Microsoft scrapped its 'Project Natick' underwater data center trial — here's why it was never going to work

Microsoft recently canned Project Natick, its pioneering underwater data center project, after just over a decade – and experts have told ITPro the project was never a truly feasible option for businesses. 

The firm began the project back in 2013 and saw some fairly successful results after sinking the first servers in 2018. With 855 servers submerged and left unattended for just over two years, Microsoft found the success rate higher than the 135 unsubmerged servers they used for comparison. 

Only six of the submerged servers failed compared to eight of those on dry land, suggesting potential in the project. Underwater data centers are particularly attractive from a sustainability standpoint owing to the ease with which liquid cooling can be performed. 

As Cadence Design Systems engineer Mark Seymour told ITPro, however, there are limiting factors that make the underwater data center less appealing despite its cooling potential. 

“It's obviously a potential solution to use liquid cooling by going into the biggest volume of liquid we have available,” Seymour said. “But whilst it's very attractive from the cooling point of view, obviously it’s less attractive from the operational point of view.”

Seymour added that underwater technology presents a problem for firms as they are not able to update it or upgrade it as easily as they could on land. 

“It's just probably not the easiest way to be flexible in a very fast-changing world,” he added.

Servicing data centers in these environments would present a similar challenge, with workers required to travel underwater to perform system alterations or fixes. 

Seymour added that, even now, there are complaints about servicing servers in tanks, so “you can imagine” the sort of issues people might have should the servers be below sea level. 

According to Microsoft, no more underwater data centers are being built, though the firm acknowledged its important learnings from the project. 

"My team worked on it, and it worked. We learned a lot about operations below sea level and vibration and impacts on the server. So we'll apply those learnings to other cases," Noelle Walsh, head of Cloud Operations and Innovation at Microsoft, told Data Center Dynamics.

Others have attempted to undertake similar projects such as in China, where 100 units have been submerged off the coast of Sanya, a coastal city in the country's south. 

Enterprises can take different approaches to data center cooling

Though server submersion may be an impractical approach to large-scale liquid cooling, there are other options for data center operators looking to cool their systems.  

“There's plenty of ways of putting the heat in closer contact with liquid than via the intermediate of air, that's really what we're talking about,” Seymour said. 

One such method is direct-to-chip, which Seymour described as “probably the most ready for market and most widely used”. This is where liquid is piped into a “cold plate” on the chip that helps remove heat from servers. 

The alternative to this method is to completely submerge the equipment in liquid, technically a more effective way of removing heat, though one that relies on a large amount of liquid use and compatibility of the equipment with the liquid itself.

For enterprise use, there are different advantages to consider. direct-to-chip, for example, is “probably most straightforward” for those looking to reuse an existing data center with a mind to improve cooling systems.