A group of servers will be quietly bobbing on the open ocean somewhere off the northern Pacific coast later this year, processing requests, storing data, and harnessing the energy of the waves below. There are no complaining neighbors. There is no zoning board to petition. There isn’t a local politician in the way. Only water, calculations, and the specific aspirations of those who are fed up with the land.
Things are moving in this direction. To be honest, it’s difficult not to find it both impressive and a little unsettling as you watch it happen.
The U.S.-based startup Panthalassa, which is currently valued at about $1 billion, is developing what it refers to as a fleet of wave-powered floating data centers. Earlier this month, Palantir led a $140 million funding round led by longtime contrarian investor and co-founder Peter Thiel.
Thiel used almost cosmic language to announce the deal: “The future demands more compute than we can imagine.” These days, extraterrestrial solutions are not just science fiction. The ocean frontier has been opened by Panthalassa. It would be simple to brush off that kind of talk, the kind that sounds like it belongs in a Jules Verne paperback, but the money is real, and the issue it aims to address is also very real.
In ways that weren’t the case five years ago, building data centers on land has actually become challenging. According to recent reports, nearly half of the US facilities that were supposed to open this year have either been postponed or completely canceled. Although the causes vary, such as disagreements over water rights, grid capacity, and community opposition, the trend is constant. Locals are discovering that organizing is effective in towns all over the nation.
They are making noise, filing injunctions, and attending zoning meetings. The data center, which was supposed to be constructed discreetly on the outskirts of town, is now the subject of front-page news and is extremely complicated. Despite spending more than $600 billion on infrastructure in the last few years, tech companies are discovering that money is insufficient to overcome barriers.
In contrast, the ocean does not have a city council. Most people are unaware of this place’s longer history. In 2018, Microsoft submerged an experimental underwater data center off the coast of Scotland; it was recovered in 2024. Although the results were intriguing—the isolated, low-oxygen environment actually resulted in lower hardware failure rates than comparable land-based facilities—the project was ultimately abandoned because it was judged unsuitable for commercial deployment. There have reportedly been comparable studies conducted in China. The concept has been, in a sense, floating around for some time. The urgency has changed.
The primary benefits of an underwater data center are natural cooling from the steady ocean temperature and isolation from the chaotic, unpredictable conditions of land, according to Md. Jahidul Islam, a professor of electrical and computer engineering at the University of Florida. The deep ocean consistently and consistently maintains a temperature a few degrees above freezing, which is crucial when you take into account that traditional data centers use a startling amount of their energy budget simply to prevent servers from overheating.
The Colossus training facility, built by Elon Musk’s xAI in Memphis, Tennessee, uses enough electricity to power 200,000 homes every year. A major component of that appeal is cooling. The ocean provides free, endless cold, something that no air conditioning system could ever provide.
However, Islam is cautious about the boundaries of that appeal. He points out that those same benefits could turn into drawbacks. Equipment maintenance at depth is a true engineering conundrum. The ocean is an uncontrolled environment. When your servers are seated in a climate-controlled room in Virginia, the failure modes introduced by acoustic phenomena, pressure differentials, and marine growth simply do not exist. The maintenance economics’ viability at scale is still genuinely unknown and has not yet been demonstrated.
Panthalassa is wagering that they can solve the puzzle. According to their plan, the first commercially viable installation will take place in 2027 after an experimental fleet is deployed in the northern Pacific this year. The company is creating systems that use wave motion as a direct source of power, which would theoretically make these facilities not only cool but also mostly self-sustaining. This is a very different model from anything that is currently being used on a commercial scale.
There is a perception that the industry is currently wealthy and desperate enough to try nearly anything. The communities surrounding the proposed sites are becoming more conscious of what they are being asked to absorb due to the enormous electricity demands of AI training—OpenAI’s planned facilities would draw more power than the entire state of New England at peak consumption.
The ocean completely avoids that political conflict. The question of whether it causes new issues—logistical and environmental ones that haven’t yet fully emerged—will probably be answered the hard way over the coming years. The subsea cloud is undoubtedly no longer a theoretical concept. Now the money is in the water.
