Starcloud, the Redmond, Washington startup building data centers in low Earth orbit, has added a $250 million extension to its March Series A round, valuing the company at $2.3 billion and pushing total capital raised since its 2024 founding to roughly $450 million. The August 21, 2026 tranche was led by Manhattan West Ventures, with participation from NVIDIA, Cisco Investments, Benchmark, EQT Ventures, NFX, 776, Soma, Cedar Capital, Goanna Capital and Standard Capital.
Nvidia Backs A Space-Ready GPU Roadmap
NVIDIA put roughly $25 million into the round, according to a person familiar with the deal, marking the chipmaker's first equity commitment to an orbital compute company. Starcloud CEO Philip Johnston told TechCrunch the diligence was “way more technical” than a typical growth check, driven by data collected from Starcloud-1, which flew the first NVIDIA H100 GPU to orbit in November 2025 and trained a large language model in space. NVIDIA is now designing its first purpose-built space GPU, dubbed Vera Rubin Space-1, and Starcloud plans to fly the chip in late 2028.
Racing To Lock In Launch Capacity
Johnston said much of the new capital is aimed at securing rocket transport as launch options tighten. SpaceX plans to phase out its Falcon 9 workhorse in 2028 in favor of the still-unproven Starship, while Blue Origin's New Glenn, ULA's Vulcan and Rocket Lab's Neutron are not flying regularly. Starcloud has already requested FCC permission for 88,000 spacecraft and is preparing to buy a dedicated Falcon 9 slot for its Starcloud-2 8 kW compute satellites, which will fly on rideshare missions in 2027 for U.S. government and commercial inference customers. The larger Starcloud-3 spacecraft is architected around Starship's payload envelope.
Scaling A Woodinville Satellite Line
The company, now 25 employees and growing, is building production lines at a 100,000-square-foot facility in Woodinville, Washington, near the plants where SpaceX and Amazon assemble Kuiper and Starlink satellites. Starcloud is pitching orbital compute as a way to sidestep terrestrial permitting bottlenecks and use continuous solar power and radiative cooling to scale toward gigawatt-class AI training clusters — a thesis it shares with rivals like Cowboy Space and Muon Space.
Reporting based on coverage from TechCrunch, SpaceNews and Business Wire.