Silicon Valley bets $140M on floating AI data centers powered by ocean waves

Silicon Valley Invests in Ocean-Based Data Center Projects

Peter Thiel, co-founder of Palantir and PayPal, is leading a $140 million investment in Panthalassa, a startup developing floating AI data centers powered entirely by ocean waves 3

3

. The Series B funding round values the Oregon-based company at close to $1 billion and includes prominent Silicon Valley investors such as Salesforce CEO Marc Benioff, PayPal co-founder Max Levchin, and early Google investor John Doerr 3

3

5

5

. The investment reflects mounting pressure on tech companies to find sustainable solutions for AI's energy needs as traditional land-based infrastructure struggles to keep pace with demand.

The new capital will help Panthalassa complete a pilot manufacturing facility near Portland, Oregon, and accelerate deployment of its wave-riding "nodes" designed to generate electrical power directly in the ocean 1

1

. Instead of transmitting renewable energy to shore, these autonomous floating data centers would power onboard AI chips and transmit inference tokens to customers worldwide via satellite communication 1

1

.

How Wave-Powered Data Centers Work

Each Panthalassa node resembles a massive steel sphere sitting atop a tube-like structure, standing approximately 85 meters tall—nearly as tall as London's Big Ben or New York's Flatiron Building 1

1

3

3

. The platform uses wave motion to drive water upward through the tube into a pressurized reservoir, where it spins a turbine generator producing electricity for AI chips housed in hermetically sealed containers below the surface 1

1

3

3

.

"Panthalassa's idea transforms an energy transmission problem into a data transmission problem," Benjamin Lee, a computer architect at the University of Pennsylvania, told Ars Technica 1

1

. CEO Garth Sheldon-Coulson, a former AI and energy researcher at Bridgewater Associates, explained that "one of the key insights that we had was that it's very important to use the electricity in place. We will never be transmitting electricity back to shore" 3

3

4

4

.

The nodes also leverage seawater cooling, offering significant advantages over traditional data centers that consume vast amounts of electricity and fresh water for cooling 1

1

. "Ocean-based compute might offer a massive cooling advantage because the ambient temperature is so low," Lee noted 1

1

.

Addressing AI's Huge Energy Demand

The International Energy Agency projects that data centers powering AI could consume 945 terawatt-hours annually by 2030—exceeding Japan's entire electricity consumption 2

2

. Panthalassa's wave-powered AI approach aims to bypass overwhelmed electrical grids and deliver carbon-free computing in international waters 2

2

.

Sheldon-Coulson describes ocean waves as "twice-concentrated sunlight" because they're created by wind, which is generated by solar heat. "The waves are like a battery for sunlight and we can be capturing from it 24/7," he told the Financial Times 3

3

4

4

. He believes wave and wind power, alongside solar and nuclear, are the only clean sources capable of generating "tens of terawatts" of energy 3

3

.

Harsh Ocean Conditions and Technical Challenges

Despite the promise of offshore computing solutions, significant obstacles remain. "Wave power is an old technology and it can work, but the ocean is a harsh environment," says Jonathan Koomey, a former Lawrence Berkeley National Laboratory researcher and expert in data center energy consumption. "The salt and the waves are effective at causing trouble for machinery" 2

2

.

Panthalassa designed its nodes to withstand these conditions by making them largely solid with no hinges, flaps, or gearboxes that might break down 3

3

4

4

. Company job listings describe the goal of nodes "surviving for more than a decade in the harshest ocean conditions" while lasting "without human maintenance or intervention" 1

1

.

Latency issues present another hurdle. Relying on satellite links like Starlink to transmit data between nodes and customers means dealing with limited bandwidth and signal delays compared to fiber-optic cables 1

1

2

2

. "Satellites could communicate perhaps hundreds of megabits per second per terminal, which is feasible for real-time responses to prompts and queries," Lee explained. "But frequent communication and coordination between nodes may be challenging" 1

1

.

This makes the platforms most practical for neural network workloads that run for hours or days before returning results—like training advanced models or running scientific simulations—rather than consumer-facing AI applications requiring fast response times 2

2

4

4

.

From Prototypes to Commercial Deployment

Panthalassa has tested several earlier prototypes, including Ocean-1 in 2021 and Ocean-2 during a three-week sea trial off Washington state in February 2024 1

1

5

5

. The Ocean-3 prototype is scheduled for testing in the northern Pacific Ocean later in 2026, with commercial deployments planned for 2027 1

1

5

5

.

The nodes are assembled on land, towed horizontally behind boats, then flipped upright in deep water before using their hull shape to self-propel to designated locations without engines 3

3

4

4

. Sheldon-Coulson told CBS he hopes to eventually deploy thousands of nodes 1

1

.

Broader Context of Data Centers at Sea

Panthalassa's approach is more ambitious than previous ocean-based efforts. Microsoft's Project Natick experimented with underwater servers in 2015 and 2018, showing that sealed, seawater-cooled systems could achieve lower failure rates than land-based systems, though Microsoft decided against commercialization 1

1

. Chinese companies have deployed underwater data centers near Hainan Island and Shanghai, while Singapore's Keppel is constructing a floating data center 1

1

.

Other companies exploring offshore solutions include Aikido Technologies, which is developing floating data centers integrated with offshore wind platforms, and Mitsui O.S.K., studying ship-based computing systems 2

2

. The funding surge also parallels growing interest in space-based data centers harnessing solar energy, with Starcloud raising $170 million in March 2026 5

5

.

Whether autonomous floating data centers can scale remains uncertain. Jacqueline Davis at the Uptime Institute notes that "power and networking are the top two root causes of data center outages" and "can each be uniquely difficult to manage in a remote environment with little to no staff" 2

2

. For now, Panthalassa is betting that offshore computing powered by ocean waves can provide a sustainable solution for AI's energy needs faster than expanding land-based infrastructure.