TLDR
- Washington-based Starcloud successfully operated Google’s Gemma AI model in space in November 2025 using an Nvidia H100 GPU
- The satellite chip is 100 times more powerful than any previous GPU sent to orbit and also trained NanoGPT on the works of Shakespeare
- The company wants to build a 5-gigawatt orbital data center using constant solar energy to meet Earth’s growing energy needs
- The next satellite launch, scheduled for October 2026, will feature multiple H100 chips and Nvidia’s Blackwell platform
- Competitors including Google, Lonestar Data Holdings and Aetherflux have announced similar aerospace data center initiatives
A Washington-based startup has achieved a first in artificial intelligence by training and deploying an AI model in space. Starcloud launched its satellite in early November 2025 with an Nvidia H100 graphics processing unit on board.
The Starcloud-1 satellite is now orbiting Google’s large language model Gemma. This is the first time such a model has worked on a powerful GPU outside Earth’s atmosphere.
The H100 chip is 100 times more powerful than any GPU previously sent to space. The satellite sent its first message to Earth saying “Greetings, Earthlings” and presented itself as ready to observe from orbit.
Starcloud CEO Philip Johnston said this achievement proves that space can support data center operations. The company trained NanoGPT, created by OpenAI founder Andrej Karpathy, using Shakespeare’s entire work on the orbital chip.
Tackling the energy crisis in data centers
Data centers on Earth consume enormous amounts of energy and water while also causing greenhouse gas emissions. The International Energy Agency predicts that electricity consumption from data centers will more than double by 2030.
Johnston said the energy costs of Starcloud’s orbital facilities will be ten times lower than those of data centers on Earth. The company plans to build a 5 gigawatt orbital data center with solar panels about 4 kilometers wide and high.
These space-based facilities would be able to continuously capture solar energy without Earth’s day-night cycles or weather interruptions. “Anything you can do in a data center on Earth, I expect you can do in space,” Johnston said.
The satellites will have an operational life of five years, based on the expected life of the Nvidia chips. Starcloud is a member of the Nvidia Inception program and a graduate of Y Combinator and Google for Startups Cloud AI Accelerator.
Real-World Applications and Future Plans
The satellite can answer questions about its location and operational status in real time. Users can ask where it is located and get answers like “I’m over Africa and in 20 minutes I’ll be over the Middle East.”
Starcloud develops customer applications using satellite images from Capella Space. The system can identify rescue boats from capsized ships at sea and detect forest fires as they break out.
Johnston said these capabilities would enable real-time intelligence for first responders. The company integrated the satellite’s telemetry to provide information about altitude, orientation, location and speed.
Starcloud’s next launch is scheduled for October 2026. It will include multiple Nvidia H100 chips and integrate Nvidia’s Blackwell platform for improved AI performance.
Increasing competition in space data centers
Several companies have announced plans for orbital data centers. Google unveiled Project Suncatcher on November 4, which aims to deploy solar-powered satellites with tensor processing units.
Lonestar Data Holdings is developing a commercial lunar data center for the lunar surface. Aetherflux announced plans to launch an orbital data center satellite in the first quarter of 2027.
Morgan Stanley analysts identified potential challenges including radiation exposure, maintenance issues, space debris and data management regulations. Technology companies continue to pursue the technology for access to unlimited solar energy and expanded operations.
The October 2026 Starcloud satellite will include a cloud platform module from Crusoe, allowing customers to deploy and operate AI workloads directly from space.
#Starcloud #successfully #operates #model #space #Nvidia #GPU #Blockonomi