Big Tech
Sep 16, 2026
Google will launch a prototype satellite carrying its Tensor Processing Units on SpaceX's Transporter-18 rideshare next week, the first in-orbit test of Project Suncatcher. Built with Planet Labs, the mission will test how the AI chips handle radiation, heat, and launch stress. Two more satellites testing laser links are planned for 2027.
[For more news, click here]
Google will send four of its Tensor Processing Units, the custom chips that train and run its AI models, into low Earth orbit next week. The prototype satellite, built with Planet Labs, will fly on SpaceX’s Transporter-18 rideshare mission and marks the first in-orbit test of Project Suncatcher, a research effort exploring whether space could one day host large-scale AI computing infrastructure.
The mission will remain modest as Google has said it is meant to gather data and identify failure points, not to demonstrate a working orbital data center. Even so, it puts real hardware behind one of the more ambitious ideas circulating in the AI industry, that limits on electricity supply on Earth could be sidestepped by moving computing off the planet.
The Case for Computing in Orbit
Training and running AI models consume enormous amounts of electricity, and power supply has become a constraint on new data center construction. Google says that solar panels in low Earth orbit can give satellites access to near-continuous sunlight and generate up to eight times more power than on earth. Alongside Google, SpaceX and Starcloud are pursuing their own plans to deploy data centers in low Earth orbit, drawn by the same promise of abundant solar energy and freedom from power limits.
The partnership with Planet gives Google a practical route to orbit in which Planet builds and operates its own constellations of Earth-imaging satellites, giving it long experience with small spacecraft. SpaceX’s Transporter rideshare missions, meanwhile, pack many small satellites onto a single Falcon 9, with Transporter-15 alone carrying 140, which keeps the cost of an experimental flight relatively low.
Radiation, Heat, and Launch Stress
Radiation in space can disrupt or degrade electronics, including through data errors known as bit flips, where a stored value switches unexpectedly. Google has already tested its TPUs while running AI workloads at the University of California, Davis, where the chips were exposed to a proton beam at the university’s Crocker Nuclear Laboratory. In a blog post, Google said its Trillium TPUs withstood a total radiation dose greater than they would receive during a five-year space mission.
As ground tests provide limited data, Google has said orbital testing is needed to understand how the hardware performs in the real environment. The prototype will assess how the chips cope with launch forces, radiation, and extreme temperatures once they are in orbit.
Another engineering hurdle faced by data centres on earth is cooling where data centers rely on air and liquid to carry heat away from processors. However, in the vacuum of space, conventional airflow cooling is impossible. Google will evaluate a design that combines heat pipes, which draw heat away from the chips, with radiators that release it into space.
Orbital Data Centers Remain Years Away
Experts have spoken about how orbital data centers are still years from commercial viability. For now, launch costs remain high, the engineering constraints are significant, and satellite production faces bottlenecks that would limit how quickly large constellations could be built.
After this first mission, Google aims to launch two satellites in 2027 to test the high-bandwidth laser links that future computing clusters would need to exchange data. Those links matter because large AI workloads are spread across many chips that must communicate constantly, and in orbit that would mean travel by laser between satellites rather than through fiber inside a building as they do on earth.
Next week’s launch will show whether chips built for climate-controlled server halls can operate reliably in orbit, and how much of what engineers learned in the lab holds up once the hardware is actually up in space. What Google learns will shape whether Suncatcher stays a research project or becomes an early step toward a new kind of computing infrastructure.
Related Articles
HUMAIN Expands MIS AI Data Center Project to 250MW
SpaceX Unveils Starbase Louisiana, a $100 Billion Spaceport Project
Related Articles
