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

Starcloud’s latest funding round values ​​the space computing company at $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 Demo Day presentation, was led by Benchmark and EQT Ventures. It’s another sign of interest in outsourcing data centers on orbit where resource and political hurdles are slowing their development on Earth, but the business model relies on unproven technology and significant capital expenditures.

Starcloud has now raised a total of $200 million, and launched its first satellite with an Nvidia H100 GPU in November 2025. The company will launch a more powerful version, Starcloud 2, later this year with multiple GPUs, including an Nvidia Blackwell chip and AWS server code, as well as a computer for Bitcoin mining.

The company will also begin developing a data center spacecraft designed to launch from Starship, a reusable heavy rocket being built by Elon Musk’s SpaceX. Starcloud 3, as the spacecraft has been called, will be a 200-kilowatt, three-ton spacecraft that will fit into a “pez dispenser” system designed by SpaceX to deploy Starship’s Starlink satellites.

CEO and founder Philip Johnston said he expects this will be the first orbital data center to rival terrestrial data centers in terms of cost, with costs of about $0.05 per kilowatt-hour of power — if the cost of a commercial launch reaches about $500 per kilogram.

The challenge is that the spacecraft has not flown yet; Johnston says he expects commercial access to open in 2028 and 2029. This is the reality facing all large space data center projects: The cost of powerful space computers will be prohibitive until a new generation of rockets begins launching at a high operational cadence, which may not happen until the 2030s.

“If it is delayed, we will continue to launch smaller versions of the Falcon 9,” Johnston said. “We won’t be competitive on energy costs until the spacecraft flies frequently.”

“There are two business models: one is to sell processing power to other spacecraft in orbit; one is to sell processing power to other spacecraft in orbit; one is to sell processing power to other spacecraft in orbit,” Johnston explains. For example, the company’s first satellite analyzes data collected by Capella Space’s radar spacecraft. Then, in the future, when launch costs come down, more powerful distributed data centers can pull work from their terrestrial counterparts.

This shows how new this industry really is. When Nvidia CEO Jensen Huang unveiled the company’s Vera Rubin Space-1 chipset modules at his company’s annual GPU technology conference last week, he failed to note that none of them had been produced or shared with the company’s development partners.

In fact, the number of advanced GPUs in orbit is in the dozens, while Nvidia is estimated to have sold nearly 4 million to terrestrial CPUs in 2025.

Or consider that SpaceX’s Starlink communications network, the largest satellite network in orbit with 10,000 spacecraft, produces nearly 200 MW of energy, while the data centers include more than 25 gigawatt Energy is currently under construction in the United States, according to Cushman and Wakefield.

Johnston says his company is far ahead of the competition, with the first terrestrial GPU deployed in orbit. It was used to train an in-orbit AI model, the first of its kind, according to Starcloud, and to operate a copy of Gemini. Beyond performance, Johnston says Starcloud now has valuable data about what it takes to run a powerful chip in space.

“The H100 is probably not the best chip for space, to be honest, but the reason we did it is because we wanted to prove that we can run the latest terrestrial chips in space,” he told TechCrunch. This hard-won knowledge — another GPU, the Nvidia A6000, failed during launch — will influence future designs.

There is a long list of technical challenges to solve, including efficient power generation and cooling of hot chips. Starcloud-2 will have the largest deployable radiator ever launched on a private satellite; Johnston said he expects at least two additional versions of that spacecraft will head into orbit.

Then there’s the challenge of synchronization. Larger data center workloads, often for training purposes, require hundreds or thousands of GPUs to work in tandem. Doing this in space would require either fantastically large spacecraft, or powerful and reliable laser links between spacecraft flying in formation. Most companies working on this technology expect these workloads to come long after simpler inference tasks are performed in orbit.

Along with Starcloud, Aetherflux, Google’s Project Suncatcher, and Aethero – which launched Nvidia’s first space-based Jetson GPU in 2025 – are developing their space-based data center business.

The biggest problem is SpaceX itself, which has asked the US government for permission to build and operate 1 million distributed computing satellites in space.

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

“They’re building a slightly different use case than ours,” he told TechCrunch. “They plan to primarily serve Grok and Tesla workloads. Maybe at some point they will offer a third-party cloud service, but what I think they are unlikely to do is what we do.” [as] Energy and infrastructure player.