Rocket seats, not GPUs, are the bottleneck Starcloud is buying against. The Redmond, Washington company that runs AI data centers in low Earth orbit closed a $250M Series A extension on August 21, valuing it at $2.3B post-money, more than double its March figure and a $450M total since founding two years ago.
Manhattan West captained the extension, with Lauren Selig joining as a board observer. Benchmark, EQT, Soma, NFX, and 776 re-upped, while NVIDIA, Cisco Investments, Cedar Capital, Goanna Capital, and Standard Capital entered as newcomers.
Securing rocket capacity is a budget line in its own right, a reminder that the orbital compute thesis rides on launch availability. SpaceX controls most of that supply. Falcon 9 retires in 2028, Starship remains unproven, and rivals like New Glenn and Vulcan still fly irregularly. Starcloud has asked the FCC to authorize 88,000 spacecraft and wants Starship contracts signed as soon as possible.
The NVIDIA relationship behind the round predates the money. Starcloud’s first satellite carried an H100 into orbit in November 2025, the first data center-grade GPU to fly, and trained the first AI model in space. NVIDIA and Starcloud are also building a space-rated version of NVIDIA’s next-generation Vera Rubin platform, hardened for radiator-only cooling and radiation loads.
Two Starcloud-2 satellites, rated at 8kW each, are booked for rideshare launches in 2027, serving orbital inference work that includes US government agencies. Rival Cowboy Space raised $275M in May on a similar bet, keeping rocket upper stages in orbit as one-megawatt data centers.
