SPEAKER_00 0:00

So as I mentioned, as Ron asked about different use cases, missile warning and missile tracking and uh space domain awareness, you know, um in the broadest sense I could imagine you know hundreds of thousands of satellites helping, but that may not be necessary.

SPEAKER_01 0:30

Hello, welcome to the Space Policy Show, brought to you by the Center for Space Policy and Strategy, part of the Aerospace Corporation. And I'm Colleen Stover. So all of our shows are available on YouTube and by podcast. Uh you can browse all of them also on our website that's csps.aerospace.org slash space policy show. So today's show, we're talking about ODCs, orbital data centers. What's going on? Let me read a couple headlines to you. So this one is from January 2026. SpaceX plans to launch 1 million satellites to power Orbital AI data center. Then a startup, Orbital. Orbital seeks FCC approval for 100,000 satellite data center constellation. Also this year, StarCloud, another company, is targeting a constellation of 86,000 satellites and 20 gigawatts of orbital compute capacity. And then even China's getting in the game. Here's their headline. China backs orbital data center startup with $8.4 billion in credit lines. Yeah. What is going on here? What advantages could a mesh network of smaller data centers in orbit offer? So terrestrial data centers are in the headlines, oftentimes, for concerns about their high usage of energy to run the large centers, and then also the use of water to cool them. Terrestrial data centers are facing challenges from local and state governments, and also the nearby populations of neighboring people. Not to mention that terrestrial data centers can also be vulnerable to attack during conflicts on the ground. So demand for new and expanding data centers has continued, however. Meanwhile, space offers the endless power of the sun, essentially being off-grid. Space is cold, so there's no water cooling. Space doesn't really have neighbors or local government to contend with. Space is big, so scaling new centers may be easier up there. And it's a bit harder to reach space with your attack missile or drone. So in terms of the types of data processing that we're talking about, space can offer an alternative place to store data as well as train models, but it can also do some in-orbit computing, whether all that computing data goes down to the ground or is used for in-space decision making, what we call edge computing. So today's guests have been thinking about these different use cases and also about some of the bottlenecks and the operational challenges of putting up and managing all those objects in orbit. To talk about this today, I have co-host and expert Ron Burke. He is a CSPS partner and principal director for Aerospace's Civil Systems Group. We also have Andrew Abraham. He is a senior engineer in the space object risk department. And we have Tim Bixler. He is an aerospace technical fellow working in the cloud computing architecture area. So first I'm going to ask you guys to introduce yourself a little bit to our audience and kind of why I asked you to come here. Tim, can we start with you?

SPEAKER_00 3:50

Absolutely. Thank you, Colleen, and uh great to be here. Um, my focus area as a tech fellow within aerospace is in distributed systems, applied AI, space cybersecurity. And, you know, I have uh about 30 plus years of experience in designing and operating large-scale data and competing services uh for companies from Bloomberg to uh pioneering in the cloud with Amazon Web Services. So happy to be here and a part of the dialogue. Thank you.

SPEAKER_03 4:17

Okay, well, hello. Um I'm Andrew Abraham, um senior space traffic management uh specialist here at the Aerospace Corporation. Um thank you for having me today. Um I've spent a lot of time working with uh various government uh entities such as NASA and Office of Space Commerce and others on space traffic management related um challenges and opportunities. I've also interacted with a number of um uh constellation operators, commercial operators, right? And um have absorbed a lot of how they uh conduct their operations and have really thought through how they might apply it to orbital data centers. And as you see, the numbers can grow to be quite large. So uh happy to join you guys today.

SPEAKER_01 5:04

Thank you. And co-host extraordinaire.

SPEAKER_02 5:08

All right, thank you, Colleen. So good to be here with you and with Tim and Andrew on this very important topic. Uh you've talked about numbers, you've talked about push-pull. You know, one of the things that uh we're aware of is that there are many different applications of on-orbit data centers that range anywhere from edge computing to training and inference uh applications. And and and we note that the not only the design and development of the systems, but the operation of the systems um as a span uh to to accommodate those different different applications. We are really fortunate, and I am very much looking forward to hearing from our experts uh today um as they uh sort of unpack uh the uh the different dimensions of being able to operate and and deliver these applications in space.

SPEAKER_01 6:02

Thank you. Yes. Yeah, we're lucky to have uh aerospace thinking about this. Um so I wanted to kind of start um you know noting that there's different kind of applications. I mean, there's there's the kind of space-to-space things that people are talking about, there's the space-to-earth things that people are talking about. But how uh do we uh are we kind of envisioning ODCs to be used either by government or by commercial entities? And Tim, I'll start with you.

SPEAKER_00 6:29

Oh, absolutely. Thanks, Colleen. Um, you know, as I think about the government angle, um, promising areas that we've been looking at, we're starting to study, are in missile warning and missile tracking, um, as well as space domain awareness. So those are those seem like they're um high value areas to to dig deep further. Um, you know, when you when you think about those missions and quite honestly, other kinds of missions uh and applications, there are really kind of six areas that I look at as uh potential value areas. And, you know, one is um I'll briefly go over um resiliency, you know, continuing um continuity of mission um area activity despite degradation or denial of ground control, uh data routing, uh including tipping and queuing uh from sensors to ODC end users without ground in the loop. You have data fusion, uh combining data from multiple sensors and phenomenologies into products for further dissemination. Uh you have data exploitation, uh, performing uh processing that uh requires large data sets. Um Ron mentioned uh inference, large language models and inference uh training and things like that, targeted typing, etc. And a few others, um uh data reclamation, leveraging data in processing that would otherwise be dropped on the floor uh due to bandwidth or download link constraints. And then finally, uh load relief, uh moving the processing burden from large data sets to ODCs, um uh, you know, including high volume or or uh computationally intensive tasks. So to wrap it up, those are you know, like like terrestrial solutions, um, you know, the specific combinations of compute and storage and networking and all those things that you know I've I've spent my career thinking about um are highly dependent on the process, the application processing requirements uh that are being hosted. Um so I'll I'll pause it there and see if others um, but that's how I think about um the government use cases, and I think largely in a broad sense, uh could be applicable uh commercially as well.

SPEAKER_01 8:34

Yeah, exactly. So you mentioned, I mean, a lot of those things, you know, in terms of um um, you know, loading a lot of data or processing things, inference, those are those are transferable across other areas, right? Andrew, I mean, commercials kind of looking at similar things.

SPEAKER_03 8:54

Absolutely, yeah. And I think when it comes to commercial, there's a very diverse ecosystem. So it really depends on who you're talking about, right? So a lot of the benefits that Tim just mentioned can be applied uh to the commercial world as well. So I'll just use like concrete example. Um, let's say, you know, your end-use customer is um uh agricultural entity, a farmer, big farm, right? So they're monitoring crops. Um so you could take this uh picture every single day of the exact same cornfield and you can downlink that picture to the exact same ground site, and you can imagine there'd be a bottleneck that would form, right? Everybody's trying to do the same thing, so really using that same ground antenna, or you could leverage an orbital data center or the infrastructure of an orbital data center and process that image up in space and only send maybe a few text fields that say, you know, this cornfield is dry, you might need to take an action. This other cornfield is uh well watered, don't worry about it, right? That kind of thing. Um so leveraging um, you know, the ability to minimize that downlink pipe is great. Uh, national sovereignty is another uh important issue, right? So you kind of fly the flag of the country that you receive your license from, uh, but you're distributed globally. So there could be some real advantages of you know protecting your data or moving your data around, but still being flagged under the flag of choice that you happen to have. Um so those are kind of um on one end of the spectrum. Um on the opposite end of the spectrum, uh it really comes down to one word, and that's scale. Right? So there are some very large um uh satellite operators that are very vertically integrated, and right, they own launch, they own even some launch pads. Um they are planning to deploy very, very large numbers of orbital data centers to really scale more rapidly, they argue, than what you can otherwise scale using ground-based infrastructure. And earlier you mentioned uh some advantages and some ground-based limiters, right? So, one example of that is uh when you try to um build a ground-based data center, you have to go through local zoning, right? Permitting and all those laws with each local town. And each one has kind of their own take on things. As opposed to um, you know, if you try to deploy orbital data centers, yes, you have that one kind of upfront effort, and it's not trivial where you're trying to uh get a launch license, um, an FCC license as well. Uh, but once you kind of um met that hurdle, you can scale as much as your license permits you to do. And like you said, some licenses are for 100,000, some are for up to a million uh or build data centers. So really scalability is uh kind of a key aspect to this.

SPEAKER_01 11:49

Yeah, and that the the the um global uh customer base of commercial companies is really interesting.

SPEAKER_02 11:58

It's just an a really impressive span of applications uh that uh that are that are foreseen. Um and we talked a little bit about the fact that there are different configurations of the the space systems to to accommodate uh these different applications. Um so welcome. Well let's let's talk a little bit about some of the factors uh that uh that present trade-offs um in trade space uh for the developers. You've you've mentioned um Andrew, um, you know, for launch and for operations. What are engineers that are developing these ODCs? What are they grappling with?

SPEAKER_03 12:38

Yeah, uh so the design space is pretty interesting. Um as Colleen mentioned earlier, you know, a lot of it with scalability is it comes down to power, right? So a lot of these orbital data centers from an astrodynamics perspective are trying to take advantage of these sun-synchronous orbits, right? Where the uh the orbit is kind of always exposed to the sun. And that's a very unique set of orbital real estate, right? Uh but if you deploy large amounts of solar panels in those particular orbits, there's real key advantages, right? Solar panels on the ground, they you know, they they have issues with cloud cover, they have issues with dealing with nighttime, right? You're not producing a lot of power or any at night. Well, you don't have those same issues during the day. So deploying solar panels up in those particular orbits possibly makes a lot of sense. It always does come back to the economics, though. So one challenge of getting that amount of solar panels up in orbit is launch, right? And so there is potentially a real launch bottleneck. Um, so the FAA has been um has publicly stated that they're motivated to increase the launch cadence from somewhere just over 200, 250 launches per year today to about 1,000 launches in 2030, uh, in 2030 and maybe maybe beyond that in 2035, as many as 10,000 potentially. Our current launch infrastructure cannot support that right now. So we we really need to invest in things like uh better launch infrastructure at our current facilities to really increase cadence of launch, new launch facilities as well. Um, I know there's uh potentially a new facility in Louisiana, for example, and we probably need several others. Um new launch vehicles? Oh, go ahead, Colleen.

SPEAKER_01 14:35

Yeah, sorry. Um just to interrupt, um I'm I'm interested in these bottlenecks, to be sure. Um, and I know that um we have had a lot of um discussion about launch cadence in this country, but what other kind of engineering um I don't know if they're bottlenecks, engineering feats do people need to kind of overcome in terms of ODCs?

SPEAKER_03 14:57

Sure. Okay. Well, uh we can come to the bottlenecks later. Um in terms of the other engineering issues, um so uh thermal management is a big deal, right? You're you're generating large amounts of power. Um honestly, well over 99% of that power turns into heat, waste heat, right? You got to get rid of it. Um on the earth, you know, you have uh air-cooled CPUs right in your computer right now that you're using to record this. Um that there is no air in space, right? So uh you really have to focus on having very good radiators. Radiators are heavy, so you know that goes back to the launch issue and up mass. Um they they but they one advantage of radiators is um they're really good at rejecting heat when the radiators are very, very hot, right? So you have deep space, it's very cold, it's three Kelvin. From engineers' perspective, that rounds to zero, right? And if you can run your radiator very hot, um the the heat rejection scales as temperature to the fourth power. So running it a little bit hot turns into a big benefit in terms of heat rejection. Um you can do things like um you can make them liquid-cooled radiators, you can uh maybe have a heat pump that moves heat um from the CPU to a radiator device, you can maybe invest in a new generation of chips that natively just can run at higher temperatures, right? So there's a big trade space that you can pursue there. But the bottom line is if you can make that radiator hot, then you can minimize the mass and really maximize the performance.

SPEAKER_00 16:38

I'd like to I'd like to add a little bit. Um, Andrew, that was excellent. Excellent insights. A few other areas that I think about, um, maybe outside of the physical plant and what have you, are um secure on-orbit enclave processing of data. So imagine a scenario where you have an existing uh constellation and you want to augment through services, compute, storage, and what have you through these ODCs, um you'll need to decrypt and process and re-encrypt data. And so on-orbit enclaves. So we'll have to you know think about that from a network architecture standpoint. Um the communication links and coordination protocols between existing constellations and what I call augmented ODC services or capabilities. Um today we have several protocols that make the internet work. Uh in space, we have um when we're trying to coordinate between different constellations or you know, different ODCs or existing investments in these ODCs, um we're gonna have to um you know employ and think about some open and interoperability um uh challenges uh to resolve. And then I also think about um space-to-ground cybersecurity assurance. Um just because the you know the processing is happening further away, it's harder to get there. It's not a reason uh to um uh to think that um you know we've addressed cybersecurity.

SPEAKER_01 18:06

Tim, you touched on um bandwidth a little bit. Is there are there concerns even like from from government and commercial in terms of the kind of throughput that might be needed for this edge computing?

SPEAKER_00 18:20

Yeah, I mean, um I think as um as the you scale, as you have more more assets flown and what have you, certainly bandwidth is a planning consideration. Um but I think about um if you're gonna truly use, if you're gonna, if you're gonna take processing and do it uh closer to where the data is captured, uh then you have to deal with not just downlink. You have to deal with um sort of the lateral network and comms concerns and what have you.

SPEAKER_01 18:50

So um would you call that a bottleneck?

SPEAKER_00 18:55

Um I'd say that it's uh it's a constraint that needs to be addressed. Um and and you know, when I think about this, um, you know, utilizing existing links and what have you to downlink, you know, these kind of foundations that that Andrew's mentioned, that I've mentioned, um they're they're kind of important because um uh offering ODCs kind of implies in the future that we'll be able to use traditional um CI C D, continuous integration, continuous deployment techniques, right, uh for um uh for ongoing operations. Um, you know, and and that's you know that um so those types of deployment techniques for uh software refresh everywhere from the anywhere from the operating system to the application and things like that.

SPEAKER_02 19:39

So all this discussion around scale uh prompts me to just make a a uh an observation from a back of the envelope calculation. Uh so if you if you take a million satellites uh that are 70 meters wide and place them edge to edge, shoulder to shoulder, uh it would wrap around uh the equator of the earth uh at about 1.7 times. So it's just you know something to try to help us to get our our minds around the scale of what we're talking about.

SPEAKER_03 20:11

That's funny, Ron. I actually did that calculation just the other day just to see what what it would be. Um so yeah, obviously they'd be touching. That's that's not good. Um other aspect to like there's a lot of aspects to the engineering here, but um, so because of your calculation, I think it's obvious you don't want to put all the orbital data centers in precisely the same orbit, right? So from an astrodynamics perspective, there is this tight, you know, sun synchronous um plane that you do want to populate, but you're probably gonna populate in various rings um or close to rings um at different altitudes, so to speak, right, to to avoid uh the traffic jam there. That being said, I think there are substantial space traffic management concerns that would remain, especially if you're trying to coordinate um, you know, your fleet with uh another company's fleet, right? Um that having that coordination is is absolutely a critical aspect to being able to kind of co-locate um these data centers.

SPEAKER_02 21:15

Absolutely. And in a recent discussion, uh it was mentioned that it'd be beneficial if the uh orbits were going in the same direction if the trans, you know, if they if the actual transit transfer was going in the same direction so that they weren't crossing paths. Something else to consider.

SPEAKER_03 21:33

Yeah. Um, another thing to consider is um so a lot of the current large constellations, like Starlink comes to mind, they're autonomous. They're very autonomous. And that is the only way at scale that you can really manage a fleet, right? So they make autonomous decisions, including maneuvers, including collision avoidance maneuvers. So what happens when you have one autonomous constellation parked next to or amongst? Other autonomous constellations. They can work kind of intra-constellation easily, but communication between constellations can become a challenge. And for example, who has the right of way? Who you know takes control and says, I'm going to do this maneuver, you don't do this maneuver, because if we both maneuver, we may maneuver into each other. Right? So those are some of the key space traffic management things we're working out.

SPEAKER_01 22:27

And what other kind of bottlenecks might exist in terms of being able to make these kinds of numbers reality? Are there things besides orbital capacity and managing the traffic? Are there terrestrial bottlenecks outside of launch?

SPEAKER_03 22:48

There certainly are some terrestrial bottlenecks. I mean, you you need to get all the inputs to produce all these orbital data centers, right? So, you know, just getting the raw materials, you know, for example, it's it's not always super straightforward to find pure silicon, you know, to create the AI chips. Um, if you're going to scale to a large number of chips, you you need to actually manufacture a large number of chips. So creating additional fabrication facilities is not trivial. Um, they they're very big, they're very capital intensive, and it takes a lot of super sophisticated engineering talent to get them to work.

SPEAKER_01 23:26

So we have just workforce, manufacturing workforce alone. This is my this is my my thinking.

SPEAKER_03 23:32

Absolutely.

SPEAKER_01 23:32

Tim, do you have anything to add about that? About bottlenecks in general?

SPEAKER_00 23:37

Yeah, you know, one thing that I think of that's helpful is to is to set some context when talking about this stuff. Um if your goals are to uh support a million, you know, constellation, a satellite constellation, uh even hundreds of thousands, um uh, you know, that's one set of concerns. Uh, and that's the scale that Andrew's mentioning. You know, the other thing we're thinking about is um the smaller, purpose-built, sort of augmented uh ODCs, uh ODCs that augment existing mission systems, um, and what have you. So those are the the concerns are very different, right?

SPEAKER_01 24:14

So you'd you'd look at can you give an example of what what you mean by that, the augmented um capabilities for existing systems? Just a yeah.

SPEAKER_00 24:24

So as I mentioned, as Ron asked about um uh different use cases, missile warning and missile tracking and uh space domain awareness. You know, um in the broadest sense, I could imagine, you know, hundreds of thousands of satellites helping. But that may not be necessary. Um, you know, what what what would likely, you know, there'll be two different kinds of at least two different kinds of deployment types. Uh one where um I have an existing um investment constellation that's doing a mission, um, and I need additional uh compute and storage services available in an augmented way um uh to offload. Because today, um, you know, as I mentioned before, one of the challenges is dropping data, right? The the hardware um and and you know, the uh like sensors and cameras and things like that, they're they're physical, they're backplane, uh, you know, even if it's just a gigabit uh backplane, um, that amount of data can't be sent back down to the ground. So it has to be filtered out. So that's a common practice today. So um, you know, so I can imagine um uh you know, sort of a purpose-built smaller ODCs, not small by any normal means, but not millions, you know, and in terms of uh having millions of them and what have you. Um, anyways, does that help answer Colleen?

SPEAKER_01 25:45

Yeah, no, that that works. So Ron, there's also you have some insight into some of the um bottleneck, bottlenecking, is that a word, um, around testing facilities.

SPEAKER_02 25:57

Yeah, we we've we've certainly seen that you know, while there's um a number of test facilities across the US and with our allies, uh, relative to the demands, uh, there is even a higher demand uh than there is supply in those kinds of capabilities. Uh certainly welcome going, Tim, over to you to to expand on that from your experience.

SPEAKER_00 26:19

Yeah, um, Ron, what I'd like to share is a little bit about other areas uh that I see. And then I think Andrew uh can share a little bit more about the testing and and launch. Um spacecraft design, you know, therm power thermal radiation uh for operational stability at scale, largely on proven area. And uh so we have work to do there. Uh supply chain readiness for radiation ready, GPUs, CPUs, storage, and other component parts that go into ODCs. Um also tech refresh, uh, you know, timelines, logistics, um, you know, how do you swap out parts um and what have you? Um open interoperability standards between heritage constellations and and ODCs, and even ODCs uh that you know uh different ODCs um uh that need to communicate with each other on on behalf of the customer. And then also policies for um data sovereignty and other legal legal constructs. Um, you know, those are areas where I think it's um you know that you know we could spend some time studying and and helping uh you know uncover uncover what needs to happen uh to make the ecosystem work.

SPEAKER_03 27:32

And just just some other thoughts. Um it a lot of it keys bottlenecking back to launch, but uh in a slightly different perspective here. Um so I'm a huge proponent of ground tests, but at some point it may not be economically wise um to do all of your testing on the ground. Sometimes it makes sense to get a prototype up in space. And we've actually seen that recently um where there's been a couple of prototypes that were launched on typically these rideshare missions where you're deploying, you know, dozens, maybe a hundred different objects from different uh companies at a time. Unfortunately, there seems to be a new bottleneck uh appearing where you know, because launch is becoming such a constraint that launch capacity is being vectored towards um building out large constellations and these um rideshare uh missions that you know facilitate a lot of prototyping are finding it very difficult to create future missions and get real rides to space. There are some um newer entrants to the launch market that are trying to kind of step into that niche and fill that void, but a lot of that's not yet proven. So, you know, that's just another bottleneck that we have to monitor.

SPEAKER_01 28:44

Yeah, and then Tim mentioned the kind of replacing parts. So in terms of replenishing, I can just see the little robots like taking out certain GPUs, bringing them back, putting in new ones.

SPEAKER_00 28:56

Well, and you know, I just to put a bow on it, I um, you know, when I when I think about these problems, if if we don't set the deployment type, whether it's a um I'll call it a mega uh constellation, millions of satellites, what have you, or a smaller, augmented, special purpose built, um, it it really matters. It matters how we talk about each of these areas. Um, maybe not every one of them, but um uh, you know, it's um you know, this as you as you scale, and in my experience and my career, um you have a whole you have a couple different, you have other concerns, right, that that arise. So I think making sure that the deployment type is that context is honed in on is important uh when we're having a public dialogue about it.

SPEAKER_02 29:40

We've been talking at a high level about a number of challenges that need to be overcome uh from a technology standpoint. Uh we haven't talked so much about the regulatory challenges and and essentially opportunities. Uh maybe good to delve into that regime uh for for a bit as well. What what are what are some of the um additional challenges, Andrew, that you've been seeing for space traffic management and some of the regulatory processes and practices that need to be uh experienced and and understood more fully as we go forward?

SPEAKER_03 30:19

Absolutely. Um there's several of them. Um, and I think this is kind of natural growing pains for an industry that's expanding this quickly. Um but kind of starting on the earth, near the earth, um, you know, we keep coming back to launch, but there is um, you know, integration that has to occur between air traffic management and space traffic management. So even something as seemingly trivial as, you know, as the launch rates pick up, we need to reserve airspace or do something about airspace to permit a high number of uh launches per year and not mess up uh uh air traffic too badly, right? Not maybe do it on Thanksgiving, the busiest air traffic day of the year, and right back up all the airports around the country. Um, those are real uh growing pains that we need to work through. Um so then moving forward, uh, you know, as we're going through launch, um, there's a lot of launch collision avoidance uh concerns that you have to be cognizant of. As you really populate out those orbital shells, it becomes increasingly difficult to try to thread that that rocket needle and avoid other space objects on the ascent. Um then when you deploy the objects, um there's this um kind of uh trackability and identification gap that needs to be filled and really needs to be considered. Um, you know, investing in technologies that can more quickly and more reliably communicate the position and identity of these large amounts of objects is just going to be very critical to go from the air integration into the space integration. Of course, you have your on-orbit life cycle. I touched on that a few minutes ago with autonomy and collision avoidance procedures, um, but then also be cognizant on the way back down, you know, most of these objects are going to re-enter. A few of them may have a controlled re-entry with even a potential recovery, but the majority of them are going to, you know, intentionally dispose and burn up on re-entry. And so, you know, when it comes to the objects that are recovered, there's a license that's needed to recover an object to make sure that it's re-entering airspace and and also still avoiding air traffic to maintain safety. For objects that are disposed, um, you know, we want to make sure that um as the number of objects grows, we don't have issues where you know some objects um hit infrastructure on the surface of the earth and damage property or even, you know, god forbid, hit people. And so there's certain metrics that have to be hit with that um as well. And then, you know, finally, um, we have to be cognizant of uh, you know, the effect on the upper atmosphere. A lot of this material is kind of ablating away and turning into essentially smoke, and those particles are kind of percolating through the atmosphere, and you know, we still need to study the effects of you know, what does that do to the atmosphere? Are there any ozone interactions, for example? Um, are there any things that we could typically ignore with smaller numbers of objects? But now that we're getting into larger numbers, we really need to be cognizant of. And oh, one last thing that I just thought of. Um humanity has never operated a hundred thousand active objects in space or a million. Um, and so when you're looking up from the surface of the earth into the night sky, you know, you you can see uh satellites. And so um it we have to be cognizant of the effect to astronomers, especially where you know they're they're gonna be trying to execute their astronomy mission and they're just gonna have you know a significant number of space objects fly through their field of view and potentially ruin their science. There are good ways to handle that, um, but that still is uh maybe not perfect solutions there. So we have to do a lot of active research and mitigation uh to decrease that light pollution.

SPEAKER_02 34:20

All right, Andrew, you've given us uh a few things to think about there, a few a few engineering challenges to work through. Tim, uh over over to you. What what are some of the key uh issues relative to being able to support national security applications?

SPEAKER_00 34:36

Ron, um really good question. For national security space solutions, some of the metrics of note or that we've been studying are latency, resiliency, and accuracy. When it comes to latency, we think about uh reducing free space transmission time or eliminating the need to send data to and from the ground. Um we also think about in latency reducing the kill chain data processing delays by distributing compute uh from centralized ground-based models to ODCs on orbit. And then as data processing matures and we're more confident, um, we think about uh the potential for reducing human in-the-loop delays with trusted AI and decision making. When it comes to latency, I think of um uh maintaining kill chain closure with loss of ground uh C2 command and control or loss of individual ODC nodes. And and also uh providing alternative paths to ground uh to uh for data routing to ground. Um and finally, for accuracy, increasing uh track accuracy by reclaiming OPIR data, um, overhead persistent uh infrared data uh not currently downlinked, and um also increasing the probability of identification by implementing threat typing algorithms with reclaimed OPIR data andor fused uh data on orbit from multiple sensors. Back to you, Ron. Fabulous.

SPEAKER_03 36:08

Just one thing to add to that. Um I just want to make an observation that um all this space-based infrastructure is kind of uh feeding into one another and working as a big system, right? So Tim mentioned a bunch of different aspects, resiliency, um, you know, uh moving data around, um, satellite laser interconnects, a whole bunch of things. Um, and so there's systems that are dedicated entirely to that that have you know already been deployed in space. A lot of those are going to be leveraged by these orbital data centers. And so what you're seeing is really just another layer integrated on top of the infrastructure that's already being built out.

SPEAKER_02 36:52

Thank you. And and Tim, I I understand that you're working on uh a paper, uh Center for Space Policy and Strategy paper. Can you just outline the scope of that?

SPEAKER_00 37:02

Yeah, Ron, I'd love to. You know, this year we've worked on a study, um, uh, and uh some of the insights that I've shared are based on that. Uh and we are looking to publish a paper, but um, it's on specifically on ODC's orbital data centers for national security space. Um, it's a place to start. It's it's um you have to bound any problem that you go after uh in order to do it well. And um, you know, aerospace, we're we're an independent technical advisor, and um, you know, our customers are asking what to do, what to do with um uh these types of solutions, these future solutions, and and how to plan. So what I can tell you is what the paper is not about, all often the most important thing. You know, it's this paper is not about um hardware and engineering, uh thermal, avionics, uh processors, very important topics, uh, but just not the scope for this paper. Um, as I mentioned, it's for national security space, it's not necessarily geared towards uh civil space commercial or intel collection and and so forth, uh intelligence community collection techniques. Um and you know what it is, it's it's purely uh uh a mission-driven question that we're trying to answer. Does this technology do ODCs, can they help national security missions? And I'd I'd also say um, you know, why now um you know policymaker interest is is heating up. As I mentioned, we're getting some questions uh from from folks, and and we think the the DOW needs to have a position on ODC soon. Um and uh and then also we want to, while this paper will be largely qualitative in nature um by design, uh we are following up and interested in following up on more rigorous analysis. Okay. And um yeah, yeah, thanks. Thanks, thanks, Ronna. We're excited about that and hope to publish that soon. Looking forward to it.

SPEAKER_02 39:00

Colleen over to you.

SPEAKER_01 39:02

Yeah, well, I'm I'm looking forward to it too. CSPS is putting out a paper uh written by our our uh one of our guests here today. So um we will definitely match this show and that future paper together uh when it's released publicly. I wanted to ask each of you, and Ron, I'm gonna get you too. Um, but I'll I'll start with you, Andrew. Is there something that what do you think that we really need to leave the audience with? Or is there something that maybe we didn't cover today that you would like to cover?

SPEAKER_03 39:31

Um, I mean, there's always there's there's so many little things that we could kind of wrap into this conversation. It's it's a really big integration question, but I think the real fundamental uh question for me comes down to the launch bottleneck. And I know I mentioned that many times, but here's the angle that I'm gonna throw at you. Um, really looking at uh the economic viability of these orbital data centers for the particular use cases that we mentioned, a lot of it comes down to um getting launch more reliable and a lower cost. So, really um, you know, I hate to boil it down to one number, but the the dollars per kilogram is always kind of one of the first metrics that people watch for. And a lot of the economic viability really hinges on having that number be a really low, much lower than it is today. And so, you know, the big question in my mind is how low can that practically go? And so that's something that I'm gonna be following for sure over the next couple of years.

SPEAKER_01 40:29

Yeah, that's a really, a really good point. I know um closing that kind of business case is important for commercial. At what point does it actually start to pay off? Um, not fighting with um your terrestrial neighbors over uh power and water. Right. Tim, how about you? Is there um what what do you want to leave the audience with?

SPEAKER_00 40:51

Yeah, Colleen and team, uh there are two points. You know, one is um applications drive requirements. Um, and then the other one is the deployment type matters. So on the application side, you know, given the amount of commercial investment happening, I think it's prudent for us to start thinking about and helping our customers uh determine uh what types of government applications could take advantage of ODCs. Um one thing that aerospace does, we do modeling and simulation really well. We get ahead of problems uh to help characterize uh um, you know, and so forth. And that help characterize the art of the possible. And that helps everybody. It helps our customers, helps industry, and so forth. You know, the second one is the deployment type, you know, whether, and I mentioned this a little bit earlier, but whether um extending terrestrial capabilities or enabling better process data processing that originates on orbit, um, I'd encourage a dialogue to distinguish between the you know million um uh constellations, I'll call it a mega constellation for lack of a better term, uh, with more um smaller, uh more purpose-built ODCs. Um getting that context right, framing in the in the conversation is kind of important as we get into details.

SPEAKER_01 42:05

Interesting. So kind of looking at that um instead of the quantity, the sheer numbers, but looking at the quality of smaller, tighter kind of formations, if you will.

SPEAKER_00 42:14

Yeah, and and whether whether just what I'm encouraging is just these two different deployment types, because when you peel through the details that we've we've talked about today, and if you're talking about a million satellites versus um a hundred or a couple thousand, um the the technical details uh and so forth, the trades, they become different.

SPEAKER_02 42:37

So your your prompt um cause calls me to uh to reflect on the flight proven paradox, right? Um everyone when it comes to space would like to be dealing with a system that is already proven in space. Um these ODCs is an area is an area that uh has that in the future. We we only have relatively modest examples of edge computing and uh and AI systems that are operating in space today. So I look forward to uh to us uh you know accelerating the deployment of ODCs and using the initial one or ones um as test beds and proving grounds back to that point, uh, so that we can build some experience. We have the benefit of some experience uh to deal with all of the wonderful challenges that Tim and Andrew raised and the span of application and determine uh the performance that can actually be realized uh for the span of applications that were discussed.

SPEAKER_00 43:44

Yeah, and I'd like to add, um first of all, thank you for inviting me, uh Colleen, Ron, and Andrew. Uh it's been it's been great to collaborate with you and continue to in the future. You know, this topic area is it's extremely interesting to me, you know, as the space industry utilizes and seeks uh to further utilize uh distributed computing techniques, right? And to help to help um the US um maintain and accelerate our space advantage. So definitely an exciting area and uh hope to get to do a lot more work in in the future.

SPEAKER_03 44:16

It's a very exciting time to be in the space industry. So I'm glad I can be a small part. Yep.

SPEAKER_01 44:22

Well, thank you everybody for joining today. Um it was as usual, I learned a lot and it was a good discussion. And I hope that uh the audience enjoyed um everything that you brought up and they are inspired to look into more things. And we will definitely be um uh releasing the paper that aerospace is working on uh in tandem with this show. So to our audience, um, you can watch and listen to all of our shows just about anywhere. And to get notified, you can subscribe at csps.aerospace.org, or of course, you can follow us on social media. So see you next time. Thank you.