AI model successfully trained on orbit using Nvidia GPU, validating space-based compute capability.

Starcloud successfully trained and ran AI models on an Nvidia H100 enterprise GPU launched into Earth's orbit on November 2 on a test satellite. The company used the GPU to train NanoGPT, a lightweight, open-source model created by OpenAI founding member Andrej Karpathy, and has also been "running inference" on the model to generate answers and output. Additionally, Starcloud has been operating Google's open-source Gemma model on the GPU, effectively creating a chatbot in space.

The NanoGPT implementation was trained on the complete works of Shakespeare, according to Starcloud Chief Engineer Adi Oltean. The test satellite, Starcloud-1, sent back a message reading: "Greetings, Earthlings! Or, as I prefer to think of you—a fascinating collection of blue and green."

This achievement represents an early step toward placing data centers in Earth's orbit, which could trigger a new space race. Major players including SpaceX, Google, and Amazon founder Jeff Bezos have highlighted potential benefits such as near-limitless solar energy. On Earth, AI data centers are raising environmental concerns and straining the electric grid, making orbital alternatives increasingly attractive.

However, Starcloud-1 remains just one satellite, about the size of a small refrigerator, carrying a single H100 GPU—vastly smaller than the Earth-based AI data centers that house tens of thousands and even millions of GPUs. Significant technical and financial challenges stand in the way, most notably the difficulty of cooling GPUs in the vacuum of space, which offers no air to dissipate heat. Starcloud is exploring air-based or liquid-based cooling solutions and envisions deploying "the largest radiators deployed in space" to manage heat dissipation.

The company is preparing a second satellite, Starcloud-2, featuring even more GPUs, with plans to launch it sometime next year and eventually offer access to customers. As Oltean stated: "This is a significant first step toward moving almost all computing off Earth to reduce the burden on our energy supplies and take advantage of abundant solar energy in space."