A team of technicians in San Francisco recently completed rigorous testing on a satellite developed by Google for an ambitious project to operate artificial intelligence (AI) data centers in orbit. The satellite, named MVP, underwent vibration tests designed to simulate the shaking it will experience during launch, confirming the durability of its components, including solar panels and specialized AI chips. Google plans to launch the satellite into Earth’s orbit on Thursday aboard a SpaceX Falcon 9 rocket from Vandenberg Space Force Base in California.

Project Suncatcher, as the initiative is known, seeks to develop AI data centers powered by solar energy in space. While MVP represents only a preliminary step, carrying four tensor processing units roughly equivalent to the computing power of a single data center server, it is intended to test the viability of operating such hardware in the harsh environment of space. The satellite’s solar panels are expected to generate about one kilowatt of power, enough to operate the AI chips and run simple AI queries over a mission planned to last one year, although the satellite could remain in orbit for up to six years before re-entering Earth’s atmosphere.

Google executives, including James Manyika, the company’s senior vice president for research, caution that the trial is exploratory and unlikely to yield operational data centers within the next several years. Manyika likened the effort to Google’s early and prolonged research into driverless cars, emphasizing the long-term nature of the project. Experts also note that scaling such technology would involve significant engineering challenges and substantial investment, with Brandon Lucia, a professor of electrical and computer engineering at Carnegie Mellon University, describing the endeavor as “uncharted waters.”

The project was conceived by Blaise Agüera y Arcas, a Google vice president and AI researcher, who envisioned the potential to harness solar energy in space to power AI systems amid rising energy demands. After conducting radiation tests at the Crocker Nuclear Laboratory in California, which exposed AI chips to particle accelerator-generated doses simulating five years in space, Google found that the chips were resilient, with software resets mitigating radiation-induced errors known as “bit flips.”

Following internal presentations that included Google’s CEO Sundar Pichai and co-founder Sergey Brin, the company moved forward in collaboration with Planet Labs, a satellite imagery provider in which Google holds an investment. Planet Labs helped expedite the project by providing a pre-existing satellite platform for Google’s AI chips. An innovative cooling system was developed to manage the heat generated by the AI processors in the vacuum of space, employing layers of conductive materials and radiator panels, since traditional fans cannot operate in orbit.

Google’s roadmap includes launching additional satellites in 2025, with future ambitions to deploy constellations comprising fleets of more than 80 satellites operating in concert to deliver AI processing services. The company is also exploring designs for much larger satellites, potentially the size of a soccer field, to accommodate extensive computing infrastructure.

While Project Suncatcher remains experimental, company officials acknowledge the potential for significant transformation in data center operations over the coming decade. Travis Beals, senior director of product management for the project, noted that if the initiative achieves success without setbacks, it may indicate insufficient risk-taking and limited learning, highlighting the exploratory nature of this venture into space-based AI computing.