Starcloud, which builds satellites that run AI inference in orbit, raised a $250 million Series A extension led by Manhattan West, doubling its valuation to $2.3 billion, the company said. Nvidia joined as a new investor with a $25 million check, alongside Cisco, Benchmark, EQT and other backers, according to Starcloud and Via Satellite.
The round brings Starcloud's total funding to $450 million since it was founded in 2024. Starcloud said proceeds will expand manufacturing capacity, fund engineering work with Nvidia and secure launch contracts for its next satellite, Starcloud-3, which is intended to fly on SpaceX's Starship rocket.
Starcloud has asked the Federal Communications Commission for permission to eventually operate a constellation of 88,000 spacecraft. Its first satellite, Starcloud-1, launched on a SpaceX Falcon 9 in November and was the first to run an Nvidia H100 GPU in low Earth orbit, according to the company.
Starcloud chief executive Philip Johnston told TechCrunch that launch capacity, not funding, is now the company's binding constraint, since SpaceX's Falcon 9 program is scheduled to end in 2028 and competing vehicles from Blue Origin, ULA and Rocket Lab are not yet flying at comparable volume. "One of the biggest costs is now on securing your launch capacity," he said.
A space startup treating rocket seats as scarcer than capital is a specific version of a problem showing up across AI infrastructure: the bottleneck keeps moving. First it was chips, then power, and now, for anyone trying to put compute somewhere unusual, it's the ride to get there.