Microsoft is running an undersea data center to keep it cool


When waste heat is below the temperature at which you can run a turbine of some kind, the conversion efficiency is a lot lower. Like 10%. Silicon based chips can’t take more than about 100 C. So the temperature isn’t high enough to make steam.


Quite possibly greater than the alternative for the immediate surroundings(depending on currents); but (if it’s even vaguely sensible to consider the “yeah, I’m going to need that zero-touch, watertight to at least a few atmospheres, and marine-grade corrosion resistant…” mark-up); probably less overall warming than the servers plus beefy AC as in the conventional setup; and no worse than power plant style ‘use lake/sea as sink for the internal heat exchangers’ data center cooling, if that is an option.

Probably not a win if it’s being plunked down in endangered frigid goby habitat; but if it lets you skip the, not insignificant, energy costs of substantial chillers it could be lower impact overall.


I’m sure that this has nothing to do with a desire to host their servers in international waters away from pesky subpoenas.


There have been some demos, not sure if any have remained for the long term, of waste heat reuse(here’s one using it for residential heating); but dealing with the ‘wastr heat not very hot’ problem is tricky. Silicon ICs get really unhappy at not terribly high temperatures; and some amount of thermal resistance between the actual junctions and the cooling medium is unavoidable; so the coolant has to remain at a temperature even lower than the already tepid maximum safe junction temperature.

And, as Carnot tells us, heat engines are at their best when they have a very large thermal delta to work with; so it’s not a good match.

This doesn’t make it useless(worst case you can use it to pre-heat something that will then be heated further by other means; the fact that temperatures will be too low to let the computer waste heat contribute to overcoming the enthalpy of vaporization if you use water is a downer; but bringing something to 50-60 degrees from ambient is work that someone else doesn’t have to do); but it makes for tepid efficiency that really makes the cost of the capture mechanism stand out.

My personal favorite; despite it being basically a novelty, has to be the Sterling engine chipset cooler. The difference between chipset and ambient is enough to get the teeny little fan going(as long as all that heat sink you see keeping the cold side cool gets enough airflow; and it essentially acts as it’s own thermostatic control since the greater the thermal delta between chip and environment the faster the cooling fan runs to try to correct that.

No hugely practical; but fun.


Plus there’s the future-proof aspect. We’ll be needing more data storage in the future, and the ocean levels will be simultaneously rising to accommodate even more data centers. It’s a win-win!





Ah, when I wrote that I was thinking of low-temperature geothermal plants that use a fluid other than water and it looks like those are about 10 - 13% efficient.


I suspect that’s the actual R&D value to this project: get smart people thinking really hard about getting those reliability numbers up to the lots-of-nines territory. Which seems pretty deployable, even if data-boats aren’t the future.


I have owned computers that were considered boat anchors, but this beats them all.


PNM concludes wind and gridscale solar require “fossil fueled” load following backup. What backup planned?


Agreed - this was the meeting/presentation area at an event in their Reading, UK campus:


It’ll depend on the local environment, but you’re on target, it’s a very real concern

…aquatic organisms evolved in relatively thermally buffered environments, and thus they are generally more sensitive to temperature fluctuations than their terrestrial counterparts. […] Temperature increases as little as 1 to 2°C can alter communities because they are lethal to some species and can affect growth and reproduction of others. – Dodds & Whiles, 2010

To date most studies have been on the effects of industries increasing the temperature of fresh water. (Consider how various harmful aquatic flora and fauna, such as zebra mussels, take advantage of power plant and factory heat outflows.) There’s relatively little data on ocean thermal pollution, although it’s known to be a real thing, and that global climate change is having profound effects on the seas.

Right, we should not scoff at small steps, and studying the environmental impact of this data center could lead to an immense increase in our understanding of how to do it better. (Poorly phrased, but I hope you get my point!)


It wasn’t meant to be dismissive. I just wanted to suggest looking at the entire system.


I think you are absolutely correct!


I would say “states without proof” rather than “concludes” based on my cursory examination of that link.

I suspect PNM just doesn’t have the financial resources to build out enough infrastructure to avoid fossil fuels - they burn a lot of coal right now - and that’s why they made the statement as they did. They have to plan within their economic means, since the USA isn’t going to help with tax funding.

There are entire nations using no fossil fuels right now; Costa Rica used none for 300 days in 2017, and is well on the way to being entirely and permanently sustainable. Iceland, with their unique geologic advantages, produces 100% of their electricity without fossil fuels. Several other nations are already committed to a path that will eliminate use of fossil fuels, for example Sweden, Scotland, Denmark, and Finland.

It’s doable. But it’s expensive, and PNM might not have the resources.


Time for a sequel then!

closed #38

This topic was automatically closed after 5 days. New replies are no longer allowed.