Google Launches Project Suncatcher Satellite to Test AI Data Centers in Space

Google launched a refrigerator-sized prototype satellite into low-Earth orbit on October 1, 2026, marking the first real-world test of its Project Suncatcher initiative, which aims to explore whether AI data centers can operate in space.

The satellite, developed in partnership with aerospace company Planet, carries four of Google's Tensor Processing Units running a version of the company's Gemma AI model. It was launched aboard a SpaceX Falcon 9 rocket into a dawn-dusk sun-synchronous orbit, where it can maintain near-constant access to sunlight.

The mission has two primary goals: testing how the hardware handles the physical stress of launch and the harsh radiation and thermal conditions of space, and validating the use of optical inter-satellite links for distributed machine learning tasks.

Space presents unique engineering challenges for computing. Heat cannot be dissipated through air or water in a vacuum, so the satellite relies on pipes and radiators to manage thermal load. Radiation is another concern. Google researchers found in proton beam testing that its Trillium v6e TPUs are surprisingly resilient to radiation, but High Bandwidth Memory subsystems remain sensitive.

The economic case for space-based AI computing depends heavily on launch costs. Google researchers estimate that for the model to be competitive with terrestrial data centers, launch costs would need to fall to approximately $200 per kilogram by the mid-2030s. Current costs are significantly higher.

Google plans to test connections between multiple satellites in 2027, moving toward the vision of kilometer-radius formations of 81 satellites working together as a distributed compute cluster.

The company is clear that space-based computing will not replace ground-based data centers in the near term. Instead, it envisions a coexistence where specific workloads, particularly those requiring global coverage or benefiting from space-based solar power, are better suited for orbital infrastructure.

The project reflects the broader pressure on the AI industry to find new energy sources. Amazon recently signed a 20-year nuclear power deal to support a $3 billion upgrade to the Calvert Cliffs nuclear plant in Maryland, and Japan is developing a $15 billion AI data center project integrated directly with power generation.