Google Ventures into Space AI
(Reuters)
Google is set to explore the potential of artificial intelligence in space as part of a broader initiative involving prominent tech figures, including Elon Musk. The initiative plans to launch satellites designed to function as floating data centers, leveraging the vast resources and energy available in space.
Next week, Google will conduct its first in-orbit test of Project Suncatcher, which aims to determine if space can accommodate large-scale AI computing infrastructure. The project intends to utilize solar power and ample space to support AI applications.
In partnership with SpaceX and Starcloud, Google’s mission seeks to deploy data centers in low Earth orbit. These centers will capitalize on nearly continuous sunlight to power energy-heavy AI tasks while aiming to mitigate terrestrial electricity limitations.
The upcoming mission, marketed as a Transporter-18 rideshare launch with Planet Labs, will test how Google’s AI hardware endures launch pressures, radiation, and temperatures in space. The prototype satellite will contain Google Tensor Processing Units to see if these AI chips function reliably in the unique conditions of low Earth orbit.
Space radiation poses risks to electronics, causing issues like data corruption, necessitating the orbital testing of the hardware. Previously, Google assessed its TPUs under AI workloads at the University of California, Davis, but in-orbit performance remains critical to understanding their capabilities.
Additionally, the company will evaluate its innovative cooling design, which incorporates heat pipes and radiators, essential in a vacuum where conventional cooling methods are not feasible.
Experts indicate that achieving a commercially viable solution could take years due to high launch costs and production challenges. The initial Suncatcher mission is geared towards collecting in-orbit data to pinpoint failure risks rather than proving an operational orbital data center.
Future plans include launching two additional satellites by 2027 to test high-bandwidth laser communications essential for future computing clusters.
Additional reporting by agencies.
