A data center in space: what Google Suncatcher is really testing

Headlines about a “data center in space” sound like science fiction, but the actual project is more modest—and, from a technical standpoint, more interesting. On October 1, 2026, Google launched a prototype satellite into orbit equipped with its own TPU chips, which are designed for artificial intelligence computations. This is not a fully operational data center, but rather the first test to see if such chips can survive and function in a space environment.

What was actually omitted

The prototype, called MVP, was launched by SpaceX’s Falcon 9 rocket as part of the Transporter-18 mission from Vandenberg Air Force Base in the U.S. The satellite was built by Planet, which operates a fleet of small observation satellites, and Google installed its chips in it. According to reports, the device is about the size of a refrigerator and contains four TPU chips, whose performance is roughly equivalent to that of a single server in a traditional data center. It is powered by solar panels with an output of about 1 kW.

What the test checks

Google itself emphasizes that this is not a fully operational orbital data center. The goal of the first flight is to determine where such a device might fail, so that weaknesses can be addressed in future missions. Two questions are particularly important.

Radiation

The chips were previously tested on Earth in a laboratory at the University of California, Davis, where radiation was simulated. According to the company, they performed “remarkably well,” but some things can only be verified in actual space.

Heat dissipation

Heat dissipates more slowly in a vacuum than on Earth, where air or water helps with heat transfer. According to the mission description, the satellite therefore runs artificial intelligence calculations in 15-minute bursts and then shuts down to allow the radiators time to dissipate the heat.

What Suncatcher Does Not Confirm

• That data centers in space are practical. According to analyses, they would make sense primarily on a large scale and only if the cost of launching payloads into orbit drops significantly.
• That a satellite locally processes data from other satellites. We did not find any mention in the official project description that the goal is to process data from satellites without sending gigabytes back to Earth. The main stated objective is to verify whether machine learning can be deployed on a large scale in space using solar power.
• Initial results. We were unable to find any results from orbit in the available sources. Therefore, we cannot confirm the mission’s success.

What's Next

The original plan called for two prototype satellites in early 2027, which would also test laser links between satellites. Google decided to integrate the chips into an existing Planet satellite and test them sooner. According to reports, the laser links will not be tested this summer, but rather on the two satellites in 2027. This is a crucial part of the concept, as a space data center would require a high-speed network between satellites comparable to that found in terrestrial data centers.

This isn't the only attempt

Other companies are also discussing distributed computing systems. According to reports, SpaceX has announced plans for a constellation of computing satellites, the number of which could reach into the millions. These are plans, and their implementation cannot yet be confirmed.

Conclusion

The Suncatcher prototype is more of an experiment than a product: it verifies whether the chips can handle radiation, solar power, and heat dissipation. Until the results are published and the satellites’ ability to communicate via laser links is demonstrated, this remains a research concept, not a replacement for terrestrial data centers.