Starcloud raises $250M to scale on an unbuilt chip

Starcloud's $250 million Series A extension, led by Manhattan West, brings total capital raised since 2024 to $450 million after a $170 million Series A in March and values the orbital data center developer at $2.3 billion post-money. Existing backers Benchmark, EQT, Soma, NFX and 776 were joined by new investors Nvidia, Cisco Investments, Cedar Capital, Goanna Capital and Standard Capital. Cisco's Aleem Rizvon framed the bet as bringing secure data-center expertise to the emerging orbital market.

Starcloud itself is small: 25 employees in Redmond, Washington, building production lines for the Starcloud-3 spacecraft at a 100,000-square-foot factory in Woodinville. That factory is the bridge between a demonstrator and the stated ambition of an 88,000-satellite constellation delivering 20 gigawatts of capacity, for which Starcloud has already sought FCC permission. The gap between the factory floor and the constellation is the size of the bet.

At bottom, the extension is a continuation of Starcloud's relationship with Nvidia. The company says it put an Nvidia H100 into orbit in November 2025 and agreed in March 2026 to deploy Nvidia's Space-1 Vera Rubin module in late 2028; the catch is that the chip is still unbuilt, and Nvidia plans to put the same space-hardened Vera Rubin on orbit with Cowboy Space, Axiom Space, Kepler Communications, Planet and Sophia Space. Starcloud's release nonetheless mentions the world's largest company by market cap eight times.

What Starcloud has actually demonstrated in orbit—training NanoGPT, Andrej Karpathy's minimalist educational model, on the complete works of Shakespeare before running Google's lightweight Gemma model—is closer to an engineering proof than a workload. These are real firsts for the orbital category, yet the announcement includes no revenue figures and no customer contracts.

Cowboy Space, formerly Aetherflux, closed a $275 million Series B in May at a $2 billion valuation, pledging to vertically integrate orbital data centers with rockets, and Starcloud's new mark puts two orbital data center developers past the billion-dollar line this year.

Starcloud's $2.3 billion post-money sits above Cowboy Space's $2 billion from May, and neither announcement carries revenue figures; the premium suggests investors are pricing the chip relationship over the constellation economics.

For digital-infrastructure investors, the logic is familiar: the bottleneck has shifted from land and chips to electrons. Starcloud's answer is extreme—put the compute above the grid and avoid interconnection queues entirely. But leaving the atmosphere does not leave the queue. An FCC filing, a factory ramp and a launch manifest become the scarce things, and a $250 million round does not buy a place in any of those lines; it gets Starcloud to the factory door.

Terrestrial data centers clear capital when they carry a hyperscaler guarantee; everything else fights for attention. Starcloud has Nvidia on the cap table without a hyperscaler landlord, and Nvidia is spreading the same chip across half a dozen orbital data center developers, which makes Starcloud's position less singular than its release implies.

The market is pricing this correctly for a company at Starcloud's stage, but the price is fragile because the Nvidia relationship underpins the valuation. If Vera Rubin ships and flies on schedule, the next round will be about capacity; if the schedule slips, that support disappears. The same calculus has played out on the ground, where Nvidia's guarantee, more than its equity, is what would make a $500 billion project financeable. In orbit, the guarantee is starker: Starcloud's valuation rests on the chipmaker's launch calendar, and the first Vera Rubin launch is the only date that matters.