Starcloud, a Redmond, WA-based developer of orbital data centers for AI training, has raised $170M in Series A funding led by Benchmark and EQT Ventures. The startup deploys satellites equipped with NVIDIA H100 GPUs, powered by continuous solar energy and cooled via radiation to deep space. The capital will accelerate Starcloud-2, a commercial GPU cluster launching in 2026, and scale toward gigawatt constellations.
Orbital Data Center Funding Surges
The round values Starcloud at $1.1B post-money, bringing total funding to $200M and marking it as YC's fastest unicorn at 17 months post-demo day per TechCrunch. It follows recent activity like Axiom Space's $1.2B+ raised for space station-integrated compute and Aetherflux's $80M+ for solar-powered orbital AI. Sophia Space secured $13.9M for AI-ready platforms. Starcloud differentiates with its first orbital H100 launch and nanoGPT training in November 2025.
Earthbound AI Power Grids Strain
AI hyperscalers face grid constraints and permitting delays as data centers eye 100GW+ demand by 2026. Terrestrial energy costs and shortages limit LLM training clusters. Space offers unlimited solar (>95% capacity factor) and free radiative cooling, slashing costs to $0.002-0.005/kWh.
Satellites Enable Hyperscale AI Training
Starcloud-1, launched November 2025 (NORAD 66303), became the first satellite to run Gemini inference and train nanoGPT entirely in orbit. Proprietary thermal and power systems support modular scaling to 5GW via ~100 launches. Dawn-dusk orbits ensure near-constant sunlight without Earth infrastructure dependencies.
As Philip Johnston, CEO, noted:
"An H100 is probably not the best chip for space… we wanted to prove that we could run state of the art terrestrial chips in space."
Benchmark Backs Space Compute Bet
Benchmark and EQT Ventures lead, joined by Macquarie, NFX, YC Continuity, and angels including ex-Boeing CEO. This mix signals conviction in orbital compute amid AI energy crunches, following Starcloud's YC S24 seed and NVIDIA backing. Investors gain exposure to hyperscaler offtake deals in negotiation.
In-Orbit Market Scales to $39B
The in-orbit data centers market stands at $471M in 2025, projected to reach $39B by 2035 at 67.4% CAGR per BIS Research. Drivers include falling launch costs and radiation-hardened GPUs from NVIDIA. Lonestar Data Holdings raised $15.5M for lunar storage, while SpaceX eyes a million-satellite AI constellation.
Ex-SpaceX Team Drives First-Mover Edge
Co-founders include Philip Johnston (ex-McKinsey satellite strategist, Opontia exit after $42M raise), Ezra Feilden (PhD materials, Airbus/Oxford Space deployables), and Adi Oltean (ex-SpaceX Starlink principal engineer, 26-year Microsoft GPU scaler with 26 patents). Their blend tackles space hardware, AI compute, and scaling challenges. Team of 17 ex-SpaceX, Microsoft, NASA, Airbus launched Starcloud-1 post-YC.
Starcloud-2 Targets Blackwell GPUs
Starcloud plans Starcloud-2 for late 2026 with NVIDIA Blackwell B200 and AWS blades, plus Starcloud-3 on Starship. Partnerships with NVIDIA and Crusoe pave binding offtake agreements. The 88,000-satellite constellation aims at GW-scale AI for hyperscalers like OpenAI.