Artificial intelligence hardware startup Starcloud Inc. has raised $250 million in new funding, valuing the company at $2.3 billion. Manhattan West led the round, with participation from more than a dozen backers, including Nvidia Corp. and Cisco Investments. The capital extends a Series A round announced in March.
From orbit to AI training
The funding comes nearly a year after Starcloud launched an Nvidia graphics card into orbit aboard a small satellite. The system completed the first AI training run in space, proving that data processing beyond Earth is possible. The company’s long-term vision involves deploying 88,000 satellites with a combined computing capacity of 20 gigawatts.
Orbital data centers could offer key benefits. They eliminate the need for complex cooling systems, and satellites generate power continuously using solar panels. In space, panels avoid energy losses from nighttime or weather disruptions like storms.
Most satellites currently send raw data to Earth for processing. Starcloud’s next system, Starcloud-2, aims to change that. Set for launch in 2027, it will carry AI chips, storage, and a data backup module to handle both training and inference in orbit. The company claims this approach could accelerate analysis by streaming processed results instead of raw data.
Scaling up in space
Starcloud has more ambitious plans beyond its second satellite. Starcloud-3, with a 200-kilowatt power footprint, will include graphics cards, cooling equipment, and heat dissipation gear—enough to power a few dozen homes. For context, the largest data center campuses under construction today consume energy equivalent to millions of households.
Related: NTT DATA and Palo Alto aim high
The most ambitious project, Starcloud-4, is designed as a cylindrical structure attached to a 2.5-square-mile solar array. It would house liquid-cooled servers in container-shaped modules, a setup the company believes could transform orbital computing.
To achieve these goals, Starcloud is expanding its manufacturing capacity. The new funds will support production lines for mass-producing Starcloud-3 satellites and secure launch slots. SpaceX is reportedly its launch partner, though the aerospace company is also developing its own AI satellite, AI1. With a 230-foot wingspan, AI1 matches the computing power of Starcloud’s next-generation system. SpaceX plans to begin mass production by late 2027.
This shift toward orbital data centers isn’t just about efficiency. For industries dependent on real-time data, like weather forecasting or disaster response, processing information in space could reduce latency to nearly zero. While that may not impact everyday tasks like video calls, it could be critical for tracking hurricanes or other time-sensitive operations.
Starcloud’s plans suggest a future where space becomes a hub for data analysis, not just collection. The timeline is ambitious, but if even part of it succeeds, the next decade could reshape how and where AI workloads operate.