NVIDIA Sends First AI Server to Space
Space data centers cut energy costs 90%. NVIDIA H100 orbits Earth on Starcloud-1, testing AI in orbit.
“ChipPub” Publication: 20% Discount Offer Link.
The energy costs of space data centers will be only one-tenth of those on the ground.
On November 2, NVIDIA launched its H100 GPU into space for the first time.
As the current mainstream training chip in the AI field, the H100 is equipped with 80GB of memory, and its performance is hundreds of times greater than any computer that has previously entered space. In orbit, it will test a series of artificial intelligence processing applications, including analyzing Earth observation images and running Google’s large language models (LLMs).
This test flight is carried aboard the Starcloud-1 satellite from Starcloud, a startup based in Redmond, Virginia. It is the first step in the company’s ambitious plan to migrate the world’s energy-intensive data processing infrastructure to space. Starcloud is a member of the NVIDIA Inception startup program.
Supporters believe this idea is promising: In the distant depths of space, data centers won’t occupy valuable land, won’t require as much energy and water for cooling, and won’t emit greenhouse gases into the atmosphere.
