Starcloud Raises $170M, Bets on Orbital Data Centers
Starcloud closed a $170 million Series A led by Benchmark and EQT Ventures, valuing the company at $1.1 billion. Total funding stands at $200 million, 17 months after its Y Combinator demo day.
The pitch: data centers in space. The evidence so far is limited but real.
What They've Actually Done
In November 2025, Starcloud launched its first satellite carrying an Nvidia H100 GPU. The company claims it trained an AI model in orbit, calling it a first, and ran a version of Gemini. One GPU didn't make it: an Nvidia A6000 failed during launch.
So: one successful GPU on orbit, one failure, one claimed AI training run. For 17 months of work, that's a narrow track record to justify a $1.1 billion valuation. Investors apparently see past that.
The Roadmap
Starcloud 2 will carry an Nvidia Blackwell chip, an AWS server blade, and a bitcoin mining computer. It will also fly what the company says is the largest deployable radiator ever flown on a private satellite. Thermal management is the unglamorous constraint that determines how much compute you can actually run in a vacuum.
Starcloud 3 is more ambitious: a 200 kilowatt, three-ton spacecraft designed to launch from SpaceX Starship. The company targets power costs of $0.05 per kWh, but that number is contingent on launch costs reaching around $500 per kilogram. CEO Philip Johnston expects commercial Starship access to open in 2028 to 2029.
That's a lot of "if." The $0.05/kWh target requires Starship to hit a cost milestone it hasn't hit yet, on a timeline that has slipped before.
How Small This Market Still Is
Nvidia sold nearly 4 million GPUs to terrestrial hyperscalers in 2025. Advanced GPUs currently operating in orbit number in the dozens. The gap between those two figures is the entire business case for space compute.
U.S. data centers with more than 25 gigawatts of capacity are currently under construction, according to Cushman and Wakefield. The SpaceX Starlink network of 10,000 satellites produces roughly 200 megawatts total. Orbital compute is still orders of magnitude below what's being built on the ground.
The Competition
Starcloud isn't alone. Aethero launched Nvidia's first space-based Jetson GPU in 2025. Aetherflux and Google's Project Suncatcher are also in the field. Nvidia unveiled Vera Rubin Space-1 chip modules at GTC, though none had been produced or shared with development partners at the time of announcement. SpaceX has requested U.S. government permission to build and operate one million satellites for distributed compute.
The field is early and crowded. Most of these companies are announcing hardware that doesn't exist yet.
The Case For It
Power is the binding constraint for terrestrial AI compute. Data centers need land, water, and grid access. Orbit has sunlight and no neighbors to complain about the noise. The physics argument for space compute is sound even if the economics aren't there yet.
Whether Starcloud specifically is the company that closes that gap is a different question. For now, they have $200 million and a single working GPU in orbit.
Source: Techcrunch