Posted on 12/31/2022 2:35:50 PM PST by SeekAndFind
Datacenters use a lot of power and despite our best efforts, a big chunk of that still comes from burning fossil fuels. But what if instead of relying on local utilities for power, these facilities generated their own – maybe using a relatively itty-bitty nuclear reactor?
In a recent report, Omdia analysts Alan Howard and Vladimir Galabov made the case that using small modular reactors (SMRs) to power large datacenters might not be as crazy as it sounds.
As the name suggests, SMRs are essentially just miniaturized reactors. Instead of a massive facility producing gigawatts or more of power, SMRs are designed to produce just a fraction of that. The International Atomic Energy Agency (IAEA) says that depending on the SMR in question, the reactors can produce anywhere from tens to hundreds of megawatts of electrical output.
These reactors are by no means a new concept. In fact, they've been powering US Navy vessels for the better part of a century without incident. The first was the USS Nautilus in 1955. Since then, nuclear power has been a mainstay of US Navy propulsion, and today, the US operates a fleet of 83 nuclear-powered ships.
However, it's only been more recently that nuclear startups have begun developing and, in some cases, deploying SMRs in a commercial setting. For instance, two Russian-built SMRs capable of 35MW each are at the heart of a floating power plant off the nation's Arctic coast.
So how many SMRs would it take to free a datacenter from the grid? The answer to that question depends on a couple factors, Howard tells The Register.
The hyperscalers and cloud providers don't like to talk about how much power their datacenters consume, he explains, adding that the megawatt rating often cited by colocation providers really reflects the upper limits of the facility's sellable capacity, not its actual power draw or the significant fraction of power required to cool them.
But for the sake of argument, let's say your datacenter campus, including the compute, thermal management, and ancillary systems consume about 125MW. Assuming each SMR produces 35MW, four reactors ought to do the trick.
While SMRs are more than capable of powering a datacenter, analysts say the typical 200,000 square foot facility probably isn't a good candidate for an on-site nuclear plant. Instead, Howard argues that SMRs are more appropriate for large datacenter campuses, particularly those located in power-constrained regions like Virginia or Ireland.
According to the report, the sweet spot for SMRs will likely be for facilities exceeding 100MW, though Howard suggests that smaller datacenters could also partner with local utilities to form a sort of co-op in which other power-hungry industrial plants – a steel mill, for example – could purchase excess capacity.
Microreactors – an even smaller take on SMRs – may be a viable option for smaller datacenters or as an alternative to battery or diesel generators commonly used as backup power in the event of an outage, he added.
Despite the Navy's success, for many, nuclear power still evokes images of the Three Mile Island incident, and the meltdowns at Chernobyl and Fukushima. These accidents have served to stigmatize nuclear power as a dangerous and unnecessary risk.
Howard and Galabov, however, argue that recent developments around SMR have resolved many of the design and safety challenges associated with older reactor designs. "SMRs are considerably smaller than the large power plant reactors most of us are familiar with. Thus, SMRs pose far less risk due to their scale, simple design, and inherent safety characteristics of the reactor," they wrote.
"The biggest challenge is going to be convincing people in industry and the people where these things are gonna [be deployed], that it is safe, and viable, and environmentally friendly," Howard added.
While SMRs might have a better track record than large, pressurized water reactors, there's still the lingering issue of nuclear waste. While nuclear energy may be cleaner than coal or natural gas, it isn't renewable. The reactors produce power using heat generated by the controlled fission of elements like uranium or thorium. The byproduct of these reactions is a concoction of radioactive waste that can take tens or even thousands of years to reach safe levels.
The good news is that, depending on how the SMRs are built, they may not need to be refueled all that often. According to Omdia, the reactors used in nuclear submarines only require refueling roughly every 10 years and newer designs could push that to 30 or even 40 years.
The bad news is research suggests that SMRs aren't nearly as clean as their larger siblings. A study published this summer found that SMRs produce 35x more waste compared to larger reactor designs.
For SMRs to gain mainstream adoption among datacenter operators, they must be economically viable. SMRs may be powerful enough to run a datacenter, but unless they can do so cheaper than using renewables and fossil fuels, it's going to be a tough sell.
With commercial SMRs still in their infancy, it's hard to say how much they might cost to operationalize. With that said, SMR startups like NuScale claim their reactors will have a levelized cost estimate (LCOE) of between $40/MWh to $65/MWh, when they reach commercial availability during the latter half of the decade.
LCOE refers to the estimated revenue required to build and operate a generator over its lifetime. And on the optimistic end of this equation, this would put NuScale's reactors within spitting distance of natural gas and onshore wind at a LCOE of roughly $37/MWh, according to the US Energy Information Agency [PDF]. However, solar fares a bit better at $33/MWh.
For SMR vendors, this comparison is only going to become more favorable with time. Over the next two decades, the Energy Information Agency expects the LCOE for wind and natural gas to steadily creep upwards, while solar is expected to hold steady.
Nuclear does, however, have a distinct advantage over solar or wind. Renewable energy isn't available in every market, and where it is, its efficiency depends heavily on Mother Nature's cooperation. If the sun is not shining or the wind is not blowing, they're not producing power.
While you might think SMR would face major regulatory hurdles, Howard and Galabov note that this is less of a headache than you might think. While the US has been slower than other nations, the Nuclear Regulatory Commission recently cleared the way for SMRs on domestic soil.
Despite this progress, it's going to be some time before commercial SMRs are available. "The most optimistic deployment of an SMR here in the United States is by 2030," Howard said. "The notion of it being used on a datacenter campus, that's going to be – and I'm only speculating here –between 10 and 15 years away."
However, the US is hardly the only country actively exploring SMR tech. According to the report, there are already several SMRs under construction or being licensed in Argentina, Canada, China, France and South Korea. But just like the US, many of these reactors are still the better part of a decade from deployment. ®
I think nuclear power would help our country greatly, as long as we avoid incidents like Chernobyl.
these are small scale reactors that probably cannot melt.
I can see them being used in some cases, and in some countries, and for some workloads.
for example would I use one of these in Redmond for a microsoft data center, no probably not energy is cheap in that area of washington and there are nimbys
would I use one of these to mine bitcoin in the middle of a desert or in an arctic area with no population,
or if I were Musky somewhere on Mars, oh yes you betcha.
Great, nuclear for data centers that will track and control our lives and make sure our bug protein vs actual meat ratio is correct. Generating power for us? Not so much.
Those nuclear reactors better not be too small.
Under the illegitimate pedo bidet regime, the software will be forced to be outsourced to the chicoms. What could go wrong?
So NOW we can have nuclear reactors AFTER the Commies are in charge.
You didn’t mention your backup diesel generators and their fuel storage tanks. All that expense would go away with SMRs.
Here is someone trying to make a point that could be quite convincing and then he puts in a word like itty-bitty. Then he decides to make it relatively itty-bitty. I think itty-bitty as a stupid term stands alone, but adding relatively to it removes any usefulness of cuteness.
In case y’all forgot:
If such reactors pop up in Scotland, will they be referred to as “wee reactors”?
𝘐 𝘵𝘩𝘪𝘯𝘬 𝘯𝘶𝘤𝘭𝘦𝘢𝘳 𝘱𝘰𝘸𝘦𝘳 𝘸𝘰𝘶𝘭𝘥 𝘩𝘦𝘭𝘱 𝘰𝘶𝘳 𝘤𝘰𝘶𝘯𝘵𝘳𝘺 𝘨𝘳𝘦𝘢𝘵𝘭𝘺, 𝘢𝘴 𝘭𝘰𝘯𝘨 𝘢𝘴 𝘸𝘦 𝘢𝘷𝘰𝘪𝘥 𝘪𝘯𝘤𝘪𝘥𝘦𝘯𝘵𝘴 𝘭𝘪𝘬𝘦 𝘊𝘩𝘦𝘳𝘯𝘰𝘣𝘺𝘭.
My understanding of the aftermath of Chernobyl was that the flaw in the RBMK reactors was worked out. Seeing how things are these days, bureaucrats, technocrats, politicians, and/or diversity hires will ensure we’d have another event on the scale of Chernobyl, or much, much worse.
They have been trumpeting thorium reactors for now close to twenty years. They should be operational by now.
I just wrote a dissertation in my law school class on the SMR. They’re a pretty neat concept. The big issue is finding sufficient people to work on the reactors. Simple truth is that there aren’t enough smart people to operate the SMR (300 MWhr reactors) and the Mini reactors. The navy pays millions upon millions of dollars in bonuses to attract and retain operators. Foreign countries are willing to pay millions of dollars to hire them after Nukes leave the navy.
Russian reactors used liquid graphite and were inherently unsafe. As they got hotter, they got more reactive. As they got more reactive, they got hotter.....
They didn’t work very well at the Antarctic research station. They reverted to diesel generators.
That is the trick, isn't it? Proper vigilance is never maintained.
I don't think graphite melts.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.