Via futurumgroup.com
The Y Combinator-backed startup plans to put GPU clusters in space, promising energy costs up to 90% cheaper than terrestrial facilities.
A startup wants to solve AI’s insatiable electricity appetite by doing something that sounds like a plot device from a sci-fi movie: putting data centers in orbit. Starcloud, formerly known as Lumen Orbit, has raised roughly $200M in total capital and reached a $1.1B post-money valuation, making it the fastest company in Y Combinator’s history to hit unicorn status.
The company pulled off its $170M Series A on March 30, 2026, just 17 months after its Y Combinator demo day. Benchmark, EQT Ventures, Sequoia, and a16z scout funds all participated, alongside In-Q-Tel, the venture arm of the US intelligence community.
Solar power, radiative cooling, and GPUs in the sky #
Starcloud’s pitch is deceptively straightforward. AI training and inference require enormous amounts of electricity. Terrestrial data centers face rising power costs, grid constraints, and increasingly hostile permitting environments. Space, meanwhile, offers something Earth cannot: uninterrupted sunlight and a natural vacuum that dissipates heat through radiative cooling.
The company claims this combination could reduce energy costs by up to 90% compared to ground-based facilities. CEO Philip Johnston has said the savings could reach a tenfold reduction.
Starcloud’s near-term hardware roadmap centers on two satellites. Starcloud-1 is designed to host the first NVIDIA H100-class GPU in space. Starcloud-2, targeted for a 2027 launch, is a commercial GPU cluster satellite operating in a sun-synchronous orbit, which keeps the craft in consistent sunlight as it circles the planet.
In February 2026, Starcloud submitted an FCC proposal for a constellation of up to 88,000 satellites. The target: approximately 20 gigawatts of orbital computing capacity. For perspective, a single large terrestrial data center might consume 100 to 300 megawatts.
The road from PowerPoint to orbit #
It is worth noting that Starcloud has not yet put a GPU in space. The company’s milestones so far are fundraising, FCC filings, and hardware development.
Latency is one obvious concern. Data traveling to and from low-Earth orbit faces round-trip delays that may be acceptable for AI training workloads, which can be batched, but potentially problematic for real-time inference.
There’s also the question of orbital debris. An 88,000-satellite constellation would be larger than SpaceX’s Starlink network, which has already drawn scrutiny from astronomers and regulators over space congestion. The first successful commercial deployment, likely Starcloud-2 in 2027, will be the real test of whether orbital data centers graduate from audacious concept to competitive infrastructure.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our