The AI infrastructure crisis has a new, unconventional answer: float it in the ocean and let the waves handle the power bill. Oregon-based startup Panthalassa just closed a $140 million Series B led by Peter Thiel, valuing the company at close to $1 billion. The round also drew in John Doerr, Marc Benioff’s TIME Ventures, Max Levchin’s SciFi Ventures, and South Korea’s Hanwha Group, among others. The pitch is deceptively simple — move the data centre offshore, generate power directly from ocean waves, and skip the terrestrial grid entirely.
The Problem Nobody Has Cleanly Solved Yet
Terrestrial data centres face mounting constraints: limited grid capacity, cooling water scarcity, supply chain bottlenecks, permitting delays, and impacts on local communities and infrastructure. This isn’t a future risk — it’s already operational reality. In parts of the United States, Ireland, and the Netherlands, grid operators have effectively stopped accepting new connection requests from large compute facilities. Business WireCapacity
The numbers behind the crunch are stark. AI data centre planned capacity in the United States alone now exceeds 50 gigawatts of demand, a figure straining electricity networks not designed for such concentrated load. And getting a new facility online is no quick fix — the time between site selection and power-on for a greenfield data centre in many Western markets now routinely exceeds three years. That lag is incompatible with the pace at which AI compute demand is growing. Capacity
What Panthalassa Actually Built
Panthalassa’s floating units, which it calls “nodes,” are large steel structures about 85 metres long. Most of each unit sits below the surface and includes a sealed container holding computer servers, cooled by seawater. Capacity
The energy generation mechanism is clever engineering rather than magic. Each lollipop-shaped node consists of a buoyant spherical head connected to a long, submerged vertical tube and structural frame. As ocean waves pass, the node bobs up and down, but the surrounding water moves only in small orbital paths. This relative motion between the structure and the water column induces oscillations within the tube, effectively driving seawater up and down through the system. That oscillating flow is channeled into the spherical chamber through a high-pressure jet, where it is converted into usable mechanical energy, passing through internal turbines to generate electricity. Tom’s Hardware
Critically, the units are self-propelled. The nodes can use the shape of their hull to navigate autonomously, powered by ocean waves, to targeted deep-sea areas — eliminating the grid connection bottleneck of land-based data centres. Data from the onboard AI chips gets relayed back to shore through low-Earth-orbit satellite links, cutting out the need for costly undersea cables. Capacity
The cooling economics are worth noting separately. Keeping chips cool is one of the most expensive operational challenges facing conventional data centres. Panthalassa’s approach addresses it without a single cooling tower, chiller unit, or drop of municipal water — the surrounding ocean does the work for free. Capacity
Why Wave Energy, Specifically
Solar and wind get most of the clean-energy press, but they share a critical weakness: intermittency. Waves provide something that solar and wind usually do not — consistency. During the night, solar panels go dark. When the wind is calm, turbines stop. Waves are in constant motion, caused by wind patterns that develop over thousands of kilometres. News9live
Wave energy along the US coast alone is estimated at 2,500 terawatt-hours per year, a figure that dwarfs current projections for offshore wind in the same geography. CEO Garth Sheldon-Coulson has framed this resource in expansive terms: “There are three sources of energy on the planet with tens of terawatts of new capacity potential: solar, nuclear, and the open ocean.” That’s a bold claim, but the underlying physics of wave energy density support the general direction of the argument. Capacity
One of the biggest challenges historically with wave power is the need to build costly infrastructure to move energy from the ocean to where it’s needed. Panthalassa’s approach sidesteps that problem by using power onsite to run already-trained AI models, while tapping cold ocean water to cool the hardware — solving two problems at once. GeekWire
The Honest Uncertainty Here
This is genuinely promising — and genuinely unproven at scale. Operating in harsh marine environments raises concerns about durability, maintenance, and ecological impact. Regulatory approvals and large-scale deployment could also pose significant hurdles. The Hans India
Wave energy as a technology category has attracted meaningful capital before without reaching commercial scale. Sweden’s CorPower Ocean and the UK’s Marine Power Systems have each crossed the $100 million investment threshold globally, yet neither has deployed at the kind of scale Panthalassa is targeting. The Ocean-3 nodes are still in the pilot phase. Panthalassa plans to deploy its Ocean-3 pilot node series in the northern Pacific Ocean, demonstrating AI inference capabilities and refining its manufacturing process in preparation for commercial deployments in 2027. That’s an important qualifier — this is a 2027 story, not a 2025 one. Tom’s Hardware
The investor lineup does suggest this isn’t a pure moonshot bet. Thiel’s Founders Fund has a track record of picking technically ambitious companies early. The presence of Hanwha — a South Korean industrial and defence conglomerate with serious manufacturing capability — is arguably the most interesting signal. Capital from operators with fabrication experience is different from pure VC money.
Conclusion
Panthalassa is attempting something genuinely novel: not just a new data centre design, but a complete rethink of where AI infrastructure can exist and how it gets powered. The $140 million raise, and the calibre of investors behind it, signals that the land-based data centre model is cracking under pressure and the industry is starting to fund alternatives in earnest. Whether wave-powered ocean nodes can deliver at commercial scale by 2027 is the real question. The physics are sound, the problem is real, and the funding is serious — but the ocean is an unforgiving place to manufacture and operate complex infrastructure. Watch the pilot deployment closely.