Change language to
0:00

Google will launch a prototype satellite next week to test whether its Tensor Processing Units (TPUs) can run reliably in orbit under Project Suncatcher. The mission will fly on SpaceX’s Transporter-18 rideshare with Planet Labs, but it is a research test rather than an orbital data centre ready for customers.

Google’s Project Suncatcher is exploring whether satellites could eventually host large-scale AI computing infrastructure using near-continuous sunlight. The company’s earlier research describes a possible network of solar-powered satellites linked together with high-bandwidth lasers.

Project Suncatcher graphic asking how machine learning reaches space

Subscribe to our Newsletters for more Tech Stories

Project Suncatcher’s first orbital test

The prototype will carry Google TPUs into low Earth orbit. Engineers will study how the hardware handles launch forces, radiation and extreme temperatures, including radiation-related data errors known as bit flips.

Cooling is another problem. Conventional airflow cooling does not work in a vacuum, so Google is testing a design that combines heat pipes with radiators. The company says it has already run TPU workloads in a facility at the University of California, Davis, but orbital testing is needed to measure the hardware in its intended environment.

Reuters reports that the mission is designed to collect in-orbit data and identify failure points, not to demonstrate an operational commercial data centre. Google’s own account describes the launch as an early test before any larger constellation is attempted.

Google AI chips in space are still a long way from customers

Google plans to launch two satellites in 2027 to test the high-bandwidth laser links needed to connect future computing clusters. That is a communications test, not a promise of a Google data centre in orbit by next year.

The idea has attracted interest from other companies, including SpaceX and Starcloud, which are also examining data centres in low Earth orbit. The engineering case remains difficult: launch costs are high, satellites must be produced at scale, and radiation, thermal management and communications all need to work together. Reuters reports that experts consider commercial viability years away.

The UAE has its own early AI-in-space effort: the UAE’s first AI satellite, Altair-1, is scheduled to launch in October. Google’s mission is aimed at testing computing hardware in orbit, while Altair-1 is a separate UAE project. For wider context on Google’s enterprise AI position, see GlobalData’s assessment of Google and Amazon.

Reuters’ report on Google’s first Suncatcher orbital test and Google’s Project Suncatcher research announcement provide the source details. Google’s follow-up facts page says the early mission is intended to learn what works before the company considers scaling up.

Is Google building a data centre in space now?

No. The first mission is a prototype test collecting in-orbit data. Reuters reports that it is not intended to demonstrate an operational commercial data centre.

Why does Google need to test TPUs in orbit?

Radiation, launch forces, extreme temperatures and vacuum cooling create conditions that cannot be fully reproduced in a ground-based facility. The test will show where the hardware and thermal design fail or hold up.

When could AI computing in space become commercially available?

There is no commercial launch date. Google is targeting a two-satellite laser-link test in 2027, and experts cited by Reuters say commercial viability is still years away.