Google is sending a refrigerator‑sized prototype satellite into low Earth orbit to test whether its custom AI chips can work in space. The launch — on a SpaceX Transporter‑18 rideshare — is the first real hardware test of Project Suncatcher. It will carry four Tensor Processing Units (TPUs) and try to answer a simple question: can orbital AI survive rockets, radiation, and the very real problem of heat?
What Google is testing in orbit
The MVP prototype, built with partner Planet, carries four of Google’s TPUs and about a kilowatt of solar power. Google says lab tests showed the chips survive big vibration and heavy radiation doses. But lab tests are not orbit. The point of this flight is to collect real data on vibration during launch, radiation effects in space, and how well the cooling system — heat pipes and radiators — actually works while circling the Earth.
Why cooling and heat are the hard part
Here is the kicker: the chips can only run for roughly 15 minutes before they have to shut down to cool off. In a vacuum you can’t just blow cooler air over hot parts. You must radiate heat away, which gets tricky during eclipse periods and when you try to run many chips at once. That 15‑minute number tells you how far away continuous orbital compute still is. If you want a satellite AI data center, you need a real solution for cooling, not a clever PowerPoint slide.
Bigger questions: security, control, and who benefits
Beyond the engineering, conservatives should ask who controls this new infrastructure. Space‑based compute raises national security and data‑sovereignty questions. Do we trust one or two Silicon Valley giants to run compute nodes above the United States, or worse, let hostile actors race them to similar capabilities? Elon Musk, Jeff Bezos, and other tech titans have all flirted with the idea of orbital data centers. This should not be a playground for tech bros to escape regulation while deciding how Americans’ data and critical services are routed.
Bottom line
This launch is a useful, small step: a live experiment that will tell engineers what still needs fixing. But it is not a green light to send whole data centers into orbit without thinking about cost, cooling, comms, debris, and security. Congress and regulators should watch closely and demand answers about national security, data control, and economic sense before we hand over any piece of critical infrastructure to an untested orbital model. Testing is fine. Reckless reliance on space compute would be anything but.
