Startup testing AI computer chips in outer space
In this DML Report…
Starcloud, a Redmond, Washington-based startup, plans to send an NVIDIA H100 GPU into orbit aboard its Starcloud-1 satellite, roughly the size of a small refrigerator, to test AI data processing in space. The H100, nearly 100 times more powerful than any chip previously sent to orbit, powers advanced AI models and will analyze satellite data for applications like identifying wildfires, tracking crops, and monitoring weather directly in space, reducing delays from transmitting data back to Earth. Google's Gemma language model will operate in orbit for the first time as part of this mission. Earth-based data centers, which support online services including streaming and AI tools, consume significant electricity and water for cooling while emitting heat, noise, and greenhouse gases that impact communities. Starcloud aims to relocate such operations to space for improved efficiency.
The orbital approach leverages abundant solar energy for power and the natural cold vacuum of space for cooling, potentially saving millions of gallons of water annually and reducing carbon emissions by ten times compared to ground-based facilities after accounting for launch costs. Starcloud CEO Philip Johnston stated that post-launch, the system could achieve these environmental savings while accelerating computing for applications, games, AI tools, and cloud services. Benefits include faster results for Earth monitoring, enabling better ground-level decisions on disasters, weather forecasts, and resource management. Falling launch prices make space data centers increasingly viable.
(read more below)
Future plans involve incorporating NVIDIA's next-generation Blackwell GPUs for enhanced performance and building a five-gigawatt orbital data center approximately 2.5 miles wide to handle large AI workloads. Johnston predicts widespread adoption by the 2030s, with many new data centers operating in orbit rather than on land, combining innovation with sustainability to support a cleaner digital infrastructure. The primary challenge remains the environmental impact of the launch itself.