Starcloud raises $250M for orbital data centers

An NVIDIA H100 is already running above Earth. Starcloud says its orbital data center has operated the GPU since its launch in November, using it to run Google’s Gemma large language model and train OpenAI’s NanoGPT. The company has now raised a further $250M, taking its total funding to $450M and its valuation to $2.3B, according to Payload.

The round was led by Manhattan West, with NVIDIA and Cisco among the participants. Their involvement points to a possible link between the equipment built for data centers on Earth and the hardware needed in orbit. Starcloud’s CEO, Phillip Johnston, said the H100 mission produced technical data valuable enough to influence NVIDIA’s Space-1 Vera Rubin Module, its first GPU designed for space.

That next chip is not ready yet. Starcloud hopes to fly it in late 2028, while engineers work through radiation shielding, the forces of launch and the operating temperature that determines how large the spacecraft’s radiators must be. The funding will also help expand Starcloud’s manufacturing footprint in Washington and advance its Starcloud-3 spacecraft.

So what changes in practice? Starcloud can move from a single demonstrated orbital GPU toward more spacecraft and a larger manufacturing operation, while giving NVIDIA real operating data for space hardware. This is still a prototype-stage effort, not a deployed orbital cloud service: the company’s experience comes from one known H100 in orbit, and the future Vera Rubin chip still has to be built.

The immediate constraint may be getting those spacecraft into orbit. Starcloud has asked the FCC for permission to operate 88,000 spacecraft in LEO, or low Earth orbit, a scale that would require hundreds, if not thousands, of launches. Payload reports that the funding also responds to growing concern across the sector about a looming launch shortage as SpaceX prepares to sunset Falcon 9 and pauses new rideshare bookings.