Why Did Microsoft Sink 864 Servers to the Bottom of the Ocean?

In 2018, Microsoft’s Project Natick submerged a sealed, nitrogen-filled pod containing 864 servers 117 feet beneath the surface off Scotland’s Orkney Islands — and when it was retrieved two years later, those servers had failed at just one-eighth the rate of equivalent land-based servers.

What Was Project Natick?

Project Natick was Microsoft’s experimental undersea data center program. In June 2018, engineers lowered a pressurized steel cylinder — roughly the size of a shipping container — onto the seafloor near Scotland’s Orkney Islands. The location was chosen deliberately: Orkney runs almost entirely on wind and tidal power, meaning the entire operation could draw on clean, renewable energy. The pod sat undisturbed on the seafloor for 107 days shy of two years before crews winched it back to the surface in 2020, barnacle-encrusted but largely intact.

Why Did the Underwater Servers Fail Less Often?

The results were striking. Servers inside the pod failed at roughly one-eighth the rate of identical servers operating in a conventional land-based data center. Microsoft’s engineers pointed to two primary reasons.

Nitrogen instead of air. The pod was filled with dry nitrogen gas rather than ordinary air. On land, oxygen gradually corrodes circuit boards and degrades the tiny solder joints connecting components. Humidity compounds the damage, causing expansion and contraction that stresses electronics over time. Inside the sealed, nitrogen-filled pod, none of that happened. There was no oxygen to oxidize the hardware, and no moisture to penetrate the circuits.

No human hands. Every time a technician enters a conventional server room, small disruptions occur — a cable is nudged, a board is jostled, humid outside air drifts in. Over months and years, these micro-events add up. The Natick pod was sealed shut from day one. No one opened it, touched it, or disturbed it for two full years. Microsoft’s team concluded that human intervention may be one of the leading causes of server failure in land-based facilities.

How Did the Ocean Keep the Servers Cool?

Cooling is one of the most expensive and energy-intensive aspects of running a data center. On land, massive air-conditioning systems run continuously to prevent servers from overheating. On the seafloor off Orkney, the cold Atlantic water did the job passively and constantly, maintaining a stable temperature around the clock with zero mechanical cooling required. The ocean provided what no land-based building can easily replicate: a vast, free, and perfectly consistent heat sink.

Why Can’t This Just Be Replicated on Land?

It’s a reasonable question — if nitrogen and sealed environments reduce failures so dramatically, why not seal land-based servers in nitrogen too? Microsoft acknowledged the logic and said it would explore ways to transfer the lessons to terrestrial data centers. But there is a fundamental catch: land-based servers eventually need to be opened for repairs and hardware replacements. The moment a technician cracks a nitrogen seal, the gas escapes, humid air floods in, and every protective advantage is immediately lost. The ocean made the seal permanent in a way no land building can match. A pod that never needs to be opened never needs to compromise its environment.

What Did Microsoft Do With the Results?

In 2024, Microsoft confirmed it would not build additional undersea data centers. The project had served its purpose as a research platform rather than a scalable commercial strategy. However, the engineering insights were not abandoned. Microsoft stated it is actively applying what it learned from Project Natick to next-generation liquid-cooled data center designs on land — particularly around minimizing human contact with hardware and controlling the atmospheric environment around servers. The seafloor, it turned out, was the most stable operating environment these servers had ever experienced. The only way to discover that was to sink them and see.