Why Microsoft Put a Data Center Underwater

← Articles index · Home · Infrastructure · Nori · About

Topics: Infrastructure · Cooling

Microsoft's Project Natick became the public reference point for underwater data centers. The company sealed computing equipment into underwater vessels, deployed them in the ocean, and later retrieved them for analysis. The project asked a simple question with complicated engineering behind it: can servers run reliably underwater, using the ocean as a heat sink?

For searchers, journalists, and operators, Natick is often the first result associated with the category. Understanding what it did, and what it did not do, is essential before evaluating later commercial proposals.

Natick proved underwater operation is possible. It did not finish the commercial product problem.

What Project Natick was

Natick was a research and engineering program. Microsoft and partners built sealed underwater data-center vessels, placed them on the seabed, supplied power and networking, and monitored the systems during operation. After a multi-month Northern Isles deployment off Orkney, Scotland, the vessel was recovered so teams could inspect hardware condition and gather operating data.

The experiment explored several themes at once: seawater cooling, reliability in a controlled atmosphere, logistics of deployment and recovery, and whether underwater siting could reduce some land and cooling constraints associated with conventional facilities.

Why Microsoft tried it

The published motivations clustered around a few practical pressures:

• Data centers consume large amounts of energy for cooling as well as compute.
• Coastal and near-coastal regions often have population, network, and sometimes energy advantages.
• Sealed environments can reduce exposure to oxygen, moisture cycling, and human traffic that contribute to failure modes on land.
• Rapid deployment concepts are attractive if manufacturing and marine installation can be standardized.

In other words, Natick was less about novelty for its own sake and more about testing whether the ocean could change the cost, speed, or reliability envelope of cloud infrastructure.

What the results suggested

Public discussion of Natick emphasized that the underwater vessel operated successfully for an extended period and that recovered servers showed strong reliability characteristics relative to expectations for comparable land systems. Seawater provided the heat rejection path. The controlled internal environment limited some ordinary failure drivers.

Those findings are important. They do not mean every underwater concept inherits the same outcome. Vessel design, materials, connectors, monitoring, retrieval method, and scale all matter. A research vessel with a defined mission is not the same as a multi-megawatt commercial fleet supporting customer SLAs across many sites.

What Natick did not solve

Commercial subsea compute still has to answer questions Natick was not designed to close:

• How do you service or replace capacity without rare, expensive marine campaigns?
• How do modules scale from a demonstration vessel to repeatable manufacturing?
• How do sites secure firm power and diverse terrestrial fiber at customer scale?
• How is environmental performance monitored continuously for regulators and communities?
• How do operators sell reserved capacity with predictable availability?

That is why later essays distinguish research pods from modular commercial platforms. See Beyond Project Natick and the pillar guide What Is a Subsea Data Center?.

Lessons for buyers and builders

If you are evaluating underwater or ocean compute claims today, Natick is useful as a proof point and a caution:

• Treat successful underwater operation as necessary, not sufficient.
• Ask for the maintenance and expansion model, not only the cooling story.
• Separate peer-reviewed or company-published experiment results from marketing analogies.
• Compare total delivered service (power, network, latency, ops) with land alternatives in the same region.

For broader industry context, read State of the Subsea Data Center Industry and Floating vs. Subsea Data Centers.

Bottom line

Microsoft put a data center underwater to test whether sealed marine systems could cool efficiently, run reliably, and inform future infrastructure options. Project Natick strengthened the technical case that underwater compute can work. The commercial case still depends on modularity, serviceability, power, networking, and environmental accountability at scale.