A startup named Panthalassa has raised $200 million to build AI data centers that float in the ocean and draw power from waves. The concept transforms a land-based energy and cooling problem into a data transmission challenge, with computing nodes stationed far from traditional power grids and real estate constraints.
The latest funding round of $140 million, announced May 4 and backed by investors including Palantir co-founder Peter Thiel, will help complete a pilot manufacturing facility near Portland, Oregon. Panthalassa plans to deploy wave-powered nodes that generate their own electricity, cool onboard AI chips with surrounding seawater, and transmit inference results to customers worldwide via satellite links.
Each node resembles a large steel sphere riding on the ocean surface, with a vertical tube extending deep underwater. Wave motion drives water upward through the tube into a pressurized reservoir, where it is released to spin a turbine generator. The company claims this design eliminates the need for grid power at the compute location and could dramatically reduce the freshwater consumption associated with traditional data center cooling.
Panthalassa has already tested earlier prototypes, including Ocean-1 in 2021 and Ocean-2 during a three-week trial off the coast of Washington state in February 2024. The newest prototype, Ocean-3, stretches roughly 85 meters in length and is scheduled for testing in the northern Pacific Ocean later this year. CEO Garth Sheldon-Coulson has stated ambitions to eventually deploy thousands of these nodes across the world's oceans.
Experts note significant hurdles remain. Satellite bandwidth limitations could constrain real-time coordination between nodes, and physical maintenance at sea is far more complex than servicing a land-based facility. Still, the approach offers a novel answer to the energy and land shortages slowing conventional data center expansion.
Why it matters
Land-based data center expansion is running into hard limits across power grids, zoning opposition, and cooling constraints. If ocean-based compute proves viable even in niche scenarios, it could relieve pressure on congested metropolitan infrastructure and unlock AI compute capacity in regions where building onshore is politically or environmentally difficult. For cloud providers and enterprises dependent on scalable inference, unconventional infrastructure may soon become a genuine part of capacity planning rather than science fiction.