In an ambitious move that feels straight out of science fiction, Google has unveiled Project Suncatcher — a “moonshot” initiative that envisions launching AI-powered data centers into space. The goal? To harness the near-limitless potential of solar energy and overcome the growing environmental and energy challenges faced by terrestrial data centers.

The project, revealed by Google senior director Travis Beals in a company blog post, aims to deploy Google’s Tensor Processing Units (TPUs) aboard solar-powered satellites orbiting Earth. These satellites would function as data centers, running on solar energy that’s continuously available in space — an advantage that could make them up to eight times more productive than solar facilities on Earth.

“In the future, space may be the best place to scale AI compute,” Beals wrote, hinting at how the project could transform the landscape of artificial intelligence infrastructure. Google also released a preprint paper, not yet peer-reviewed, detailing early research and feasibility studies for this futuristic endeavour.

At the heart of Project Suncatcher is a simple yet revolutionary idea: build orbital data centers that can run large-scale AI computations using clean, uninterrupted solar energy. The company hopes this could drastically reduce the carbon footprint of AI processing while sidestepping the mounting concerns over energy consumption and local grid strain caused by traditional data centers.

However, turning this vision into reality presents a series of daunting technical and logistical challenges. For starters, satellite data centers would need extremely high-speed communication links, capable of supporting “tens of terabits per second,” as Google notes in its research. To achieve this, satellites would likely need to operate in tight formations — within just kilometers or less of each other — significantly closer than current satellite constellations. Such proximity raises complex issues, including the increased risk of collisions and space debris.

Radiation is another major obstacle. Electronics in orbit must endure high-energy cosmic rays and solar radiation that can damage circuitry over time. Google says its Trillium TPUs have been tested for radiation tolerance and can “survive a total ionizing dose equivalent to a 5-year mission life without permanent failures.”

Cost, of course, remains a central concern. Launching and maintaining such advanced satellites is expensive today, but Google’s cost analysis suggests that by the mid-2030s, the economics could shift. The company estimates that operating space-based data centers could become “roughly comparable” to the energy costs of running similar facilities on Earth, measured on a per-kilowatt/year basis.

Looking ahead, Google plans to put its vision to the test. The company is partnering with Planet, a satellite imaging firm, for a joint mission in 2027 to launch prototype satellites and evaluate the performance of its hardware in orbit.

If successful, Project Suncatcher could redefine not only how AI infrastructure is powered but also where it resides — turning the vastness of space into the next frontier for intelligent computing.