Seventeen months. That's how long it took for a team pitching orbital data centers at Y Combinator's Demo Day to become the accelerator's fastest-ever unicorn. The company—now called Starcloud, though it started life as Lumen Orbit—just closed a $170 million Series A at a $1.1 billion valuation, a milestone announced March 30 that has as much to do with timing as execution.
The round was led by Benchmark and EQT Ventures, firms not typically known for throwing capital at moonshots disconnected from commercial reality. Joining them: Macquarie, NFX, Nebular, Y Combinator's Continuity Fund, and Harpoon Ventures. Notable angel investors include Dennis Muilenburg, the former Boeing CEO, and General Stephen "Seve" Wilson. Total capital raised now sits at roughly $200 million.
The timeline is worth dwelling on. Y Combinator has funded over 5,000 startups since its founding and minted more than 100 unicorns. None reached a billion-dollar valuation this quickly after completing the program. Demo Day happened September 25-26, 2024. The Series A closed in late March 2026. Seventeen months—the period press outlets have characterized as the fastest path to unicorn status in YC history—in an industry where hardware timelines typically stretch across years, not quarters.
What They've Actually Built
Here's what separates Starcloud from the typical pre-revenue deep tech venture: they've already put hardware in orbit. In November 2025, the company launched Starcloud-1, a 60-kilogram satellite carrying the first Nvidia H100 GPU to reach space. The team successfully trained an AI model and ran inference workloads in orbit, according to multiple reports—a validation of the core technical thesis that might seem audacious to anyone who's spent time thinking about the challenges of running high-performance compute in a vacuum.
Starcloud-2 is scheduled for launch in October 2026. This one's more ambitious: multiple GPUs including Nvidia's Blackwell architecture, an AWS server blade, and bitcoin mining ASICs. Perhaps more importantly, it will deploy what industry reports describe as a massive deployable radiator, an attempt to solve one of the fundamental problems of space-based compute. Heat dissipation in the vacuum of space is... not trivial.
CEO Philip Johnston told TechCrunch that Starcloud-3—still in development—will weigh roughly three tons and consume 200 kilowatts of power. It's being designed specifically for deployment via SpaceX's Starship launch system, which uses what the company calls a "PEZ dispenser" mechanism to release payloads. The scale ramp between generations is striking.
The Economics—and the Catch
Johnston outlined a cost target that sounds almost too good to be true: roughly $0.05 per kilowatt-hour, competitive with terrestrial data centers. There's a caveat the size of a Starship. That number only works if launch costs drop to around $500 per kilogram. Current market rates remain significantly higher, though SpaceX's Starship is designed to push costs down dramatically—assuming it achieves full reusability and high flight cadence. That's still a big assumption.
The value proposition, as Starcloud tells it, rests on three pillars. First: continuous solar exposure in sun-synchronous orbit provides 24/7 power without the grid constraints now plaguing AI infrastructure buildouts on Earth. Second: radiative cooling in space eliminates the massive energy overhead of traditional HVAC systems. Third: as launch costs decline, the capital expenditure per compute unit could theoretically undercut ground-based facilities.
The engineering challenges, of course, are formidable. Radiation hardening for commercial GPUs. Thermal management at scale. Orbital debris mitigation for large constellations. Ground latency. Johnston has acknowledged chip reliability as a central risk the company is working to solve through testing and redundancy. Which is another way of saying: we're figuring it out as we go.
An Industry Takes Shape
Starcloud isn't operating in isolation. According to industry reports, the company filed with the FCC in early February for authorization to operate up to 88,000 satellites as a distributed orbital data center constellation. For context, SpaceX submitted an application in late January for 1 million satellites designated for "Orbital Data Center" use. The FCC accepted it for filing on February 4, 2026.
Blue Origin has reportedly filed for up to 51,600 satellites in the orbital data center category, according to GeekWire reporting from March 2026. China's government announced in February that its five-year plan includes dedicated work on space-based data centers. At Nvidia's GTC conference in March, the chipmaker formalized the trend with a "Space Computing" initiative and announced the Vera Rubin Space Module designed explicitly for orbital deployments.
The partners and validators matter. Nvidia featured Starcloud in an October 2025 blog post on the technical rationale for running GPUs in orbit. The company appears on Nvidia's official list of Space Computing program participants. Tom's Hardware reported in October 2025 that Crusoe Cloud, a specialist in energy-optimized compute infrastructure, was exploring workloads on Starcloud's platform—though formal customer contracts have not been publicly disclosed.
The Y Combinator Pedigree
Starcloud entered Y Combinator's Summer 2024 batch, when the program was still finding its rhythm after expanding to four cohorts per year. TechCrunch's Demo Day recap from September 25, 2024 listed Lumen Orbit among 13 companies "worth paying attention to," noting the heavy presence of AI infrastructure startups in the cohort. That summer batch, in hindsight, looks prescient.
The founders brought relevant backgrounds, though not the obvious aerospace pedigree you might expect. CEO Philip Johnston previously co-founded Opontia, an e-commerce brand aggregator that raised seed and Series A rounds in 2021. CTO Ezra Feilden came from Airbus Defence & Space and Surrey Satellite Technology Limited, where he worked on deployable space structures. Chief Engineer Adi Oltean previously held roles at SpaceX and Microsoft Azure. It's a mix of startup velocity and deep technical chops.
By December 2024, the company had closed an $11 million seed round at roughly a $40 million valuation. TechCrunch reported that over 200 VCs expressed interest in the allocation, which tells you something about the fundraising environment at the time. That seed round came just months after a $2.4 million pre-seed in early 2024, when the company was still called Lumen Orbit and operating out of El Segundo, California. The team later relocated to Redmond, Washington—closer to the aerospace talent pool in the Pacific Northwest, presumably.
What the Record Signals
The $1.1 billion valuation and 17-month timeline to unicorn status reflect something larger than one startup's execution, however impressive. The combination of AI compute demand, terrestrial power grid constraints, and declining launch costs has created a window for ideas that would have seemed absurd—or at least impractical—a decade ago.
Data center operators are planning facilities requiring multiple gigawatts of power, according to research from Uptime Institute published in early 2026. S&P Global reported in April that data center capital expenditure and power procurement have reached what they diplomatically called "extreme levels" as hyperscalers race to secure capacity for training runs and inference at scale. The terrestrial infrastructure simply isn't keeping pace. Permitting timelines for new power plants stretch years into the future, and utilities in some regions are openly warning about capacity constraints.
Which helps explain why Benchmark and EQT Ventures—sophisticated investors with reputations to protect—are comfortable co-leading a $170 million round for a company putting GPUs in orbit. Their involvement, alongside the broader investor syndicate, suggests conviction that orbital data centers can move from science experiment to viable business within the fund lifecycle. Whether that conviction proves justified is another question.
The Path Ahead
Whether Starcloud maintains its velocity remains to be seen, and the challenges ahead are substantial. The company must execute three satellite generations while proving unit economics at scale. It needs to navigate regulatory frameworks for massive constellations—frameworks that are still being written in real time. And it's competing with both established aerospace primes and well-funded startups attacking the same opportunity.
But for now, at least, the company that pitched at Demo Day 17 months ago as Lumen Orbit has carved out a place in Y Combinator history. And perhaps—though this part is still very much TBD—in the future of computing infrastructure. The next few years will determine whether orbital data centers represent a genuine shift in how we think about AI infrastructure, or an expensive detour funded by a moment when capital was abundant and launch costs were (theoretically) plummeting.
Either way, seventeen months is a hell of a run.
