Why subsea data centers can’t get afloat

The subsea data center is having a moment, but it looks destined to be a brief one. In the latest, Chinese state media has announced what it calls the world's first commercial subsea data center, with China Telecom and Tencent among the customers.

It says the core module of the center, weighing 1,300 tonnes, was deployed in 35-meter deep water off Lingshui on the southern island of Hainan last month. The facility is set to grow to 100 modules in scale and will improve energy efficiency by 40%-60% compared with a similar-scale terrestrial data center. 

The center was built by Beijing Highlander Digital Technology, a Shenzhen-listed company specializing in marine electronics and communications tech.

Tellingly, the center seems to have been funded not by a data center operator but as a demonstration project by Hainan Free Trade Port.

The respected academic website, The Conversation, recently carried a paper pointing to the potential energy-saving, cost and latency advantages of undersea data centers. It suggests several companies are "actively exploring" or building subsea data centers, but does not name them.

Small edge centers

The one big name usually mentioned here is Microsoft, which spent six years testing out the technology under the code name Project Natick.

After the biggest trial, a two-year deployment off the Orkney Islands, concluded in 2020, a Microsoft corporate blog declared the underwater sites were "reliable, practical and use energy sustainably." In particular, the Microsoft team saw the lean and flexible subsea tech ideal for meeting the demand for small edge centers offshore close to where the majority of people live.
Yet since that last apparently triumphant test it's been crickets. There's no sign Microsoft has deployed an underwater data center anywhere. No one at Project Natick or any of Microsoft's corporate PR offices has responded to Light Reading's queries. No other data center player is using the technology.

Most likely this is one of those great ideas brought down by small but critical practicalities.

The most obvious problem is connectivity. If you're going to deploy these edge centers in numbers, then that is a fiber cable project with a lot of cost and complexity: either one long cable serving multiple undersea sites and landing at existing cable stations, or a lot of small cables requiring fresh cable landings. Either of these are expensive and the ROI on this untried technology is unclear.

Then there are the approvals required, a daunting task if you're landing multiple cables. A lot of coastal councils aren't going to welcome cable-laying teams tramping across their shores, and those who do will still insist on the paperwork.

The other problem is security. Subsea infrastructure these days is a soft target. We have seen cable cuts being attributed to China (here and here) and Russia. In the current febrile geopolitical climate, why increase the risk?

It’s no wonder commercial companies are not sinking their money into this. The subsea data center looks to be another promising concept with a business case that's underwater.