Google Says Starship Needs 1,800 Launches Before Orbital Data Centers Make Sense

Google has sent its first advanced chip into orbit, a early step toward a long-discussed idea: running data centers in space. The company's own thinking about what that would take is sobering. By its estimate, SpaceX's Starship would have to launch roughly 1,800 times before orbital computing becomes a practical proposition.

The orbital test is meant to show how Google's custom silicon behaves outside Earth's protection. Radiation, extreme temperature swings and the lack of easy maintenance are all problems that terrestrial server farms never face. A chip in orbit cannot be swapped out by a technician, so reliability data from real flight is worth far more than lab simulations.

The appeal of space-based computing is easy to understand. AI workloads are pushing electricity demand and cooling needs on the ground to levels that strain power grids and local communities. In orbit, solar panels can collect sunlight almost continuously, and the vacuum offers a place to shed heat. Backers argue that, in principle, this could sidestep the land, water and grid constraints that increasingly slow down new data center projects.

The catch is logistics and cost. Building meaningful compute capacity in orbit means lifting enormous amounts of hardware, including solar arrays, radiators and the servers themselves. That is where the 1,800 launch figure comes in. It reflects how much mass would need to reach orbit, and it assumes a heavy-lift rocket such as Starship flying at a cadence no vehicle has ever come close to achieving. Starship is designed to be fully reusable, which is what could eventually make such a number thinkable, but it is still in development and has not yet shown routine, rapid turnaround.

Seen that way, Google's number works as much as a reality check as a roadmap. The economics only work if launch costs fall dramatically, and if the hardware can survive long enough in orbit to justify the expense. Both conditions remain unproven.

For now, the chip test is a modest but concrete move from slideware to hardware. Google is not promising orbital data centers anytime soon. It is gathering evidence on whether its processors can function in space, while acknowledging that the rocket industry has a long way to go before the rest of the plan adds up.