Data centers run hot. Every server racks up heat as it works, and cooling that heat can eat a quarter or more of the electricity a data center uses. On land, that cooling also drinks fresh water — often in places that do not have much to spare. So an idea that sounds almost too neat has been moving from concept to reality: put the data centers in the sea. The ocean is cold, it is enormous, and it does not send a bill.
It is a genuinely clever engineering idea. It is also an experiment with a living system, and the questions it raises are not the ones the marketing tends to mention.
What is actually being built
The technology is further along than most people realize. Off the coast of Shanghai, a commercial underwater data center has been running since spring — roughly two thousand servers sealed in pods on the seabed at a depth of about thirty-five meters, powered by a nearby wind farm. Japan has a floating demonstration at Yokohama. Singapore is building a larger version, and South Korea is working on an underwater model of its own.
The engineering logic is sound. Seawater cooling removes the need for the water-hungry cooling towers used on land. The pods sit on the seabed, using the surrounding water as a heat sink, and the companies behind them report that their monitoring shows only a small temperature rise in the water a short distance downstream from the installation.
The logic is sound. The gap is that one installation’s measurements say very little about what twenty installations in the same bay would do.
What the heat actually does
Here is the honest part of the story that tends to get glossed over. The data centers do not get rid of the heat. They transfer it into the water, and warm water is not neutral for marine life.
Warmer water holds less oxygen. Species that are sensitive to temperature shift or suffer. In a coastal area that already hosts fisheries, shellfish beds or coral, even a modest, persistent temperature rise in a localized area can matter. The companies point to their own monitoring — one installation, one assessment, a temperature change of well under a degree. That is a starting point for an assessment, not a conclusion about cumulative effects. No one has measured what a cluster of installations, running for a decade, does to a bay that also has to support fish, tourism and local livelihoods.
Then there are the things that are harder to measure and easier to overlook: noise from the pods, electromagnetic fields from the cables, the physical presence of the structures themselves acting as artificial reefs and changing local species composition. None of this is necessarily catastrophic. All of it is genuinely unstudied at the scale being proposed.
The regulatory gap
The uncomfortable fact is that the rules governing this were written for other kinds of infrastructure. Under the law of the sea, countries have authority over structures in their own waters, and underwater data centers go through normal planning and permitting like anything else. But the rules were designed for cables, pipelines and platforms — not for sealed boxes of hot electronics sitting on the seabed and rejecting heat into the water around them.
The gap shows up in practice. One coastal town in Maine, where a company proposed a tidal-powered underwater data center, voted to block any such project for a period while it studies the question. The federal permit process sent the application back over missing information about tribal consultation and incomplete specifications. The system is working — slowly, cautiously — but it is working without a clear framework for what it should actually be evaluating.
This is the pattern of every new technology that touches a shared resource: the market moves first, the regulations catch up, and the period in between is where the questions get answered by default rather than by design.
The bigger question the ocean is asking
Step back and the underwater data center is a small instance of a large pattern. The ocean has become the place where the digital economy is parking its physical problems — its heat, its waste heat, increasingly its ambitions. It is the same logic that once put barrels of waste in the deep Atlantic: out of sight, out of mind, at a price that gets paid somewhere else.
The difference is that this time the technology is new enough to choose differently. The ocean is not a landfill and not an infinite heat sink. It is a system with limits, and the limits are not theoretical — they are visible in every fishery that has declined, every reef that has bleached, every species that has moved.
The underwater data center is not a bad idea. It is an idea with genuine merit and genuine unanswered questions. The measure of whether the industry handles it well will not be the cleverness of the engineering. It will be whether the deployment rate is matched by the pace of honest study — whether anyone is actually measuring what the warm water does, year after year, to the life around the pods, and whether the rules get written before the damage, not after.
The ocean has been absorbing human inventions for centuries, and it has never once sent a bill for it. That does not mean the bill does not exist. It means it is being paid by something — and someone — we have not been looking at.