The Cloud Under the Sea

The Verge: Microsoft has placed a data center in the Scottish sea to determine whether it can save energy by cooling it in the sea. Data centers typically generate a lot of heat, and big providers try to move them to cooler countries to save on energy bills. Microsoft has been experimenting with subsea data centers for around five years, and previously sunk a data center on the Californian coast for five months back in 2015.

Today’s underwater data center will be deployed for five years, and includes 12 racks with 864 servers and 27.6 petabytes of storage. That’s enough storage for around 5 million movies, and the data center is as powerful as thousands of high-end desktop PCs. The data center will be powered by an undersea cable and renewable energy from the Orkney Islands. The cable will also connect the servers back to the internet.

Hat tip: Robin Hanson.


Couldn't these data centers be powered by the ocean waves?

Just connect a bouy on the surface to a rock on the bottom of the sea via a belt connected to a coiled spring. When the bouy moves up and down, it can rotate a turbine.

Come to think of it, why can't the rest of the planet be powered this way?

Because if we did it too much, the moon would move further away from the Earth.

Most wave energy comes from wind, not the moon.

Anyway, can you explain why capturing wave power would move the moon further from Earth? Is the counterfactual a situation where some of the energy is not dissipated by turbulence/shore/etc. and returned to the moon? Do you know of any calculations of this gravitational rebound?

Thank you.

Firstly, yes, I think Mobile meant "Tidal" rather than "wave" energy; you're right to point this out.

Then....mmm...he might have the moon effect backwards though.

Currently the world's water "follows the moon" as it orbits the earth. Counter-intuitively, this ADDS energy to the moon from the earth's rotation (the moon is getting further away; greater gravitational potential energy. The earth is spinning slower; less kinetic energy)

Large-scale tapping of tidal power would reduce this effect. I suppose, done sufficiently, even reverse the earth would spin faster and the moon would descend....

An easier way to think about it that tidal power INCREASES thermal losses through friction per unit time. Assuming the earth-moon-tidal system is in dynamic equilibrium, this additional energy outflow has to come from somewhere, and the only place it can come from is either the earth's rotation or the earth-moon separation....

The Earth 'gives' energy to the Moon, which causes it to move further away and the Earth's rotation to slow down.

How does this transfer occur?

The Moon's gravity causes the ocean to rise (a tide). Because the Earth rotates faster than the Moon, the bulge is ahead of the Moon, effectively pulling it ever so slightly to go faster.

If energy was extracted from the tide, the bulge would be smaller, the pull would be less and the Moon would stay put and we would not have to worry about a day being 1 second longer 50,000 years from now.

What is Trump's plan to stop the Moon robbing the Earth of it's rotational energy? Future generations are going to have to work more and more seconds per day!

We will stop the unfair tidal system by which Luna has robbed America of it's rotational energy for so long. We will restore the rate of spin America enjoyed 50 years ago; under Barack Obama we lost 3 or 4 milliseconds a year.

We will also build a wall. In space. And the moon-men will pay for it.

It was a joke, son.

Green's support tidal power because they want the moon to crash into the earth.

But Flash will stop them!

There is a literature on this topic. It's not a new idea. The costs are too high. A quick Wikipedia search would have helped answer your questions:

Anytime anyone with begins a plan with the word "just..."

But then, sometimes it's "just" an engineering problem.

There is at least one test project with reasonable (almost commercial) scale:

There has been a lot of research and they have come close to some workable schemes:

There are frequent claims of dishonest policy decisions with this one. Although I notice no one else has picked up the idea.

The other problem with wave power is that those waves tend to be used by the environment. Taking the energy of the waves may have some unknown but serious effect downstream.

Most people don't get that generating power takes a lot more than free energy - capex of the facility to extract the energy and opex to run it are more important in the long term cost. Wave power is hard because the energy is not particularly dense (so lots of capex per unit of power) and the environment is incredibly challenging for materials - moist air and salt water is just about the most aggressive possible corrosion environment that you can think off. So high opex as well.

Alas, there are only 500,000 movies. And how many distinct ones?

Asking wrong questions. Why underwater? That adds a lot of cost and risk (it ain't easy keeping the ocean out, always leakage, bilge pumps which can fail, etc). Much easier to just run pipes to and from the sea.

Suspect answer is that green regs prevent discharging warm water into the ocean. So to get around them, we're adding large upfront cost/impact to environment and a smaller increase while operating.

Reduced parts count. A huge part of a land base data center is the cooling apparatus.

Well, underwater, they can use natural convection currents to pump the "coolant" about; saves cost of pumps.

Rivers and lakes can overheat and kill fish stocks etc. in them as a result of industry using the water for cooling purposes. But for oceans -- we lack industries large enough.

This is fine.

There might be some interesting analysis of costs both financial and environmental. On the one hand something something that cold ocean will need less active cooling, on the other hand it might need more high grade steel?

On the other-other hand, the cold ocean has to be one of the worst environments on earth to place electronics. I wonder what the MTBF is?

I said "something submerged in" and voice heard something something."

Contact with water gives much better heat transfer than contact with air, even with fans. But do you need to submerge the whole building? Something on the coast with a lot of the building underwater but all critical components above sea level might be better.

From the pictures it looks like shipping container sized pods. Presumably a pod would be left alone until internal failure rate justified taking it offline.

Great title, and it is pretty cool to keep data centers under the sea.

I wonder if this will materially raise the water temperatures locally (or more exactly if it would if the data center was scaled up).

Trump and Kim should hold the next summit meeting there. If those two dictators can't agree on anything then we pull away all the scuba gear until they do. Incentives, am I right? I know how you Libertarians think.

I purposely didn't click on that link when the algorithms served it to me as "You'll never guess why Microsoft is sinking a datacenter under the sea?" I'm glad the answer was what anybody who knows anything about thermodynamics would have guessed.

Another way a server colocation company tried to save money on cooling was by stiffing its electricity and HVAC contractors. To the surprise of nobody (except, apparently, a nationally known consulting company that was keeping a 24-7-uptime client portal in the colo) the contractors recorded their mechanics liens and forced the company into bankruptcy.

Google's way was to buy Deepmind so they could have better machine learning algorithms for its thermostats. The tech scuttlebut is that they made back their money in a year. And had some impressive results in some board games to boot.

Comments for this post are closed