Starcloud raises $170 million Series A to build data centers in space

Starcloud’s latest investment takes the startup space to $1.1 billion, making it one of the fastest startups to reach unicorn status after graduating from Y Combinator.

The company’s Series A, which closed 17 months after its IPO, was led by Benchmark and EQT Ventures. It is another sign of interest in releasing data to circulate if political and political obstacles limit their development in the world, but the type of business. it depends on uncertain technology and capital expenditure.

Starcloud has now raised $200 million, and launched its first satellite with an Nvidia H100 GPU in November 2025. The company will launch a more powerful system, Starcloud 2, at the end of this year with several GPUs, including the Nvidia Blackwell chip and the AWS server site, as well as bitcoin mining computers.

The company will also begin building a data center spacecraft designed to launch from Starship, a heavy-lift rocket built by Elon Musk’s SpaceX. Starcloud 3, as the plane is called, will be a 200 kilowatt, three-ton spacecraft that is connected to the “pez dispenser” system SpaceX designed to launch its Starlink satellites from Starship.

CEO and founder Philip Johnston said that he hopes that it will be the first orbital data center that is expensive and the data center in the world, with costs on the order of $ .05 per kw / hour of energy – if the commercial launch costs $ 500 per kilogram.

The problem is that the Starship still doesn’t fly; Johnston says that he expects commercial opportunities to open in 2028 and 2029. This is the reality facing all major data center projects: powerful space computers will be cheap until a new generation of rockets can be launched at a high operational cadence, which will not happen until the 2030s.

“If it’s going to be delayed, we’re just going to continue to launch smaller versions of the Falcon 9,” Johnston said. “We won’t be competitive on energy costs until Starship flies regularly.”

“There are two types of businesses,” explains Johnston: One is selling production power to another spaceship; The company’s first satellite, for example, analyzes data collected by the Capella Space radar spacecraft. Then, in the future when the opening rates drop, the most powerful data centers can attract work from their counterparts in the rest of the world.

This depends on how new this business is. When Nvidia CEO Jensen Huang unveiled the company’s Vera Rubin Space-1 modules at the annual GPU Technology Conference last week, he didn’t realize that nothing had been created or shared with the development team.

In fact, the number of advanced GPUs in orbit is limited, with Nvidia predicting that it will sell about 4 million of the world’s hyperscalers in 2025.

Or consider SpaceX’s Starlink communications network, the largest network of satellites in orbit with 10,000 spacecraft, making up about 200 objects. megawatt capacity, while the data center has more than 25 gigawatts of energy being built in the U..S, according to Cushman and Wakefield.

Johnston argues that his company is well ahead of the competition, with the world’s first GPU being used in the field. It was used to train an AI model in orbit, first, according to Starcloud, and run the Gemini model. Beyond performance, Johnston says Starcloud now has the necessary knowledge of what it takes to run a powerful chip in space.

“The H100 is probably not the best space device, to be honest, but the reason we did it is because we want to prove that we can run Earth’s equipment in space,” he told TechCrunch. That successful note – another GPU, the Nvidia A6000, failed at launch – will affect future designs.

There is a laundry list of problems that need to be solved, including the proper power generation and cooling of the chips we run. Starcloud-2 will have the largest radiator that can be operated on a private satellite; It is expected that at least two additional versions of the spacecraft will orbit, Johnston said.

Then there is the problem of synchronization. Very large datacenter workloads, often for training, require hundreds or thousands of GPUs to work together. To do this in space would require incredibly large ships, or powerful and reliable laser links between artificial flying ships. Most of the companies working on this technology hope that the jobs will come after the long simulation work is done in orbit.

In addition to Starcloud, AetherfluxGoogle’s Project Suncatcher, and Aethero – which launched Nvidia’s first Jetson GPU in 2025 – are both developing data-centric businesses.

The elephant in the room is SpaceX itself, which has asked the US government to allow it to build and operate a million satellites for distribution in space.

Going head-to-head with SpaceX is a daunting task for any entrepreneur, but Johnston sees room for coexistence.

“They’re creating a different way of using it than we are,” he told TechCrunch. “They are very much planning to serve Grok and Tesla for heavy duty. It may be at some point that they provide some kind of support, but what I think they cannot do is what we are doing (as) an energy and infrastructure player.”