Project Suncatcher: Google Plans to Put AI Data Centers in Space

Putting artificial intelligence (AI) in space might sound like something out of a science fiction movie, but Google is seriously exploring it. Google has revealed a research project called Project Suncatcher, which aims to place data centers in orbit on solar-powered satellites. These space-based data centers would rely entirely on clean and unlimited solar energy, available 24 hours a day.

The idea behind Project Suncatcher is simple yet ambitious — to reduce the massive energy consumption and carbon emissions linked with AI cloud computing on Earth. By moving data centers to space, Google hopes to eliminate the need for power plants that release harmful CO2 into the atmosphere.

Project Suncatcher: Google Plans to Put AI Data Centers in Space

According to Google’s senior director, Travis Beals, space could be the best place to scale AI computing in the future. In his blog post, he explained that solar panels in the right orbit could be up to eight times more productive than those on Earth. Unlike Earth-based systems, which depend on daytime sunlight, satellites can generate almost continuous power, significantly cutting the need for large battery storage.

However, sending data centers into orbit is not as simple as it sounds. The plan involves fitting Google’s Tensor Processing Units (TPUs) — the company’s specialized AI chips — into solar-equipped satellites. These chips would operate in extreme conditions, where radiation levels are far higher than on Earth.

Radiation is one of the biggest challenges for electronics in space. High levels can damage chips and cause them to fail. Google, however, has already tested its TPUs for radiation resistance. The company claims that these chips could last for about five years in orbit without suffering permanent failures.

Another major challenge is communication. AI systems require high-speed, low-latency data links to process massive amounts of information quickly. Google estimates it would need data speeds of tens of terabits per second between satellites. It is a difficult task in space, where transmitting data over long distances consumes much more power.

To solve this, Google may position multiple TPU satellites in proximity. This would help maintain faster data transfer speeds and reduce the fuel needed for orbit corrections.

See Also: Google Integrates Gemini AI Into Maps for Smarter, Conversational Navigation

The biggest question, though, is cost. Sending hardware into space is extremely expensive. Yet Google’s analysis suggests that by the mid-2030s, the efficiency of orbital AI data centers could be roughly equal to that of Earth-based data centers in terms of power usage and cost.

Although Project Suncatcher is still in its early research stage, Google plans to conduct real-world tests soon. The company has partnered with Planet, a satellite imaging firm, for a learning mission that will launch two prototype satellites by 2027. These trials will test how Google’s TPU hardware performs in orbit and whether satellite-based machine learning is truly feasible.

If successful, Project Suncatcher could mark a revolutionary step in how AI computing is powered — moving from Earth’s grids to the endless energy of space.