Starcloud Reaches Unicorn Status With $170 Million Series A to Build Data Centers in Orbit as Terrestrial AI Infrastructure Hits Its Limits
The Y Combinator graduate plans an 88,000-satellite constellation for orbital AI compute, backed by Benchmark and EQT Ventures.
Overview
Starcloud, a Redmond, Washington-based startup building data centers in orbit, has raised $170 million in a Series A round that values the company at $1.1 billion, according to TechCrunch. The round, led by Benchmark and EQT Ventures, makes Starcloud a unicorn just 17 months after its Y Combinator demo day presentation, one of the fastest paths to that milestone among YC alumni.
The funding arrives as terrestrial data center expansion faces mounting constraints from land scarcity, power grid capacity, and cooling demands driven by AI workloads. Starcloud’s bet is that low Earth orbit offers a way around all three.
What We Know
The Series A closed in two tranches, with Benchmark leading the first and EQT co-leading the second, according to SiliconANGLE. Other participants include Monolith Power Systems, a publicly traded power management chip maker. The round also drew participation from Macquarie Capital, NFX, Nebular, Y Combinator, Adjacent, 776 Ventures, and Fuse Ventures, according to TechCrunch.
Starcloud launched its first satellite, Starcloud-1, aboard a SpaceX Falcon 9 in November 2025, carrying an Nvidia H100 GPU that the company describes as roughly 100 times more powerful than anything previously operated in orbit, according to SiliconANGLE. The company used it to train custom AI models and run Google Gemini algorithms in space.
A second satellite, Starcloud-2, is planned for later in 2026. It will carry multiple GPUs including an Nvidia Blackwell chip, an AWS Outposts server blade, and what the company describes as the largest commercial deployable radiator ever launched to space, according to SiliconANGLE. Its onboard solar array will generate 100 times more power than its predecessor. Crusoe, an AI data center builder, has signed on to run workloads on Starcloud-2.
The company has filed an application with the Federal Communications Commission for a constellation of up to 88,000 satellites in low Earth orbit, according to SpaceNews. The long-term vision described by co-founder and CEO Philip Johnston involves a five-gigawatt orbital data center powered by a 6.1-square-mile solar array, with liquid-cooled server containers linked by thousands of fiber-optic cables, as reported by SiliconANGLE.
Starcloud-2 will also include an AWS Outposts server blade, indicating a partnership with Amazon Web Services. The company’s Starcloud-1 satellite already runs Nvidia hardware and has executed Google Gemini workloads in orbit, pointing to relationships with both companies.
What We Don’t Know
The economics of orbital compute remain unproven at scale. Starcloud has not disclosed detailed cost-per-compute-hour projections or the expected lifetime of its satellites. Whether space-based systems can reach cost parity with terrestrial data centers within the next few years, as the company has suggested, depends on manufacturing scale, launch costs, and satellite longevity that have yet to be demonstrated.
The 88,000-satellite FCC filing dwarfs most proposed constellations, and it remains unclear how quickly Starcloud can scale manufacturing and launch cadence to approach that figure. The company plans to use part of its new funding to build manufacturing capacity and expand headcount, according to SiliconANGLE, but has not shared a production timeline.
Latency, bandwidth to ground stations, and orbital debris management are open technical challenges for any space-based compute provider. Whether enterprise and hyperscaler customers will trust latency-sensitive AI inference to orbital infrastructure, as opposed to batch training workloads, has yet to be demonstrated.
Analysis
Starcloud is entering a nascent but increasingly competitive field. SpaceX has signaled orbital data center ambitions of its own, and Blue Origin is expanding its space infrastructure portfolio. But Starcloud appears to be the first startup to demonstrate an Nvidia data center GPU operating in orbit and to complete AI model training there, giving it a technical lead and a compelling pitch to investors wary of the terrestrial power and cooling bottleneck.
If Starcloud-2 delivers on its promised 100x power increase and handles commercial workloads from Crusoe later this year, the company’s trajectory from YC demo day to operational orbital data center will have taken less than three years. The question is whether the physics and economics of space-based compute can scale as fast as the ambition.