AI Data Centers at Sea: Panthalassa Raises Over 20 Billion Yen Using Wave Power

Harnessing Waves to Power AI: Panthalassa’s Grand Ambition

As land-based AI data centers face challenges such as securing sites and power supplies, Silicon Valley investors are shifting their focus to the open sea. A group of investors, including Peter Thiel, co-founder of Palantir, has poured over $200 million (approximately 29 billion yen) into Panthalassa, a company developing wave-powered offshore AI data centers.

On May 4, the company announced that it had raised $140 million in its latest funding round. This funding will be used to complete a pilot manufacturing facility near Portland, Oregon, and accelerate the deployment of its floating “Node” units powered by wave energy.

Turning Wave Motion into Electricity and AI Processing

Panthalassa’s Node consists of a massive steel sphere floating on the ocean’s surface, connected to a tubular structure extending vertically underwater. The up-and-down motion of the waves pushes water upward through the tube into a pressurized reservoir. Releasing this water drives a turbine generator, which in turn supplies renewable energy to onboard AI chips.

A key innovation is that the Node doesn’t transmit energy to land but instead performs AI inference processing directly onboard. The inference tokens generated by the AI models are then transmitted to customers worldwide via satellite links.

Benjamin Lee, a computer architect at the University of Pennsylvania, remarked to Ars Technica, “Panthalassa’s idea transforms the problem of energy transmission into one of data transmission.” He explained that AI models are sent to the offshore Nodes, which respond to specific prompts or queries.

The Cooling Advantage of Seawater

Offshore data centers not only offer energy benefits but also significant cooling advantages. “Ocean-based computing, with its low surrounding temperatures, could provide a massive cooling advantage,” said Lee. Traditional land-based data centers consume vast amounts of electricity and freshwater for cooling, a problem that has intensified with the rapid growth of AI demand.

Panthalassa claims that its Nodes can use surrounding seawater for chip cooling, providing another differentiating factor compared to conventional data centers.

North Pacific Test Scheduled for 2026

The latest prototype, “Ocean-3,” is scheduled for testing in the North Pacific in late 2026. According to the Financial Times, the new model is about 85 meters tall, comparable in height to London’s Big Ben or New York’s Flatiron Building.

The company has already tested several prototypes of its wave power conversion technology. In 2021, it tested “Ocean-1,” and in February 2024, it plans to conduct a three-week sea trial of “Ocean-2” off the coast of Washington State.

CEO and co-founder Garth Sheldon-Coulson expressed in an interview with CBS his hope to deploy thousands of Nodes in the future.

Challenges Ahead: Bandwidth and Latency in Satellite Communication

As ambitious as the project is, it faces numerous challenges. Relying on satellite links to transfer data between the Nodes and customers presents issues such as limited bandwidth and signal latency. This is why data centers still use fiber optic cables to handle large volumes of data quickly.

Other hurdles include maintaining hardware in harsh marine environments, potential impacts on marine ecosystems, and safeguarding against security threats such as piracy and natural disasters.

Nevertheless, with skyrocketing energy consumption and land use concerns accompanying the explosive growth of AI, the idea of utilizing the open ocean holds the potential for a paradigm shift in the industry. The extent to which Panthalassa can bring its ambitious vision to fruition will be closely watched in the coming years.

---

FAQ:

Q: How does Panthalassa’s offshore data center generate power?
A: It uses a steel sphere floating on the ocean’s surface connected to a tubular structure beneath the water. Wave motion pushes water upward through the tube into a pressurized reservoir, and releasing this water powers a turbine generator to produce renewable energy. This electricity drives AI chips directly onboard.

Q: Why build data centers offshore?
A: Land-based AI data centers face challenges such as difficulties in securing sites and the significant energy and cooling water required. Offshore centers benefit from natural seawater cooling and can leverage renewable energy from wave power, potentially solving these issues.

Q: What are the downsides of using satellite communication?
A: Satellite links have limited bandwidth and higher latency compared to fiber optic cables. These limitations could become bottlenecks for real-time, high-volume data exchanges required for AI inference processing.