Who Has the Power
A company called Panthalassa has raised $140 million in new funding to develop and deploy autonomous, floating AI computing nodes powered by ocean waves, with the Series B round bringing its total funding to $210 million. The round was led by Peter Thiel, the Palantir co-founder, and the company says the money will help complete a pilot manufacturing facility near Portland, Oregon. Panthalassa also plans to deploy its Ocean-3 pilot node series in the northern Pacific Ocean later in 2026.
The pitch is simple enough: instead of building another giant AI data center on land, Panthalassa wants to push computing power out to sea. Ocean waves would generate electricity, seawater would help with cooling, and onboard computing systems would process AI prompts before sending the results back to land through low-Earth-orbit satellites. The people making the decisions are the ones with the capital, the chips, and the infrastructure; everyone else gets the consequences later.
What They Want the Ocean to Do
Panthalassa's floating nodes are designed to capture wave motion and turn it into electricity. The company says it has spent a decade developing the technology behind its power generation, onboard computing and autonomous ocean operations. Its earlier Ocean-1, Ocean-2 and Wavehopper prototypes were tested in 2021 and 2024. The company's plan is to use those chips for AI inference, the part of AI where a model responds to a prompt after it has already been trained.
Panthalassa says its nodes would operate far from shore in wave-rich parts of the ocean. The goal is to use that wave energy directly onboard instead of sending the power back to land. Garth Sheldon-Coulson, Panthalassa's co-founder and CEO, said, "We've built a technology platform that operates in the planet's most energy-dense wave regions, far from shore, and turns that resource into reliable clean power." The ocean also offers cold surrounding water, which could help cool the chips onboard.
That setup is sold as efficiency, but it also shifts the burden of infrastructure farther away from public view. The company wants the sea to absorb the heat, the motion, and the maintenance headaches while the computing work keeps serving the same AI economy that keeps demanding more power, more hardware, and more extraction.
What Breaks, Who Fixes It
The ocean may help with power and cooling, but it creates another problem: connection. Traditional data centers rely on high-capacity fiber-optic connections because they need to move huge amounts of data fast. A floating node far out at sea may depend on low-Earth-orbit satellite links. That can work for some AI responses, but it may be slower and more limited than fiber. Panthalassa's press release says Ocean-3 testing is meant to demonstrate AI inference and refine manufacturing before commercial deployments in 2027.
There is another practical question: what happens when something breaks? Panthalassa says it is developing autonomous systems meant for harsh ocean conditions. The ocean is brutal. Saltwater eats away at equipment. Storms can turn a routine repair into a major operation. Constant motion also puts stress on the hardware. For this plan to work, Panthalassa will have to show that each node can keep running for years in harsh ocean conditions without frequent human repairs.
That is the real hierarchy here: a remote machine built to keep serving AI prompts while the labor, risk, and repair burden are pushed into hostile conditions. The system is designed to keep computing moving even when ordinary maintenance becomes difficult, expensive, or dangerous.
The Wider Infrastructure Race
Ocean data centers are not new. Microsoft experimented with underwater data center servers through Project Natick, including tests in 2015 and 2018. Those tests showed that sealed underwater servers could run reliably while using seawater for cooling, with Microsoft reporting a lower failure rate than comparable land-based systems. Microsoft later ended the project. Chinese companies have also reportedly pushed ahead with underwater data center projects near Hainan and Shanghai. Keppel has explored floating data center designs in Singapore, where land constraints make the concept especially attractive.
Panthalassa's plan combines wave power with onboard AI chips and satellite-based results, and it depends on floating nodes that would need to operate far from the kind of support a normal data center gets. For now, this will not change how your phone or computer works. Panthalassa argues its approach could reduce the need for new data centers and power plants on land, but the company still has to prove the system can work reliably at sea.
If ocean-based AI moves beyond testing, it could also raise questions about marine maintenance, environmental oversight and who controls computing infrastructure in international waters. For now, the money is in, the pilot facility is on the way, and the machinery of AI expansion is being pushed farther offshore.