Two major problems in data centers.

A huge amount of water and a huge amount of energy.

Imagine a floating, self-sufficient, autonomous supercomputer that uses the kinetic energy of the ocean’s own waves to generate electricity and process trillions of data points 24 hours a day—that is exactly what they are already putting into operation. The energy consumption of new supercomputers is growing at such a rapid pace that various regions are already facing energy supply constraints, water shortages for cooling, and difficulties obtaining permits for new data centers. Faced with this scenario, a U.S. startup decided to move its servers to an unlikely location: the ocean.

The problem with modern data centers goes far beyond electricity; each facility houses thousands of processors running continuously, producing enormous amounts of heat. In many cases, millions of liters of water are used daily to keep this equipment cool. Furthermore, the construction of new facilities faces environmental restrictions, power grid limitations, and growing resistance from local communities.

The solution came in the form of platforms that use mechanical systems capable of converting the continuous motion of the waves into electricity, reducing dependence on conventional power grids. Built in shipyards using naval engineering techniques, these structures were designed to operate for long periods in deep water, withstanding the harsh conditions of the ocean environment while powering high-density artificial intelligence servers.

Another key differentiating factor is cooling.

Instead of relying on large industrial systems that consume drinking water, the ocean itself acts as a gigantic natural heat sink. The heat generated by the processors is transferred to heat exchange systems specially designed to work with seawater, significantly reducing resource consumption and increasing the operation’s energy efficiency. The energy produced by the waves is used directly to power the onboard equipment, while communication with the rest of the world occurs via low-Earth orbit satellites, allowing commands to be sent and results to be received in near real time.

The company aims to begin commercial operations in the coming years, using these platforms to process artificial intelligence workloads without increasing the strain on the electrical infrastructure of major cities. If the model proves economically viable, entire fleets of oceanic data centers could operate in a distributed manner, bringing computing capabilities to regions where it would currently be virtually impossible to install large processing centers.