Panthalassa Raises $140M to Build Wave-Powered Floating AI Data Centers in the Pacific

AI’s real bottleneck is not compute — it’s power. Panthalassa, an Oregon-based startup, just raised $140 million in a round led by Peter Thiel’s Founders Fund at a valuation approaching $1 billion. The company is building floating orbs that generate electricity directly from ocean wave energy and pair it with onsite AI inference hardware, bypassing land-based power grids entirely.

Returning investors Gigascale Capital, Lowercarbon Capital, Unless, and WovenEarth also participated in the round.

The concept addresses the specific constraint killing AI infrastructure expansion. Connecting a new hyperscale data center to an overloaded power grid in the continental United States currently takes 2-4 years of permitting and infrastructure build-out. Panthalassa sidesteps that entirely: its floating systems generate and consume electricity at sea, transmitting data back to shore via low-Earth-orbit satellites. Cold ocean water provides free, abundant cooling — eliminating the other major operating cost driver for large-scale compute.

The economic case is cleaner than it sounds. Electricity is the single largest operational cost for AI data centers, typically accounting for 40-60% of total operating expenses. Grid-connected facilities also pay substantial transmission fees and face exposure to grid instability. A system that generates power at its point of consumption eliminates both costs.

The capital raised in this round will be used to complete a pilot manufacturing facility near Portland, Oregon. Panthalassa’s Ocean-3 pilot series is scheduled for deployment in the northern Pacific in 2026, with commercial systems planned for 2027.

The target workload is AI inference — running already-trained models to generate outputs — rather than the far more compute-intensive training phase. Inference is the dominant cost driver in production AI deployments. A single query to a large language model consumes less power than training but happens billions of times per day across all production AI systems. Panthalassa’s bet is that the power and cooling advantages of ocean deployment are most decisive at inference scale, where continuous uptime and low operating costs matter more than raw GPU density.

The Founders Fund backing is notable. Thiel has a history of backing infrastructure bets that look implausible until they don’t — Palantir, Stripe, and SpaceX all raised money at stages when the conventional wisdom was skeptical. A floating wave-powered data center in the Pacific is exactly the kind of contrarian infrastructure thesis that fits that pattern.

Hyperscalers are pursuing their own power solutions — Microsoft has a nuclear deal with Constellation, Amazon has contracted with Talen Energy, and Google is exploring small modular reactors. Panthalassa’s wave energy approach is fundamentally different in that it does not require proximity to existing power infrastructure of any kind.

Whether the Ocean-3 pilots deliver the reliability and cost numbers the models predict will determine whether this is a genuine infrastructure breakthrough or a well-funded novelty.