Google to Test Orbital AI Hardware With Experimental Satellite Launch

Google is preparing to send an experimental micro data center into low-Earth orbit as part of its Project Suncatcher initiative, an ongoing research effort testing the feasibility of space-based artificial intelligence infrastructure, according to a report from Engadget.

Scheduled for launch on Oct. 1 aboard a SpaceX Falcon 9 rocket, the mission will deploy a test satellite named MVP. Rather than a full-scale facility, the payload consists of four specialized Tensor Processing Unit (TPU) chips that combined offer the computing capability of a single standard data center server.

The orbital hardware will be powered by solar panels supplying one kilowatt of electricity, equivalent to the power draw of a household hair dryer. Google intends to execute simple AI queries on the setup over a one-year testing period. The satellite is expected to remain in low-Earth orbit for six years before burning up upon atmospheric re-entry.

To prepare for orbital conditions, researchers subjected the silicon to heavy radiation exposure at the Crocker Nuclear Laboratory. Radiation in space presents a significant hazard to computer hardware by causing "bit flips," in which binary data unexpectedly shifts between ones and zeros. Google plans to mitigate operational errors by rebooting the chips when anomalies occur.

Thermal management presents another primary obstacle. The satellite utilizes a proprietary cooling system using layers of conductive material to expel heat into space. However, the mechanism currently permits only about 15 minutes of continuous operation before the chips must be powered down to cool.

Company executives and external researchers emphasize that operational space-based data infrastructure remains a long-term endeavor. "We don't expect, to be perfectly frank, that we'll have anything usefully operational in the next few years," said James Manyika, Google's senior vice president for research, technology, and society.

Brandon Lucia, an electrical and computer engineering professor at Carnegie Mellon University, told The New York Times that expanding from a single test unit to an operational network will demand substantial funding and novel engineering solutions. Major challenges include launch expenses—with single Falcon 9 flights priced around $74 million—alongside cosmic rays that can physically punch through and melt semiconductor junctions, as well as complex scaling requirements for power generation and heat dissipation.

Despite the technical obstacles, Google plans to scale testing by launching two similar satellites next year. Long-term roadmaps include evaluating an 80-satellite constellation flying in close formation to process AI queries, as well as studying a single satellite installation roughly the size of a football field. The company has not disclosed its total financial expenditure on the project.