Starcloud, a startup building satellites that run AI inference in orbit, has tacked a $250 million extension onto the $170 million Series A it closed in March, bringing the round to $420 million at a $2.3 billion valuation. The capital is earmarked for a larger manufacturing plant and development of the Starcloud-3 spacecraft, designed to fly on SpaceX’s Starship. Chief executive Philip Johnston said the raise is also a hedge against a tightening launch market as the company prepares to request FCC authorization for 88,000 satellites.
Launch capacity is the real constraint
Johnston told TechCrunch that securing launch slots has become one of the largest line items in the budget. SpaceX plans to retire the Falcon 9 in 2028 and replace it with Starship, a vehicle that has yet to demonstrate rapid reuse. Competing vehicles from Blue Origin and United Launch Alliance are not flying on a regular cadence, and Rocket Lab’s Neutron remains grounded. Starcloud is targeting rideshare missions in 2027 for two 8 kW Starcloud-2 satellites bound for U.S. government customers, and is evaluating a dedicated Falcon 9 flight alongside contracts with other providers.
The Starship bet
The company’s long-term model assumes Starship will drive launch costs low enough to make an orbital inference layer competitive with ground-based data centers. Johnston said he remains confident SpaceX will prove the rocket can be reflown quickly and often. SpaceX chief executive Elon Musk said this week that an attempt to catch a returning Starship booster will slip by a few months, with a first re-flight now expected late this year or early 2027. Johnston acknowledged that an inability to book SpaceX capacity in 2029 would pose a serious problem.
Nvidia leads the extension
Manhattan West Ventures led the extension with participation from Cisco and a roster of returning backers including Benchmark, EQT, Soma, NFX, 776, Cedar Capital, Goanna Capital and Standard Capital. A person familiar with the deal said Nvidia contributed $25 million. Johnston cited the investment as validation of Starcloud’s technical lead: the company is the only known operator of an Nvidia H100 GPU in orbit and the first to train a model on one. Most space-qualified GPUs are built for edge processing, not training workloads.
Feeding the next chip generation
Starcloud is sharing orbital telemetry from its H100 deployment with Nvidia as the chipmaker develops the Vera Rubin Space-1, its first purpose-built space GPU. The silicon has not yet been fabricated, but Starcloud aims to fly it in late 2028. Johnston said Nvidia conducted deeper technical diligence than any other investor in the round, a claim that underscores how much of the startup’s valuation rests on a launch vehicle and a processor that have yet to reach operational maturity.