00:00Let's talk more about the challenges space presents to the things we send up into it.
00:04Mackenzie Listrup joins us, a former director for the Goddard Space Flight Center,
00:08which is NASA's primary hub for building and operating scientific satellites,
00:12now principal consultant at Peridot Services, and just an absolute expert in space systems.
00:18So there are different degrees of orbital compute, right?
00:22Different variations of a space-based data center.
00:25But at the extreme SpaceX is presenting, it is at hyperscale.
00:30It is running massive AI workloads, inference.
00:34Explain the distinction between those three grades of compute and why it's hard.
00:39Absolutely. So the first is really on-orbit edge computing.
00:43So this is about taking imagery data, radar, weather, maritime, climate, or defense data
00:50that's collected in space, and then using, you know, AI, you know, edge computing,
00:57some kind of inference, some kind of processing in the space system itself before sending it down to Earth.
01:04So this is real. This is near-term.
01:06So a satellite sends raw data to an orbital compute node.
01:09The node performs filtering or AI ML inference,
01:12and then only the higher value or smaller data products get sent to Earth.
01:16So this is a process that the company Axiom says that it deployed as a prototype unit
01:22on the International Space Station and has an orbital data center node as well.
01:28Now, the second is really resilient or sovereign storage and computing.
01:32So infrastructure that is physically separated from terrestrial threats,
01:38so disasters, political borders, or ground-based attacks,
01:41that also has some real defense, continuity of government,
01:44and critical infrastructure use cases, but still not entirely explored.
01:48Now, the third is hyperscale AI work in orbit, and this is really the moonshot.
01:53It's potentially enormous, but it is not merely about the launch cost
01:58or even the orbital downlink nodes.
02:01It really requires some challenges to be met in power, in heat rejection,
02:07in radiation-tolerant computing, high-bandwidth optical networking,
02:11orbital reliability, and other issues.
02:15So it's a real challenge.
02:17So let's zero in on some of those challenges, right?
02:19Your job was to get systems ready to be in the environment of space.
02:24The sci-fi nerd in me thinks about the vast coldness of space.
02:28But in thermal, it's a very different challenge, right?
02:32It is.
02:33So, you know, space is not cold sometimes in the way that people think.
02:38And I would say that in orbit, a data center is a heat engine first and foremost
02:43and a compute platform second.
02:45So a terrestrial data center rejects heat through chillers,
02:49through air, water, or liquid cooling, through evaporative systems as well.
02:55Now, in orbit, there is no convective air heat sink.
02:58So heat has to be moved through the spacecraft and then radiated away.
03:03Okay. So this means that radiator area, radiator emissivity,
03:08the attitude control of the spacecraft, contamination, degradation,
03:11these are all effects that have now become first-order economic variables.
03:17And then there is the power challenge.
03:20There's the belief that the sun is always on, infinite solar panel,
03:24as long as there's nothing blocking the path.
03:27But again, it's not as simple about that as that when it comes to access to that power.
03:32Correct. So, you know, the pitch for orbital data centers is often that,
03:36hey, sunlight is stronger and more continuous in the right orbit.
03:40Google's project Suncatcher argues that solar panels in the right orbit
03:43can be up to eight times more productive than on Earth.
03:46So that is true.
03:48But just because you have space energy access
03:51doesn't mean that you have space energy infrastructure.
03:54So you're going to have to deal with some real constraints.
03:57And that means the close formation orbital dynamics,
04:01the radiation effects, the thermal management, the ground communications,
04:04all of this depends on a reliable access to power.
04:10And that means power generation, but also power storage, distribution,
04:15again, that thermal issue of heat rejection, and also redundancy.
04:20Now, this has to be flight qualified.
04:21It also has to be able to be reliably operated in space and maintained.
04:27And maintenance is one of these hidden costs right now
04:29that we're not talking a lot about.
04:32Mackenzie, we have literally 10 seconds.
04:34Do you think SpaceX can do it?
04:36Well, they're a vertically integrated company
04:38that has demonstrated many technical challenges before,
04:41so I wouldn't count them out.
Comments