Data Centers in Space 🪐

Instead of our usual rundown, we're going deep on one story — the kind that sounds like sci-fi until you realize the money is already moving and the rockets are already being built.

Data centers in space. What they are, how they work, who's building them, and whether it's genius or a very expensive way to create new problems.

No headlines. No bullets. Just the full picture. Let us know if you want more of this format.

Reminder: every time you use Viro AI, you're funding renewable energy and climate restoration. No extra steps. Just use it like you'd use any AI — and something positive, real, impactful happens in the world.

→ Viro AI (Web)
→ iOS app
→ Android app

Okay. Let's go to space 🚀

Last week, SpaceX went public in the largest IPO in history. The company priced at $135 a share, valuing it at $1.77 trillion. Musk owned roughly 40% of the shares.

You can do the math.

Elon Musk is now the world's first trillionaire. Not the richest person. Not even close to the richest person. The world's first trillionaire — a category that did not exist before June 12, 2026. And here's the number that puts it in perspective: the second-richest person on Earth, Google co-founder Larry Page, is worth $291 billion.

Elon Musk could lose a trillion dollars and still be the richest human being alive by a comfortable margin.

This is a scale of wealth that has no real cultural reference point. The closest analogy is probably a medieval king who owned the land, the roads, the ports, and the army — except Musk also owns the rockets.

Which brings us to what he's planning to do with them.

The thesis behind the SpaceX IPO — the thing investors paid $1.77 trillion to own a piece of — is not just Starlink satellite internet or rockets to Mars. The core pitch, repeated in the prospectus, in the roadshow, and in Musk's own public statements, is this: within a few years, the cheapest place to run AI compute will not be on Earth. It will be in orbit.

Data centers. In space. That's the bet.

And he's not the only one making it.

Before we get into the wild details, it's worth explaining what these actually are — because "data center in space" sounds like Star Wars science fiction until you understand the logic.

A data center, at its core, is just a building full of servers. Those servers process information — training AI models, running cloud services, handling your Netflix stream, storing your files. The problem is that they need two things in enormous quantities: electricity and cooling.

Modern AI data centers are staggering consumers of both. A single large facility can draw as much power as a small city. Cooling — keeping all those servers from overheating — accounts for 30 to 40% of total energy consumption, typically using massive amounts of water. Communities around the country have started pushing back on approvals for new data center construction. The power grid, in many regions, simply can't handle more.

So the pitch for space is elegant: both problems disappear.

In low Earth orbit, solar panels receive sunlight approximately 1,361 watts per square meter with zero atmospheric interference. No clouds. No night (if you're in the right orbit). No power grid to connect to. The energy is just — there. Continuous. Free after the upfront cost of the panels.

Cooling is also solved differently in space. Without air, you can't use fans or cooling towers. But hot surfaces radiate heat as infrared energy into the cold vacuum — passively, indefinitely, without water. The vacuum of space is, in one sense, the world's most efficient heat sink.

On paper: unlimited solar power, passive cooling, no land permits, no community opposition, no grid constraints.

SpaceX filed an application with the FCC in January to launch up to one million satellites as orbital data centers. Blue Origin, Jeff Bezos' rocket company (he’s the Amazon guy), filed separately for 50,000. Google announced a prototype constellation called Project Suncatcher, targeting 80 test satellites by 2027. Startups like Starcloud and Cowboy Space Corp (great name) are already building hardware.

This is not fake news.

It is an actual race, with billions of dollars and the world's largest rockets behind it.

Your inbox is flooded with newsletters. Your feed is chaos. Somewhere in that noise are the insights that could transform your work—but who has time to find them?

The Deep View solves this. We read everything, analyze what matters, and deliver only the intelligence you need. No duplicate stories, no filler content, no wasted time. Just the essential AI developments that impact your industry, explained clearly and concisely.

Replace hours of scattered reading with five focused minutes. While others scramble to keep up, you'll stay ahead of developments that matter. 600,000+ professionals at top companies have already made this switch.

Join them today, for free.

Here's where we steelman the space data center argument, because it deserves a fair hearing.

The energy math is real.
Data centers consumed 460 terawatt-hours of electricity globally in 2022. By 2026 that number could exceed 1,000 terawatt-hours — roughly the entire annual electricity output of Japan. AI is the primary driver. Every time someone runs a large model query, trains a new model, or processes video at scale, it costs electricity. The grid cannot keep pace with the demand curve that AI is drawing.

Solar power in orbit, by contrast, is effectively unlimited. The sun doesn't set on a satellite in a dawn-to-dusk orbit. There are no cloudy days. The efficiency of solar panels in space — without atmosphere to lose energy through — is substantially higher than on the ground. If the upfront cost of getting panels into orbit can be brought low enough (and SpaceX's Starship rocket is designed specifically to crater launch costs), the operating economics become compelling very quickly.

No water.
Earth-based data centers collectively use billions of gallons of water annually for cooling. It's one of the most underreported environmental costs of the AI industry. Communities near data centers in Virginia, Arizona, and Texas have started demanding water-use disclosures and fighting new construction permits. The orbital solution uses zero water — heat leaves as radiation, not evaporation.

No neighborhood.
Nobody lives in orbit. Data centers on Earth increasingly face community opposition — noise, traffic, power draw, visual impact, water competition. An orbital data center has no neighbors to upset.

Latency isn't necessarily a dealbreaker.
For AI training — the most computationally intensive workload — you don't need real-time data transfer. You batch the work, send it up, get results back. The latency hit matters less when you're not doing something interactive.

Musk has publicly claimed that orbital data centers will be the "most economically compelling" place to run AI compute "within three years." That's an aggressive timeline, but the structural logic is sound.

Now the other side. Because space is, as one researcher put it, "unforgiving."

Heat is harder than it sounds.
Yes, space is a great heat sink in theory. In practice, getting 10 megawatts of waste heat out of a satellite requires radiator surfaces comparable in size to two football fields. The chips generate heat. The heat has nowhere fast to go. Designing a satellite that can dump industrial quantities of heat via infrared radiation — without air, without water, with hardware that has to survive launch — is a genuinely hard engineering problem that has not been solved at scale.

Radiation damages everything.
On Earth, the atmosphere protects electronics from cosmic rays and solar radiation. In orbit, that protection is gone. Radiation causes "bit flipping" — random errors in computing hardware that on Earth are rare enough to be manageable but in space become a constant background problem. Every processor either needs to be radiation-hardened (expensive, heavy, slower) or running sophisticated error-correction software (compute overhead, more complexity). A solar flare can disable an entire constellation.

You can't call the repair guy.
Data center hardware fails constantly. On Earth, a technician drives over and replaces the broken server. In orbit, repairs are extraordinarily difficult and costly. Hardware that fails stays failed until the satellite is deorbited. This means either massive redundancy built into every satellite (expensive) or accepting degraded capacity as the constellation ages (problematic).

The latency is real for some workloads.
AI training can tolerate latency. AI inference — actually using a model in real time — cannot. The round-trip distance from low Earth orbit adds meaningful milliseconds to every query. For the kinds of real-time AI applications that are emerging — agents, voice assistants, live video analysis — the latency of talking to a server in orbit matters.

Kessler Syndrome is not a joke.
If SpaceX launches a million satellites, and Blue Origin launches 50,000, and Google launches its constellation, and startups fill in the gaps — the density of objects in low Earth orbit becomes a collision risk that compounds on itself. One collision creates debris. Debris creates more collisions. The cascade scenario, called Kessler Syndrome, could render certain orbital altitudes unusable for decades. Even SpaceX's own IPO prospectus warned investors that orbital data centers "may not be commercially viable" — a notable admission buried in the risk disclosures.

AI shows up in 60% of engineering work. Only about a fifth of it can be handed off without someone babysitting the output. That’s because agents are still missing the context you have. Join this live webinar on June 24 (FREE) to find out how teams pulling ahead are using a context layer to level up.

Register Now

Here's where we land.

The honest answer is that orbital data centers are a real idea with a real engineering foundation, backed by real money and real rockets, that faces real problems that have not yet been solved. Whether they become the dominant AI infrastructure of the 2030s or a cautionary tale about Silicon Valley's relationship with scale depends almost entirely on engineering execution — not on the vision, which is genuinely compelling.

What's notable from a climate perspective is the double-edged nature of the whole thing. On one hand: solar-powered AI with zero water consumption and no community footprint is dramatically better for the planet than coal-powered terrestrial data centers. If the space build-out goes well, it could represent a genuine environmental win for AI infrastructure.

On the other hand: launching a million satellites requires an enormous number of rocket launches. Each Starship launch, even at Musk's optimistic economics, has a carbon cost. Building the hardware — the radiation-hardened chips, the solar panels, the radiators — requires manufacturing at an unprecedented scale. And the debris risk is a real environmental issue, just one that plays out in orbit rather than on the ground.

The thing that gets missed in most coverage is that this race is happening because AI demand has outpaced the Earth's ability to supply the power to run it. That's the actual problem. The space data center pitch is a supply-side solution to a demand-side problem — and supply-side solutions, in tech, have a long history of creating more demand rather than solving the underlying tension.

More efficient AI servers didn't reduce AI energy consumption. They made AI more accessible, which increased total usage, which increased total energy demand. The same dynamic could easily play out in space: cheaper AI compute drives more AI usage drives more satellites drives more launches.

The only actual answer to AI's energy problem isn't to find a cleverer way to power the demand. It's to make sure that however AI is powered — on Earth or in orbit — the energy is clean, the environmental costs are accounted for, and the communities bearing those costs have a voice in the decision.

At Viro AI, we don't have a Starship. But every query you send funds renewable energy and climate restoration on the ground, right now, with infrastructure that already exists. Not in three years. Not after the engineering problems get solved. Today.

The future of AI might end up in orbit. In the meantime, we'll keep funding climate solutions from down here.

Read more

→ SpaceX IPO makes Musk first trillionaire (ABC News)
→ Google and SpaceX in talks on orbital data centers (TechCrunch)
→ Four things we'd need to make it work (MIT Technology Review)
→ The engineering challenges, explained (The Conversation)
→ Blue Origin and SpaceX race to orbit (Bloomberg)

Thanks for digging in. If you like this kind of stuff, let us know… we’ll do more.

Next time you think about opening ChatGPT, try Viro instead.
Same output. Better outcome.

Reply here or email [email protected]—we read everything.

– Nick