The extension, led by Manhattan West Ventures, adds to the $170 million Series A the company closed in March. Nvidia contributed $25 million, with Cisco, Benchmark, EQT, Soma, NFX, 776, Cedar Capital, Goanna Capital, and Standard Capital also participating. The capital funds a larger manufacturing facility and advances Starcloud-3, the company’s largest spacecraft, designed to fly on SpaceX’s forthcoming Starship.
The Launch Capacity Squeeze
CEO Philip Johnston frames the raise as a hedge against a tightening launch market. Falcon 9, the industry’s workhorse, is scheduled to end in 2028. Its replacement, Starship, remains unproven for operational flights. Competing vehicles — Blue Origin’s New Glenn, ULA’s Vulcan — are not yet flying regularly, and Rocket Lab’s Neutron has not reached the pad.
“We can see what’s coming — we’re going to need to book an enormous amount of launch,” Johnston said. Starcloud has already filed with the FCC to operate 88,000 spacecraft, a constellation scale that demands guaranteed access to orbit.
Launch costs have long been the central obstacle for orbital data center startups. One competitor opted to build its own rockets rather than depend on third-party providers. Starcloud’s strategy is to lock in Starship contracts early, betting that SpaceX can demonstrate rapid reuse of the world’s most powerful rocket.
Nvidia’s Bet on Space Compute
Nvidia’s $25 million investment signals more than financial backing. Starcloud is the only known company operating an Nvidia H100 data center GPU in orbit and the first to train a model using it. Most space GPUs are built for edge processing, not the heavy inference workloads Starcloud targets.
Johnston said Nvidia conducted deeper technical diligence than any other investor in the round, motivated by data from Starcloud’s first orbital mission. The companies are collaborating on Nvidia’s first purpose-built space GPU, the Vera Rubin Space-1 chip, which has not yet been manufactured. Starcloud hopes to fly it in late 2028.
Engineers are tracking three design challenges: the relationship between chip running temperature and radiator size, radiation shielding placement, and ruggedizing chips to survive launch forces. These engineering decisions will shape whether orbital AI inference can scale to compete with terrestrial data centers.
Near-Term Roadmap
Starcloud plans to launch two Starcloud-2 satellites — 8 kW compute spacecraft — on rideshare flights in 2027 for U.S. government agency customers. The company is also weighing a dedicated Falcon 9 purchase and contracts with other providers to support future missions.
The 25-person company is building production lines at a 100,000-square-foot facility in Woodinville, Washington, near SpaceX and Amazon satellite manufacturing operations.
What Happens Next
The orbital data centers market hinges on two variables Starcloud cannot fully control: Starship’s operational cadence and the cost curve it enables. Elon Musk said this week that SpaceX will delay a Starship catch attempt by several months and aims to re-fly the vehicle by late 2026 or early 2027. Any further slip compresses Starcloud’s timeline for scaling beyond demonstration missions.
Watch for Starcloud’s Starship contract signing — a signal that SpaceX is ready to commit capacity — and for the Vera Rubin Space-1 chip’s fabrication milestones. If Nvidia’s space GPU reaches orbit on schedule in 2028, the competitive gap between orbital and terrestrial AI inference narrows. If it slips, Starcloud’s $2.3 billion valuation rests on a launch vehicle and a chip that neither yet exist in operational form.
— David Kim, technology desk, AXO News