20 GW of computing capacity: NVIDIA-backed US startup to build data centers in space

US-based startup Starcloud has raised $250 million in a Series A extension at a $2.3 billion valuation, giving the company fresh capital to develop a new type of computing infrastructure designed to run AI workloads in orbit. The company is building data centers that place high-performance GPUs on satellites, with the goal of using space-based infrastructure to support the rapidly growing computing demands of AI. The latest funding will support spacecraft manufacturing, engineering and future launches as it works toward a large orbital computing network. The latest funding round, led by Manhattan West, brings Starcloud’s total funding since its 2024 founding to $450 million, with new investors including NVIDIA and Cisco Investments, alongside existing backers such as Benchmark and EQT. Putting AI infrastructure into orbit Starcloud’s approach is based on moving data center hardware away from Earth and into orbit. Its first major demonstration came in November 2025, when Starcloud-1 carried an NVIDIA H100 GPU into space. The company says that mission demonstrated that data center-grade computing hardware could operate in orbit, where conventional assumptions about cooling, radiation and power management no longer apply. “Last November we put the first NVIDIA H100 in orbit. Today this fresh capital empowers us to build the infrastructure to launch many more of NVIDIA’s most advanced GPUs into space,” said Philip Johnston, co-founder and CEO of Starcloud. Since then, Starcloud says it has trained an AI model in space, run a version of Google’s Gemini in orbit and demonstrated high-performance AI inference and fine-tuning using hardware operating in space. The company is now working with NVIDIA on the Space-1 Vera Rubin Module, designed specifically for the harsh conditions encountered outside Earth’s atmosphere. Unlike terrestrial data centers, orbital systems cannot rely on conventional air cooling. Instead, heat must be transferred through radiators, while computing components also need protection against radiation. Starcloud’s satellites are intended to serve as early platforms for testing space-rated computing hardware, using data from the Starcloud-1 mission to improve future systems. Giant orbital computing network of satellites The new capital will support Starcloud’s development of its next-gen Starcloud-3 spacecraft, with production lines already being built at a new 100,000-square-foot manufacturing facility in Woodinville, Washington. The company ultimately envisions a constellation of 88,000 satellites capable of providing 20 gigawatts of computing capacity in orbit, although reaching that scale would require a massive expansion of satellite manufacturing and launch capabilities. Starcloud argues that space could eventually provide an alternative way to meet the enormous energy and infrastructure requirements of AI computing. The company describes its approach as an attempt to address what it calls the “AI energy bottleneck” by moving computing infrastructure into orbit. The immediate focus, however, remains much smaller than Starcloud’s ultimate vision, with the company aiming to expand manufacturing, continue engineering work with NVIDIA and secure future launch capacity. Starcloud’s progress highlights a significant shift in how AI infrastructure could be built, with companies exploring the use of specialized computing hardware in orbit and connecting it through large-scale satellite networks. The technology remains at an early stage, however, and its long-term viability will depend on whether companies can solve the economic, technical and logistical challenges of operating and maintaining thousands of satellites in space.Get the latest in engineering, tech, space & science - delivered daily to your inbox.Bojan Stojkovski is a freelance journalist based in Skopje, North Macedonia, covering foreign policy and technology for more than a decade. His work has appeared in Foreign Policy, ZDNet, and Nature.