Alphabet has placed its first Project Suncatcher prototype satellite into low Earth orbit, launching four of its homegrown Trillium TPU chips on a SpaceX Falcon 9 rideshare mission on October 1. The company confirmed ground contact with the satellite and said it was operating as expected. The mission, however, is explicitly a research prototype — not a working orbital data center.
The satellite, roughly the size of a refrigerator, was built with satellite imagery company Planet Labs and flew aboard SpaceX's uncrewed Transporter-18 mission from Vandenberg Space Force Base in California. Together, its four TPUs provide roughly the computing power of a single server in a ground data center, supplied by about 1 kilowatt of solar power.
The mission's core purpose is to find out what Google does not yet know: whether its chips keep working after the physical stress of launch, sustained radiation, and the thermal extremes of orbit. Google said it plans to operate the satellite for about a year.
Why the chips can only run 15 minutes at a time
In the vacuum of space there is no air to carry heat away from the chips, so the prototype cannot run continuously. Google said the TPUs will run a version of its Gemma AI model for about 15 minutes at a time before shutting down to cool, using heat pipes and radiator surfaces to reject heat into space.
The orbit itself is part of the design. The satellite flies a dawn–dusk sun-synchronous orbit, keeping its solar panels in near-continuous sunlight. Google says a solar panel in the right orbit can be up to eight times more productive than a comparable installation on the ground, generating power almost continuously and reducing the need for heavy batteries.
The economics: 1,800 Starship launches
Even if the technology works, the economics remain the bigger open question. In a peer-reviewed paper published in Joule, Google estimated that for space-based data centers to become economically viable, launch costs would need to fall to roughly $200 per kilogram. Reaching that level, it calculated, could require moving about 370,000 tonnes of cargo to orbit.
The estimate assumes a roughly 20% annual cost decline and payloads of about 200 tonnes per mission. Under those conditions, Google calculated that around 1,800 Starship launches over a decade — about 180 per year — would be needed. Starship has not yet flown anywhere near that cadence, and Google has indicated the cost target is unlikely to be reached before around 2035.
A partner and a rival
The launch also illustrates an unusually tangled corporate relationship. Alphabet is a significant investor in SpaceX, with a stake reported to be valued at over $82 billion, and the two companies are close partners even as their AI divisions compete with one another.
SpaceX is also pursuing its own orbital computing plans. Chief operating officer Gwynne Shotwell said in September that the company plans to deploy "supercompute in space" in 2027, while SpaceX has separately indicated it expects to begin deploying orbital AI compute satellites as early as 2028.
What happens next
Google's next step is already defined: in 2027 it plans to launch two more satellites to test high-bandwidth laser communication links between spacecraft flying in formation — a prerequisite for the larger clusters, each carrying dozens of TPUs, that the project envisions.
Industry researchers note that the remaining hurdles are substantial: launch capacity and cost, cooling without air, chips that can withstand extreme temperatures and radiation, repairs that are impossible at orbital altitudes, and growing orbital debris and congestion.
What is confirmed: the prototype launched successfully on October 1 aboard Transporter-18, made contact with ground controllers, and is operating as expected; it carries four TPUs with roughly one server's worth of computing power on about 1 kilowatt of solar power.
What is unknown: how the TPUs will actually perform across a year of radiation exposure and thermal cycling, and whether behavior observed in ground testing will hold in orbit.
What is awaited: in-orbit telemetry over the coming weeks and months, and the planned 2027 two-satellite mission to test laser links between spacecraft.