Microsoft exploring options to power its datcenters using nuclear energy

Marees

2[H]4U
Joined
Sep 28, 2018
Messages
2,095
Microsoft looking for a Principal Program Manager, Nuclear Technology, who will be responsible for maturing and implementing a global Small Modular Reactor (SMR) and microreactor energy strategy.

This senior position is tasked with leading the technical assessment for the integration of SMR and microreactors to power the datacenters that the Microsoft Cloud and AI reside on.

https://jobs.careers.microsoft.com/.../Principal-Program-Manager-Nuclear-Technology

The new hire would join the energy innovation team at Microsoft, working with P. Todd Noe, director of nuclear technologies engineering at Microsoft. Noe said on LinkedIn: "This is not just a job, it is a challenge. By joining us, you will be part of a global movement that is transforming the way we produce and consume energy. You will also have the chance to grow your skills, advance your career, and make an impact on millions of lives."



A Future Shaped by Nuclear-Powered Tech Companies?​

Microsoft’s move into nuclear technology suggests a future where tech companies with significant cloud businesses may need to become nuclear power developers to sustain their growth. The race is on to replace traditional power sources with microreactors reliable enough to keep data centers secure in case of public grid failure.

https://www.bigtechwire.com/2023/09...to-slake-ai-related-datacenter-energy-thirst/



Analysists at tech research firm Omdia argue that small modular reactors (SMRs) could become prevalent in future years, replacing the need for data centers to draw power from the grid with their very own environmentally-friendly alternative.

Such reactors have been used in certain industries for many years, such as powering submarines in the US Navy, which has over 80 ships employing the technology.

Speaking to The Register, one of the co-authors of the report, Alan Howard, said that getting an accurate figure of the power consumption of the data centers used by cloud storage providers is tricky, as they often don't provide such figures easily or accurately.

A typical large scale data center may use 125MW of power, meaning four SMRs would be needed if they produce a typical 35MW each.

One problem, however, is the amount space needed to accommodate SMRs, typically requiring around 200,000 square feet. This means they would only really be viable in the largest data center sites. For this reason, the report recommends their use for colocations that require upwards of 100MW.

However, Howard also suggests that smaller centers could partner with other local industries to make use of surplus power. Additionally, he also claimed that even smaller reactors, known as microreactors, could be used for backup power in data centers, as a replacement for batteries and diesel generators used currently.

https://www.techradar.com/news/mini-nuclear-reactors-could-soon-power-data-centers


Rolls-Royce has begun pitching 470MW modular power plants to data centers, with a planned roll out of 2030, while Last Energy has already found customers in the UK for 20MW SMRs. Rival NuScale received regulatory approval for 50-77MW SMRs in the US this year, but it has struggled to keep its electricity costs in check. Sam Altman-backed Oklo is also planning 15MW+ SMRs, while Microsoft cofounder Bill Gates has backed TerraPower.

Data center operators are looking to SMRs as a potential solution to power constraints, with Green Energy Partners planning to build multiple small modular nuclear reactors next to the 1.6GW Surry Nuclear Power Plant to support 30 new data centers in Virginia.

Swedish nuclear company Kärnfull Next has announced plans for a campus of small modular reactors (SMRs) on the Swedish coast to build data centers.

https://www.datacenterdynamics.com/...-microreactor-strategy-to-power-data-centers/
 
I mean if Kodak was allowed to operate one, then why not Microsoft?
The new Gen 5's should be available for 2029, if they get started on the paperwork now they might have regulatory approval by then.

The only catch is they will need to pay for the entire project upfront and have a registered account that has an additional 30% of that cost in escrow to deal with cost overruns, and if that fund is depleted they will need to stop and have a review done, and in the event that review takes longer than 4 years or is denied then the project is dead.
So Microsoft better get started on that lobbying, because the draconian construction laws regarding Nuclear Reactors and the state of the inspection offices for them are old and lacking to put it politely.

Worldwide GE has already started on 80 of them to be online for 2038 in 29 Countries, of which the USA is not one... Which is funny because they are designed in Tennessee.
 
I feel the fact that they are talking about Fusion, 100 million degree plasma and the prospect of 1cent for 1 kwh for this decade is not creating the amount of reaction expected, maybe we just got numb to the extreme claim of the sorts to not deliver
 
I feel the fact that they are talking about Fusion, 100 million degree plasma and the prospect of 1cent for 1 kwh for this decade is not creating the amount of reaction expected, maybe we just got numb to the extreme claim of the sorts to not deliver
That and even if they started the regulatory paperwork last year they are looking at a 2028 ground breaking with a 2035+ date of actually being online.

Power plants take a LONG ass time to get built, especially experimental ones.
 
Power plants take a LONG ass time to get built, especially experimental ones.
Ver small 50 MW for private use I imagine could get faster, the way it is phrased:
The commercial project will largely be licensed and regulated by the state Department of Health, which previously approved three of Helion’s test devices. The U.S. Nuclear Regulatory Commission delegated authority to the state for radiation safety oversight decades ago. The process is expected to be faster and more straightforward than what’s required for traditional nuclear plants powered by fission, or the splitting of radioactive elements, which poses more serious health and safety risks than fusion.

The very short date include the regulatory authorization. But unlike radioactive type of fuel affair, fueled by some helium type element in something new in a era much more open to nuclear than the recent past, supported by the Microsoft of the world with support of the department of energy could go fast, I am including that aspect to the extreme claim, 2028 is so fast.
 
Ver small 50 MW for private use I imagine could get faster, the way it is phrased:
The commercial project will largely be licensed and regulated by the state Department of Health, which previously approved three of Helion’s test devices. The U.S. Nuclear Regulatory Commission delegated authority to the state for radiation safety oversight decades ago. The process is expected to be faster and more straightforward than what’s required for traditional nuclear plants powered by fission, or the splitting of radioactive elements, which poses more serious health and safety risks than fusion.

The very short date include the regulatory authorization. But unlike radioactive type of fuel affair, fueled by some helium type element in something new in a era much more open to nuclear than the recent past, supported by the Microsoft of the world with support of the department of energy could go fast, I am including that aspect to the extreme claim, 2028 is so fast.
The problem with helium fusion is the extremely high temperature required to achieve fusion. Deuterium can be used to reduce the temperature needed, but that reintroduces the possibility of radioactive byproducts.
 
Power plants take a LONG ass time to get built, especially experimental ones.
According to the All-inpodcast above, fusion will not be regulated like fission power plant at all but more like particular accelerator and nuclear machine in a hospital cancer treatment centre (at least in some state), going from a 1 billion, many decades regulatory desert walks to your more entry level 10 millions 6-9 months burden.
 
Last edited:
The only catch is they will need to pay for the entire project upfront and have a registered account that has an additional 30% of that cost in escrow to deal with cost overruns, and if that fund is depleted they will need to stop and have a review done, and in the event that review takes longer than 4 years or is denied then the project is dead.
So Microsoft better get started on that lobbying, because the draconian construction laws regarding Nuclear Reactors and the state of the inspection offices for them are old and lacking to put it politely.
MS can afford that kind of scratch, they have MASSIVE cash reserves.
 
kl;gkdfl;gsdf.gif
 
Just FYI for all the "OMG nuclear BSOD MS so unstable folks": Nuclear reactors DO run Windows. If you have something that uses SCADA/ICS stuff, it runs on Windows. Old, unpatched Windows, because the companies that make it are basically just of the "deal with it" mentality and because shit is expensive and you don't replace it often. So they end up running on (hopefully well) isolated networks.

So ya that power plant, nuclear or not? Probably run by a bunch of Windows 7 computers.
 
Windows 7 ? I BET that we have WinXP controlling some kind of instrument or sensor.

Anyway, small sized reactors are the way to go (just imagine a autonomous independent aluminium factory or chip fab !)
 
A small reactor like one from NuScale, for example, can provide 154 MWe for 12 years without needing to be refueled. Its biggest microreactor, the Voygr-12, requires just 0.05 square miles of space compared to 94 square miles for wind and 17 square miles for solar, according to the company.

https://www.extremetech.com/energy/...se-nuclear-reactors-to-power-its-data-centers
0.05 square miles is actually a pretty large footprint for something dubbed a "micro reactor" That's 1180 feet by 1180 feet.
 
0.05 square miles is actually a pretty large footprint for something dubbed a "micro reactor" That's 1180 feet by 1180 feet.
Given the average size of a reactor is about 1 square mile it’s pretty small in comparison.
And the average reactor is already something like 20x smaller than most other forms of power generation so really 0.05 square miles is downright microscopic in comparison.
 
They are quite nice, Rio TInto Alcan have some in Quebec, private hydro-power:

https://fr.wikipedia.org/wiki/Centrale_de_L'Isle-Maligne

They sales electricity when the shop is down.

Windows XP could be one of the most popular yes, not so long ago they still had them on connected to the exterior system.
I’ve helped troubleshoot a few of those RioTinto-Alcan Hydro facilities and Smelters.

Last one I was in was pretty much all off the shelf Linux based micro controllers talking back to big logging databases monitored through web interfaces.

Windows machines for management but mostly because Linux and Apple don’t have the MDM capabilities or didn’t at the time. But aside from the need for MSOffice there wasn’t much of an operational technical reason for it. All the terminals I worked with were all PXE boot.
 
Related:

Amazon ties up with nuclear energy provider for data centre power needs


First announced by DCD in July 2021, the 1,200-acre campus draws power from Talen Energy’s neighboring 2.5GW nuclear power station in Luzerne County, the Susquehanna Steam Electric Station (SSES). The company broke ground in 2021 and completed the first 48MW, 300,000 square foot (28,870 sqm) hyperscale facility early last year, along with a separate cryptomine facility.

Talen said AWS aims to develop a 960MW data center campus. The cloud company has minimum contractual power commitments that ramp up in 120MW increments over several years; AWS has a one-time option to cap commitments at 480MW. The cloud provider also has two 10-year extension options, tied to nuclear license renewals

As part of the deal, Talen will also supply AWS with energy via a 10-year Power Purchase Agreement (PPA) from the Susquehanna site.



https://www.datacenterdynamics.com/...s-nuclear-data-center-campus-in-pennsylvania/
 
Just FYI for all the "OMG nuclear BSOD MS so unstable folks": Nuclear reactors DO run Windows. If you have something that uses SCADA/ICS stuff, it runs on Windows. Old, unpatched Windows, because the companies that make it are basically just of the "deal with it" mentality and because shit is expensive and you don't replace it often. So they end up running on (hopefully well) isolated networks.

So ya that power plant, nuclear or not? Probably run by a bunch of Windows 7 computers.
Would love to know where you got this from.

Reactor control computers are purpose built, think like the guidance computer they used to send astronauts to the moon. They're completely disconnected from any network. Hell, the plant I work at the control computers have a whopping 16mb of RAM. And that's upgraded from 8. More computing power is just unnecessary.

Keeping I'm mind that most PWR and BWR plants are very old...control of reactivity is accomplished by withdrawing control rods as the fuel is used up. They're withdrawn completely MANUALLY. the only purpose the computer serves is to monitor, generate some alarms (far from all), and control governor valves to the turbine.

There are newer designs that use touch screens for some things. Can't speak to those as they aren't really built yet...or there are very few....but again singular purpose. Generally we don't use windows for something like that, it doesn't make sense.
 
Would love to know where you got this from.
From having seen SCADA controls. There's only a couple of companies that make them, and they suck at security.

If you want a government report about SCADA in nuclear power in relation to security: Here you go.
 
From having seen SCADA controls. There's only a couple of companies that make them, and they suck at security.

If you want a government report about SCADA in nuclear power in relation to security: Here you go.
This is not typical here. We use no such systems and I'm a little confused as to why one would want to ...

Virtually everything at our plants here is electromechanical. analog gauges, mechanical hand switches, etc. but the chart recorders are digital these days.... mostly because we couldn't get parts for the old pen and paper ones.

Good to know, I did not know this was a thing in nuclear. Thanks.
 
SCADA is pretty terrible, I certainly wouldn't want to rely on it, but I suppose it could be made more robust. I mean it's used offshore, for better or worse.
 
SCADA is pretty terrible, I certainly wouldn't want to rely on it, but I suppose it could be made more robust. I mean it's used offshore, for better or worse.
I don't know how good a protocol it is, but but if memory serves, it's unencrypted and it's trivial to break into or interfere with wireless SCADA .
 
Stolen from FB, don't know original source:
View attachment 639705

On topic: considering the power draw of modern datacenters, this is not terribly surprising. I wonder if any of the SMR designs are pebble-beds.
Yes some are pebble beds. I think China was the first to build and start up a pebble bed. I admit I don't know much more about it than that.

There's a lot of SMR designs out there. Weather or not they see the light of day will be another matter. Several molten salt ones that look promising, molten lead is another, and even a BWR based SMR exists.
 
I love seeing the first SMRs being built, anything is better than the world's economies flushing 19 billion dollars a day in subsidies down the oil and gas toilet.
 
I love seeing the first SMRs being built, anything is better than the world's economies flushing 19 billion dollars a day in subsidies down the oil and gas toilet.
I agree and I doubt that's true. Anything is better than flushing billions a dollar a day on inefficient wind/solar. We would have so much more money to devote to developing fusion or improving wind/solar efficiency, than putting the cart before the horse.
 
I agree and I doubt that's true. Anything is better than flushing billions a dollar a day on inefficient wind/solar. We would have so much more money to devote to developing fusion or improving wind/solar efficiency, than putting the cart before the horse.
According to the world's various governments own data compiled by the IMF it's $7,000,000,000,000 a year to subsidize oil and gas. It's insane, I read that and spent hours reading more, but as far as I can tell that's the actual number. That's corruption, baby, nobody's slicker than an oil man. Meanwhile we've let fossil fuel and eviro-psycho lobbyists convince government to try to regulate fission out of existence.

The death toll from Fukushima was one. Meanwhile there's a specific ratio of how many years your life is statistically shortened by how close you live to a coal power plant.

Oil does do more than just energy though, so the numbers aren't 1:1 with other forms of energy.
 
IMF it's $7,000,000,000,000 a year to subsidize oil and gas.
At the same time it is one of the most taxed items (with tobaco and some others) that exist on earth, from the land use to the extra municipal tax at the pomp... which is a strange dynamic. Extra tax on the land at the refinery at every step and yet subsidie at the other end.

FVtg9BlVIAEDsb3.jpg
 
It should be telling that people who want to pay for electricity opt for nuclear power while everyone else should use big fans and solar panels.
 
At the same time it is one of the most taxed items (with tobaco and some others) that exist on earth, from the land use to the extra municipal tax at the pomp... which is a strange dynamic. Extra tax on the land at the refinery at every step and yet subsidie at the other end.

View attachment 640555
You pay the taxes, the tax money goes to the oil company.

Isn't corruption great?
 
Back
Top