Google has unveiled Project Suncatcher, an ambitious research initiative to explore solar-powered satellite constellations equipped with Tensor Processing Units (TPUs) and free-space optical links. The project aims to investigate how large-scale machine learning (ML) workloads might one day be performed in space, tapping near-constant solar energy while reducing dependence on terrestrial infrastructure and resources.
According to Google’s new preprint paper, “Towards a future space-based, highly scalable AI infrastructure system design,” Project Suncatcher proposes compact clusters of networked satellites orbiting in sun-synchronous low Earth orbit. These satellites could continuously harvest solar energy and interconnect through multi-channel dense wavelength-division multiplexing (DWDM) optical transceivers, achieving inter-satellite links of up to tens of terabits per second. Early lab tests achieved 800 Gbps per direction (1.6 Tbps total) between two nodes—an early proof of concept for distributed compute in orbit.
The study also models the orbital dynamics of large, tightly clustered constellations, showing that with separations as close as 100–200 m, stable formations are feasible with minimal station-keeping maneuvers. Radiation testing of Google’s v6e Trillium TPU demonstrated resilience well beyond expected five-year mission doses. Meanwhile, projected launch cost declines—to below $200/kg by the mid-2030s—could make orbital data centers economically competitive with terrestrial facilities. Google plans a joint learning mission with Planet by early 2027 to test two prototype satellites with onboard TPUs and optical interlinks.
• Google envisions tightly integrated solar-powered AI compute satellites connected by terabit-scale optical links
• Bench-scale tests achieved 1.6 Tbps bi-directional optical throughput between nodes
• Constellations modeled for 650 km sun-synchronous orbits with stable formations under 1 km radius
• Trillium TPUs exhibited strong radiation tolerance up to 15 krad(Si)
• Launch partnership with Planet aims to deploy first test satellites by 2027
“Our new research moonshot, Project Suncatcher, asks how we might scale AI compute beyond Earth—leveraging abundant solar energy in space to minimize our impact on planetary resources,” said Blaise Agüera y Arcas, Vice President and Google Fellow.
🌐 Analysis: Project Suncatcher signals Google’s intent to push the boundaries of AI infrastructure beyond terrestrial limits, combining the company’s strengths in custom silicon and AI systems engineering. The initiative echoes recent industry trends in space-based compute and optical networking—areas also explored by Amazon’s Project Kuiper and SpaceX’s Starlink laser mesh. Google’s collaboration with Planet marks its first step toward validating on-orbit ML compute and inter-satellite networking.
🌐 We’re launching the “Data Center Networking for AI” series on NextGenInfra.io and inviting companies building real solutions—silicon, optics, fabrics, switches, software, orchestration—to share their views on video and in our expert report. To get involved, send a note to jcarroll@convergedigest.com or info@nextgeninfra.io.
