Orbital Inference Data Center Bets On Space GPUs

https://spectrum.ieee.org/orbital-inference-data-center

Publish Date: 2026-05-10 09:00:01

The surge in demand for electricity to power the rapidly advancing large language models is driving a global data center boom, straining terrestrial energy supplies, and prompting infrastructure operators to seek alternative power sources. Orbital Inc., a Los Angeles-based startup, has emerged with plans to build space data centers powered by solar energy, aiming to overcome terrestrial energy constraints and deliver AI inference workloads efficiently.

Orbital’s vision includes a mesh constellation of small, solar-powered satellites in low Earth orbit, where each satellite would feature GPU server racks and radiative cooling panels. These satellites, with up to 10,000 potentially in orbit, would be designed to run inference workloads such as chatbots and agents. In 2027, Orbital plans to launch a prototype satellite to validate its system in space, leveraging smaller satellite units to benefit from lower launch costs. Despite technical challenges such as radiation and thermal management, Orbital is optimistic about its long-term goals, despite experts predicting that space data centers may not be operational for at least another decade.

Key Points:
1. The boom in large language models is increasing energy demand for data centers.
2. Orbital Inc. plans to build space data centers powered by solar energy to address terrestrial power shortages.
3. Orbital’s vision involves a distributed network of smaller, solar-powered satellites.
4. Technical challenges like radiation and thermal management need to be addressed for the space data center concept to succeed.
5. Orbital’s launch in 2027 will be critical to identifying and addressing these engineering hurdles.