Microsoft hiring for “small modular reactor” strategy to power data centers | Hacker News Reader

Microsoft hiring for “small modular reactor” strategy to power data centers

datacenterdynamics.com

37 pointssien27 comments

So, I believe this has been under research since ~2015 at Microsoft. Bill Gates had invested in nuclear power and introduced the idea.

It seems like a good idea. One problem that pops out is that even with a large, geographically distributed data center, you’re talking about ~250MW. In order to get the same level of resiliency the grid has you need to have some redundancy. That gets really complicated. If the unit fails, and you need to start a 125MW power draw from the grid. Which…isn’t practical.

When I was studying nuclear in 2010 Gates was employing my professor as technical advisor for standing wave reactors, so he’s been interested in nukes for a while. Modular reactors were cool theoretical ideas at the time similar to 3D printers, crazy to see how far they’ve come.

That's surprising considering that most modular reactors are just smaller versions of the ones we have now.

Of course, all the ratios are wrong is you scale everything and you might not get to criticality. So it is a completely different design with different manufacturing techniques for many parts. It is like a model of an airplane also wouldn't fly without adjustments to many parts which of course changes various ratios.

I would have thought that SMR design would be more inspired by nuclear submarines and such which are already small, not necessarily large power stations, but admittedly I don't know anything in this area.

Nuclear subs are a lot more exotic, you need to go from 0-60 in a few seconds, so they make fuel plates instead of rods so it really bumps up the neutron population quickly.

There’s no need to adjust criticality that rapidly at a generating station.

Also, they tend to use much higher enrichment ratios on the fuel from what I have read. I guess that is all for the fast power output regulation as you write. Stunning technology and a security nightmare, just imagine a nuclear sub bumping into a click in territorial waters of a not-so-friendly nation and making the area uninhabitable.

I'm aware of that, but to get to normal sized reactors they first built smaller prototypes that had outputs that were much closer to the modular reactors that are now being investigated. So they're not really theoretical.

How are they different from the reactors in US Navy submarines?

That's a great use case for small modular reactors, especially once they're actually serialized and being delivered on time/on budget. I'm glad MS is interested in helping move the needle in that direction. The first few of any model will be slightly ugly.

No it isn't a great use case. It is a pandora's box. Mass manufacturing of nuclear devices used at sites with fewer or no top-notch experts, less stability because of the smaller volume, much more pressure on costs (because some things don't scale with the power output) and many, many more units where something can go wrong. Microsoft isn't even able to secure their Azure cloud or their products. Why do you think they will do so much better with nuclear energy? Will they automatically insure the whole region against an incident? Will they automatically pay all the damages such a case? How about the things that money cant buy such as stillborn babies, early death, mutations, the effects of extreme stress during a catastrophe? And if they contain an incident, will they pay the full clean up and permanent storage of the mess? Also, nuclear reactors tend to have a long life-time. What will happen at the end or should Microsoft as a company go out of business?

I really think we need to be a lot more humble with nuclear energy. The last cost savings caused Chernobyl to be such a disaster. Even the Japanese couldn't handle the power of nature, which was to some degree foreseeable.

Here is the job :

https://jobs.careers.microsoft.com/global/en/job/1627555/Pri...

This is a nonstarter from an insurance perspective. (Ask me how I know.)

For similar pricing, they could deploy renewables and MES or PHS without the risks associated with future radioactive waste transportation and storage and potential for meltdowns. Renewables already meet over 1/3 of world energy needs. It's rapidly becoming a waste of money and a political hot-button to invest in "bridge" energy sources when long-term solutions are viable here and now. It's good news that economic and ethical interests are rapidly aligning.

Now you've got me wondering, what kind of technical expertise do different insurance companies possess to understand risk in these niche industries? Insuring Viasat's Visasat-3 for example. How do you quote for something like that without just pulling a number out of thin air and hope it doesn't bankrupt them. Do insurers need to sign off on technical specs?

is nuclear a bridge? Renewables are inconsistent and batteries probably won't be good enough

Then you failed to understand pumped or mechanical energy storage.

Based on my limited current understanding, pumped/mechanical storage is not going to get us to 100% solar/wind. Pumped storage is great, but we need some baseline generation

Strawman: you're conflating and confusing energy storage with generation. They're entirely different. Texas alone deployed multiple multiples of 100's MW solar generation facilities that kept the lights on during the hottest heatwave in a century. Renewables account for 1/3 of all energy produced in the world now. Decarbonization of most major centralized pollution sources is achievable within 15 years.

And wherever there is a hydroelectric generation dam, there is also a "battery" that can run generators in reverse for PES, e.g., it is both generation and storage.

I am not conflating storage with generation. I am saying that mechanical storage is insufficient to fill the dips in renewable generation.

Neither is really efficient

You don't know what you're talking about and provided no evidence. Pumped energy storage goes up to 80%+ efficient and is extremely scalable.

It isn’t extremely scalable. It only works in pretty specific locations.

No large industrialized without large hydro resources manages to run on NREs + storage only. France does with nuclear + hydro.

You can compare the energy mix of the countries with accessible data here https://app.electricitymaps.com/map

Even in larger time periods (a year or so), if a country is emotting low co2 (green) then it is either nuclear plus VREs (variable renewable energies) or VRE plus hydro

1/3rd of current energy needs. Come back when you can do 100x that on renewable or more. We aren't going to be able to get by with just current outputs. Several orders of magnitude greater are needed.

Huh. Is that economical because data centers are such dense power sinks that taking power from the grid usually means building new plants or substations anyway?

This is a truly great use case for nuclear power. Fine idea.

Reply on news.ycombinator.com