Google will launch a prototype satellite next week to test whether its artificial intelligence chips can operate reliably in low Earth orbit, marking the first in-orbit trial under its Project Suncatcher research programme.
The mission will examine how Google’s Tensor Processing Units, or TPUs, withstand the physical and thermal conditions of space, including launch forces, radiation and extreme temperatures. The satellite is scheduled to fly on SpaceX’s upcoming Transporter-18 rideshare launch in partnership with satellite company Planet Labs.
Project Suncatcher is exploring whether large-scale AI computing infrastructure could eventually be placed in space. Companies including SpaceX and Starcloud are pursuing similar concepts, seeking to use near-continuous sunlight as an energy source for power-intensive AI workloads while reducing dependence on terrestrial electricity supplies.
Google said the first mission is intended to collect in-orbit data and identify potential failure points, rather than demonstrate a functioning orbital data centre. The prototype satellite will carry Google TPUs and assess whether they can run AI workloads in the space environment.
One of the key challenges is radiation. Space radiation can disrupt or degrade electronic systems and cause data errors known as bit flips. Google said it had already tested its TPUs while running AI workloads at a facility at the University of California, Davis, but that testing in orbit was necessary to understand how the hardware performs under actual space conditions.
Thermal management is another major engineering challenge. Conventional airflow cooling is impossible in the vacuum of space, so Google will evaluate a system that combines heat pipes with radiators to remove heat from the power-intensive chips. The results will help determine how effectively the hardware can operate without the cooling systems used in conventional data centres.
The proposed model would link computing hardware in orbit to other systems while using solar energy to support operations. However, experts cited in the report said the concept remains years away from commercial viability because of high launch costs, engineering constraints and bottlenecks in satellite production.
The planned mission is therefore a technology demonstration rather than an immediate alternative to land-based data centres. Its findings will be relevant to the wider infrastructure demands created by artificial intelligence, particularly the need for reliable electricity, cooling and high-capacity computing facilities.
Google has said it aims to launch two satellites in 2027 to test the high-bandwidth laser links that would be required to connect future orbital computing clusters. Those tests will address whether data can be transferred between satellites and other network components at the scale needed for distributed AI computing.
The first satellite’s performance in low Earth orbit will provide the initial operational data for that longer development path. The upcoming Transporter-18 mission is expected to establish how Google’s AI hardware responds to the radiation, temperature and mechanical conditions that any future space-based computing system would have to withstand.
