Starcloud, a startup building satellites that run AI inference in orbit, told TechCrunch it has added a $250 million extension to the $170 million Series A round it closed in March. The extension values the company at $2.3 billion.

Manhattan West Ventures led the round. Nvidia and Cisco both participated; a person familiar with the deal said Nvidia put in $25 million. Benchmark, EQT, Soma, NFX, 776, Cedar Capital, Goanna Capital and Standard Capital also joined.

The money is going to launch, not silicon

The revealing part of this raise is where the capital is headed. Starcloud CEO Philip Johnston says some of it will fund a larger manufacturing facility and advance Starcloud-3, the company's largest orbital data center spacecraft. But the deeper reason for stockpiling cash is different: locking in launch capacity now.

"We can see what's coming — we're going to need to book an enormous amount of launch," Johnston said. The company has already asked the US Federal Communications Commission for permission to operate 88,000 spacecraft. Securing launch capacity, he said, is now one of the biggest cost items on the table.

The squeeze comes from SpaceX changing vehicles. The Falcon 9 program is scheduled to end in 2028, and the far larger Starship meant to replace it remains unproven. Competing rockets are in no position to absorb the gap:

  • Blue Origin's New Glenn and ULA's Vulcan are not flying regularly.
  • Rocket Lab's Neutron is not yet on the pad.
  • SpaceX's Starship just slipped further: Elon Musk said the attempt to catch a returning rocket will be delayed by a few months, and the first re-flight will come at the end of the year or in early 2027.

Johnston's caveat is blunt: "Obviously if we can't book any SpaceX launch capacity in 2029, that will be challenging for us."

Rideshare first, dedicated rockets later

For now the company is following a more modest plan. Two of its new-generation 8-kilowatt compute satellites, called Starcloud-2, will reach orbit on rideshare flights in 2027. They will perform orbital inference work for customers that include US government agencies. Starcloud is also weighing the purchase of a dedicated Falcon 9 launch and contracts with other providers to support later missions.

Nvidia's interest is technical

Johnston presents Nvidia's investment as evidence of the company's edge in a young sector. Starcloud is, as far as anyone knows, the only company operating a terrestrial-class Nvidia H100 data center GPU in orbit, and the first to train a model with one. Most other GPUs sent to space are designed for edge processing instead.

The company is sharing what it learns with Nvidia, which is developing Vera Rubin Space-1, its first GPU purpose-built for space. That chip has not been manufactured yet; Starcloud hopes to fly it in late 2028.

"The reason they've chosen to do this investment now is because of all of this data that we got from Starcloud One," Johnston said. "They, more than any other VC, did way more technical duty on this than anybody else."

Why it matters

Orbital data centers are pitched as an escape from the energy, water and land problems facing terrestrial ones. Starcloud's raise suggests the sector's real bottleneck is logistics rather than compute. Until the price of putting a kilogram into orbit settles, it stays unclear whether inference in space can compete on cost with a data center on the ground.