Data centers at sea: Oregon's Panthalassa nets M led by Peter Thiel for wave-powered AI
Panthalassa has raised $140 million in Series B funding led by Peter Thiel to build floating, wave-powered AI computing nodes in the open ocean, GeekWire reported. The confirmed news is the money and the backer. The audacious claim at the center of the pitch is that wave energy harvested at sea can generate electricity for $0.02 per kilowatt-hour — roughly 35 times cheaper than current commercial wave energy and among the cheapest power ever built. Whether the number holds is an open physics and engineering question that will determine whether the whole bet makes sense.
The company, based in Vancouver, Washington, is one of several chasing ocean-based power as a solution to the AI industry's land-based electricity bottleneck. U.S. planned AI data center capacity now exceeds 50 gigawatts, but roughly half of those projects may be delayed or canceled in 2026 not because capital is scarce but because power is — grid interconnection queues stretch years and permitting timelines do not bend for artificial intelligence, StartupFortune reported. Panthalassa's answer is to move the data center offshore and generate power on-site, bypassing the transmission infrastructure that is now the primary bottleneck slowing land-based construction.
The funding round included John Doerr, TIME Ventures, SciFi Ventures, Susquehanna Sustainable Investments, Hanwha Group, Founders Fund, Gigascale Capital, Lowercarbon Capital, Unless, and WovenEarth, Ventureburn reported. Thiel framed ocean-based infrastructure as a way to expand computing frontiers beyond what traditional land systems can offer. The funding will complete a pilot manufacturing facility near Portland, Oregon, with the first Ocean-3 deployment targeted for August 2026 in the northern Pacific Ocean and commercial deployment planned for 2027, 01net.it confirmed.
The $0.02 per kilowatt-hour target is the crux of the bet. The company has targeted power generation costs as low as $0.02 per kilowatt-hour at scale, DataCenterDynamics reported. Current wave energy costs in the best resource areas — the roughly 30 kilowatts per meter of wavefront that appear in the northern Pacific where Panthalassa plans to deploy — are estimated at $0.70 to more than $1.00 per kilowatt-hour, according to the U.S. Open Energy Information database, citing IEA-OES data. European wave energy projects targeting 2025 to 2030 aim for €0.10 to €0.15 per kilowatt-hour. If Panthalassa hits $0.02, it would be roughly 35 times cheaper than current commercial wave energy and an order of magnitude below where the most optimistic offshore wave projects expect to land.
The Ocean-3 node is roughly 20 meters across at the surface and extends about 80 meters below into the water. It is self-propelled and can steer itself, moving at roughly 30 miles per day, according to the Core Memory Podcast. Rather than sending electricity back to shore, the system runs AI chips locally, using the surrounding ocean for natural cooling, Ventureburn reported. The design eliminates the transmission infrastructure that is now the primary bottleneck slowing land-based AI data center construction.
Wave energy has historically been difficult to harness consistently. The ocean does not always cooperate, and saltwater corrodes equipment faster than land-based installations. The industry has not converged on a single device architecture, and no wave energy project at data center scale has been demonstrated. Whether the physics of extracting energy from ocean swell can reliably power a facility that must stay online continuously is an open engineering question.
Panthalassa is not the only company chasing ocean-based power for compute. Several groups have proposed offshore wind or nuclear micro-reactors as alternatives to land-based grid power for AI facilities. The ocean offers advantages: unlimited heat dissipation, no land acquisition, no transmission rights to negotiate. It introduces its own failure modes: storm damage, salt corrosion, and the logistical difficulty of maintaining hardware on a floating platform far from shore. Garth Sheldon-Coulson, who co-founded the company roughly a decade ago, leads the effort.
What to watch: whether the Ocean-3 pilot actually deploys in August 2026 as planned, and whether the $0.02 per kilowatt-hour target survives contact with real ocean conditions. If it does, every land-based AI data center developer will want to know why they did not think of this first.