when I think about how this is all going to play out, it plays out more like the Expanse than it plays out like Star Trek. So there is a reason to go there to collect this material, to bring the material back or use it in space to build. So I think over the next 30 years you're going to see a lot of that happening. You're also going to see a lot of activity around Earth orbits because you get a lot more information and you can take action a lot more quickly when you are in orbit looking down at the planet than you are at the planet looking up. Right? Where you've only got one point of view, one field of view. When you're in orbit, you can see from horizon to horizon. And that gives you a lot of opportunities to do a lot of things which you need computing for. But I won't ask, I won't be self serving. It's funny, you say the Expanse, I say Avatar. You know, this is the plot of Avatar where they allowed some research, they funded some research, but essentially, you know, they're harvesting the, the assets of, you know, this other planet. We're out of time, so I'll get to this another day. We'd love to have you back on the show, but what you're describing reminds me a lot of a profession that, you know, my childhood friend's parents did, which was these offshore oil rig, you know, dangerous harvesting of minerals and materials in a really well paying job, hard job, dangerous job, but not necessarily one that required like, you know, I'm sure I'm gonna get this wrong, people are gonna be mad. But like, not like a tremendously cerebrally taxing job, you know, you could, I think it would be, actually I do think blue collar jobs are cerebrally taxing. It's you can't put an idiot on an oil rig, expect them to survive but and the same is true for asteroid mining. You've got to, you've got to be, you know, strength of character. You have to be able to think faster than, than some of the other folks around you. You got to be able to handle catastrophes when they occur because they will occur. Crisis management. So you know, it's just a different part of the. Can you imagine the first asteroid miner that misses his takeoff window and he just sails off into. That could happen. It's gonna happen. If the asteroid's small enough he could jump. Yeah but you know, if he doesn't jump and it's headed right towards the sun. What a ride. Yeah. Yeah. Rob, thank you so much for being on the show. Great to see you, Ryan and thank you for listening. Join us next time as we meet with another executive and talk about things that went wrong on a journey that went right.