Do Data Centers in Space Make Sense?

Starcloud already has an Nvidia H100 in orbit, as Mark points out while watching Philip Johnston's TED talk. They put it on the back of a Falcon 9. It's up there.

Clips and outtakes on Instagram ›

In this episode · 12 chapters

  1. Are we allowed to do this?
  2. Being on the TED stage
  3. Where we are today
  4. A megawatt in orbit
  5. The politics of space
  6. A creeping backlash
  7. He just swore on TED
  8. Another space scam
  9. There is an H100 up there
  10. A bet
  11. The projected cost
  12. The tie-in at the end

Mark and Jeremy watch Philip Johnston's TED talk on data centres in space, given in San Francisco in October 2025, as preparation for going live for the Artemis launch. Jeremy credits Johnston and Elon Musk for the ground swell behind space-based data centres since then.

They weigh the power and cooling load behind a megawatt of compute in orbit, the backlash against hyperscale data centres on the ground, the trolls who call it another space scam, and Johnston's projected launch cost against the lower figures in Space to Grow.

Philip Johnston's October 2025 TEDAI San Francisco talk, Let's build AI data centers in space, makes Starcloud's case for moving AI data centres into orbit. Reacting to it on the Thinking On Paper podcast, Mark Fielding and Jeremy Gilbertson back the case, noting that Starcloud already has an Nvidia H100 GPU in orbit, but name cooling as the problem that is still hard.

Do data centres in space make sense?

Mark and Jeremy back Starcloud's case, noting that it already has an Nvidia H100 GPU in orbit. They name cooling as the problem that is still hard.

Does Starcloud already have an AI chip in orbit?

Starcloud has an Nvidia H100, a data centre grade AI chip, operating in orbit after launching on a SpaceX Falcon 9, Mark says. Jeremy contrasts Johnston with founders who are only “great storytellers and they're great fundraisers”, saying Starcloud is building hardware that is “already up in space”.

What launch cost makes solar power in orbit break even?

Launch cost is the number to watch: Jeremy summarises the talk as saying that at about $500 a kilo, solar power in orbit breaks even with power on the ground. He calls it “a cool equilibrium” to track.

Which objections to space data centres does Jeremy accept?

Of the objections Johnston reads out from online critics, Jeremy rejects the connectivity one, because data reaches Earth faster through space than through fibre optic cable, but accepts that cooling is real and difficult. Energy, he says, is answered by the sun.

Why are space-based data centres still in test mode?

Space-based data centres are still in “micro test mode”, Jeremy says, because a megawatt of IT load, the power drawn by the computing equipment itself, comes with two megawatts of power and cooling load to support it, and that cannot simply be lifted into orbit yet.

What does local opposition to data centres target?

Local opposition to terrestrial data centres is aimed at data centres in general rather than at AI specifically, Jeremy argues, pointing to Karen Hao's Empire of AI and its account of their effects on water and power grids. He describes the talk's closing predictions as most data centres in space within 10 years and a Dyson sphere, a structure built to capture a star's energy, within 50.

Where Mark and Jeremy disagree

Mark questions the premise that all data centres will go to space: “Have you seen what they're building on the ground?” Jeremy answers that the industry is proving the model at small scale first, and both close by backing the idea against its online critics.

Questions

Do data centres in space make sense?

Mark and Jeremy back Starcloud's case, noting that it already has an Nvidia H100 GPU in orbit. They name cooling as the problem that is still hard.

What is Philip Johnston's TED talk about?

Philip Johnston's October 2025 TEDAI San Francisco talk, Let's build AI data centers in space, makes Starcloud's case for moving AI data centres into orbit.

What do Mark and Jeremy think of Starcloud's case?

Reacting to it on the Thinking On Paper podcast, Mark Fielding and Jeremy Gilbertson back the case, noting that Starcloud already has an Nvidia H100 GPU in orbit, but name cooling as the problem that is still hard.

Who does Jeremy credit for interest in space data centres?

Jeremy credits Johnston and Elon Musk for the ground swell behind space-based data centres since then.

Transcript

Mark Fielding Disruptors and curious minds, welcome to Thinking on Paper, the technology podcast for those who think. We're doing a reaction video. The plan is to go live for the Artemis launch and we need to do some prep for that. So we've decided to choose Philip Johnston's

TED data centers in space from San Francisco from October of 2025.

Jeremy Gilbertson So this could go one of two ways, friends and neighbors. This could be something really cool or it could be a bad, bad version of Mystery Science Theater 3000. So maybe somewhere in between.

Mark Fielding It's not going to be bad. It's Philip Johnston in a leather jacket and white trainers on stage at TED talking about data centers in space. It can't be. It's not going be that bad. He does look sharp, doesn't he? He looks like he's been shopping at the same shop as Mr. NVIDIA.

Jeremy Gilbertson He looks good. He looks sharp. Yeah, he looks sharp.

Yeah, he is giving a little Jensen vibes right here. Little Jensen vibes. All right, let me get this where it needs to be here.

Mark Fielding It is. I like the trainers. I like the trainers.

Jeremy Gilbertson beginning. Yeah.

Are we allowed to do this?

Mark Fielding Do we need copyright something for this? we allowed to do this? Are Ted gonna kick us off YouTube?

Jeremy Gilbertson I'd rather yeah, let's ask for permission later, man. All right, off.

Mark Fielding something out of the OpenAI playbook there, Jeremy? Ask for permission later.

Jeremy Gilbertson ⁓ definitely not the open

AI playbook. No, I don't want to be associated with that one. All right, here we go.

Mark Fielding ⁓ come on. Yeah I like that.

Jeremy Gilbertson So, oh yeah, a little dad joke there. That's okay. That's okay. But can you imagine though, Mark, you're on stage number one, you see the Ted timer down there, 1439, you've practiced your stuff, you're ready to roll, but you're also like, holy crap, I'm getting ready to like put something into space that has never been put into space before. And you have that kind of resting in the back of your mind too. That's a lot, a lot to think about. A lot to think about.

Being on the TED stage

Mark Fielding Yeah, yeah. And being on stage at Ted, it's, it's a big, a big gig, isn't it? There's going to be some nerves. There's going to be some, got to, you to start with a joke, don't you? So good on him.

Jeremy Gilbertson Yeah, 20 seconds,

21 seconds in we get a dad joke. So I love it. Good job, Philip off we go.

Where we are today

Jeremy Gilbertson All right, let's stop right there. So we are, the date today is February Since October, some pretty key individuals have been echoing the same sentiment that Philip has said back in October and also ⁓ what was previously echoed by Gerard O'Neill to a certain extent back in the 70s that we could make use of space the ground swell for space based data centers since October has been monumental and it's all down to two men, Philip Johnston and Elon Musk, who in October time, he started tweeting about this and sometimes tech Twitter feels like a game of who can impress Elon Musk the most. in as he posted about it.

There's just this surge of, yes, we're going to put the data centers in space, data centers in space, data centers in space, Nvidia on board. As he said, the H100 chip is up there.

Mark Fielding one thing that I just keep pushing back on me is we're to put all the data centers in space. Have you seen what they're building on the ground?

Jeremy Gilbertson That's yeah, it's it's it's one of those things. It's a mindset thing, right? It's it's it we're in we're in micro test mode with this we're in not minimum viable product, but we're in like minimum viable test, right? And we're kind of proving this model.

Mark Fielding would agree with that? What you just said?

A megawatt in orbit

Jeremy Gilbertson You will think about it. can't can't put like a megawatt of IT load into space right now that has two megawatts of power and cooling load to support the one megawatt of IT load, meaning the compute it. You can't just lift these things.

Mark Fielding H100 do? What's this? Nvidia H100? There's one? What can it... What are we looking at? Do you know?

Jeremy Gilbertson chip by chip. I'm not, I'm not sure, but it's a very small processing compared to like what the terrestrial data centers are doing today. All right, let's keep

The politics of space

Mark Fielding The politics of space. was just thinking, what's he going to say? What's he going to say? What's he going to say? What's the stick? What's the stick? And I wasn't expecting him to bring that up. So yeah, I like that. We've been speaking a lot about the politics of space. And of course he's right.

Jeremy Gilbertson Okay, sorry, I gotta do a back to the future reference. 1.21 gigawatts! Yeah, all right, so yeah, that's an ass load of power. Moving along. It is, yep.

Mark Fielding It's that technical term, okay?

Jeremy Gilbertson All right, so pausing right there. Yeah, for sure. We'll go, just to call back to Empire of AI, Karen Hao we talked about, there was a whole chapter, I think, dedicated to the effects of these massive hyper-scale data centers in these communities related to water, related to power grids and all of that stuff. that's a real thing. That's a real thing that's happening and a solve that needs to happen. But on the flip side, my mind is starting to think about, so,

Mark Fielding Good on Tuscon, Arizona.

Jeremy Gilbertson Yeah, that's definitely something. But then you also think about the impact and cost of launching things off the earth and the effects of that as well. ⁓ them in space, build them on the moon. You don't need to launch them, you build them up there, don't you? You get the humanoid self-replicating robots to make the data centers in space.

Hey, all we need is 12 metric tons. 12 metric tons.

A creeping backlash

Mark Fielding So is there also a creeping backlash against AI in general? is there more? I don't know what happened in Arizona, but is there other people starting to rise up and say no to AI and no to data centers? Or is it just the environmental impact?

Jeremy Gilbertson This, I think this specific thing is more data center focused rather than AI focused. I mean, I think both are high ⁓ impact news pieces, right? But I think that, I don't think it's getting as granular as data centers specific to AI that are the challenge. I think it's data centers in general.

Mark Fielding He jokes, but it's a big bit of kit.

Jeremy Gilbertson Yeah, yeah, that's it. That's a good, good scale there. And like, where would this sit to he might be getting to this, but like we're talking about in our in our book club space to grow, we're talking about orbital debris, and we're talking about lower Earth orbit geostational orbit, we're talking about the graveyard orbit that's above geo, like, it's interesting to think about, like, where all this stuff lands in the context, but good, good to know the sun gets through it.

Jeremy Gilbertson No, not from us. We're proponents. And hey, dude, really cool game changing ideas are going to get their asses kicked in the beginning, man. Like, you know, we always talk about the calcified systems want to perpetuate themselves. And, you know, the, the, the idea that, you know, something like this is going to get poo pooed on by people building stuff the normal way. Dude, hang in there, Philip. You're kicking ass, man.

Mark Fielding It's always a zero sum game for so many people, isn't it? It's this or that. You can't have this because this and you can't have that because this is bad. And it's not like that. It shouldn't be like that.

Jeremy Gilbertson agreed. Let's rewrite the book. ⁓ Moving on.

He just swore on TED

Mark Fielding He just swore on Ted.

Jeremy Gilbertson my gosh, I gotta pause this right here and let's like, we can maybe just look through these together. This is amazing. And yeah, he just said shit on the TED stage. That's amazing.

Mark Fielding I wonder if he's the first. This is ridiculously expensive and won't work due to connectivity issues, cooling issues, energy and mass. Okay, that's quite an educated troll. Even if it's wrong.

Jeremy Gilbertson Ish so so things move faster in a vacuum, right? So like the connectivity one is bullshit right there because you know, the stuff can get back to Earth quicker than it goes through fiber optic cable. So right off the bat, you know the connectivity piece, the cooling issue is real. The cooling issue is difficult. It's a tough challenge to solve the energy. You got the sun right there, right? And if we're figuring out, you know, solar rays in space, right? Let's see what else.

Another space scam

Mark Fielding yes, another space scam. So we see this in microcosm on our feeds all the time. People just saying it's a scam. ⁓ This sucks. What a waste of money. Aha. What a horrible idea. Where will the heat go? Stupidity on megawatt scale. I quite like that. Stupidity on megawatt scale. Yeah.

Jeremy Gilbertson should, they should, they should put that as their tagline on their website, you know, just to, just for a week and just see, see what happens. Like it's so easy to, to shit on new ideas, man. And it's so easy to ⁓ be a detractor in, know, the trolls come together, setting piles of money on fire in a parking lot. We're doing that already. Hey, author of that comment. We are doing that already with terrestrial data centers, like for real.

Jeremy Gilbertson Should we move on? we let him unpack this? All right, he probably does a better job than us.

Mark Fielding dad joke reprised in the trolling comments.

Jeremy Gilbertson Yeah, there you go. We just talked about that. 100%.

Jeremy Gilbertson Hell yeah, dude. that's like, all right, yeah, cause a lot of folks talk about doing really big things. we could do this and we could do that. And they're great storytellers and they're great fundraisers. Yeah, so he's building some stuff and it's already up in space as we're sitting right now. It's up there.

There is an H100 up there

Mark Fielding It's up there. There's an H100 up in space, isn't it? They put it on the back of a Falcon 9. It's up there.

Jeremy Gilbertson Good on you, man. Yeah.

Mark Fielding we spoke to Philip Metzger last week, didn't we? What was he saying? That the plan is to have one launching every two minutes or that that would be how many you'd need to launch to keep up with the scale that they're predicting.

Jeremy Gilbertson keep telling my kids this is gonna be like new the amount of flights that come out of, Hartsfield Jackson Airport or Heathrow, whatever, like per day, this is gonna be a very similar

Jeremy Gilbertson I'm holding up the high frontier.

A bet

Mark Fielding Do you think, okay, let's, I'll make you a bet now. Is he going to say Gerald O'Neill or the high frontier in the next two minutes? Yes or no? Okay. Okay.

Jeremy Gilbertson No, no, he led with Asimov. He's going to stay there. Let's real quick, let's pause there. So that's it. That'd be a cool equilibrium to kind of track, Mark, metric. launch costs hit 500, that means we're at break even on what's happening terrestrially from a solar power perspective. I think that's the important differentiator, right?

The projected cost

Mark Fielding Yeah, and also his projected cost there, we've heard a lot lower than that in space to grow. We've read a lot lower than that. I mean, it's already down to 14, 15, 1600 a kilo. If I would expect it to be lower than that, maybe two, 300 a kilo in my expert opinion. It's just hedging his bets. I think it's good.

see where he lands. ⁓ He lands this reusable rocket.

Jeremy Gilbertson Huh, that's it. Didn't know that. That's pretty cool.

Jeremy Gilbertson That is freaking rad, and timely and yeah, wow.

Mark Fielding Full circle things coming full circle. Yeah.

Jeremy Gilbertson at all the lady hats in this image. So many lady hats. We gave more of a shit. Like just when people like, we're both in hoodies. I'm literally wearing shorts and socks right now. Like, remember when used to go to the airport, you used to feel like you needed to dress up because it was like a big thing.

Mark Fielding We were better dressed back then, weren't they? People were much better dressed. Look at us! Look at us! What are we doing, Jeremy? Look at us! We would never get invited there.

Mark Fielding You know, you gotta bring, gotta call back.

The tie-in at the end

Jeremy Gilbertson Philip, dude, that was, I loved it, tie-in there at the very end. So in 10 years time, he's predicting most data centers will be in space in 50 years time. We've got the Dyson Sphere happening. ⁓ Couple of mic drop predictions. What'd you think about the UN tie-in at the very end?

Mark Fielding I think that the geopolitics of all of this, the more we learn at Thinking on Paper, the more people we speak to.

it will all come to a head as things usually do over human nature and human politics. And Philip's right. So if we can get the energy right before the real politics comes to a head, then that's perhaps the way to do it.

Jeremy Gilbertson Well, this dude, this was fun. I think we do more reaction videos. Tell us what you think about this guys and Ted, hopefully you don't send us a cease and desist because we're just looking to elevate your content. was so brilliantly executed. Mark, final thoughts.

Mark Fielding yeah, that was Philip Johnston speaking on the TED stage about data centers in space and star cloud and their Nvidia H100, which is now up there doing GPU things in space. That was our first reaction video. There will be more. I like it. I felt like I was learning something then. It was nice to watch that first time live with you. So we'll be doing more of these in the meantime, thinking on paper. X, Y, Z.

like and subscribe, share this with a curious friend, share this with a naysayer, share this with somebody who says stop building stuff in space, stop talking about building stuff in space, we've got problems here on earth, share them this, tell them to come and think on paper and until next week when we're back down the quantum rabbit hole and we have GRU talking about hotels on the moon, we'll stay, no until next week, be disruptive, stay curious.

Jeremy Gilbertson Keep thinking on paper.