NVIDIA-backed startup runs AI model from orbit, observing earthlings
Starcloud, a Washington-based startup backed by NVIDIA, has just launched a satellite carrying the first AI language model that actually works from space.
Their Starcloud-1 satellite carries an NVIDIA H100 GPU and runs Google's open-source Gemma LLM, executing inference and responding to queries while orbiting above us.
Why does this matter?
Starcloud dreams big: they want to build massive solar-powered data centers in orbit to help reduce Earth's energy and water use.
Their current satellite even trained on Shakespeare so it can answer questions in old-school English.
They're already using their tech with Capella Space to spot things like shipwreck lifeboats and forest fires from space.
With a new, even more powerful satellite planned for 2026—and big names like Google and SpaceX also eyeing space computing—the race to take AI off-planet is officially on.