Google tries to greenwash massive AI energy consumption with another vague nuclear deal Google has signed a strategic agreement with nuclear project developer Elementl Power to support the early development of three potential fission reactor sites in the US. But with no selected reactor tech and no construction timeline, the announcement sounds more like a handwaving exercise to distract onlookers from the massive …
If you build solar/wind you have to actually do something, and you'll need batteries if it is for your own 24x7 use. Sign a deal for nuclear power and no one has to do anything since nuclear approvals move so slowly it is impossible for outside observers to tell the difference between "doing our best to make this happen" and "never intended for it to go beyond the press release stage".
I beg to differ:
Vasquez: All right, we got seven canisters of CN-20. I say we roll them in there and nerve gas the whole fucking nest.
Hicks: That's worth a try, but we don't know if it's gonna affect them.
Hudson: Let's just bug out and call it even, OK? What are we even talking about this for?
Ripley: I say we take off and nuke the entire site from orbit. It's the only way to be sure.1,2
________________
"as The Register pointed out recently, Google's nuclear plans - along with ... others - may be too little too late to address the growing concerns that there isn't enough power to fuel the growing demand from datacenters and AI."
Yes, there isn't enough power right now to meet the future rising demand, so we'll be adding more power, and some of it won't be clean. Nuclear isn't going to be able to help much for the next five years. But once we have modular manufacturing up and running, it could become one of the fastest growing clean energy sectors. The near-term delay doesn't mean it will be "too late" for anything--other than arbitrary near-term deadlines that don't mean anything and which we aren't going to meet anyway, no matter what we do.
"Another abstract agreement ..."
Aren't all agreements abstract?
"... through a company that has yet to build anything is unlikely to help now, when we really need it to,"
We'll still be needing it 5, 10, 15, and 20+ years from now. We can only go from where we are. As for the "yet to build anything" knock, that definitely does not apply to Kairos Power. Starting with nothing in 2016, they built a rapid-prototyping lab and a large assembly prototyping lab--and built and knocked down many prototypes, they built a salt production facility, they built and ran the world's largest molten salt test loop, they set up a regulatory affairs center and are currently setting up an operator training facility, they have started setting up their Triso fuel production line, they've started construction on their reactor factory, they've started construction on their Hermes 1 demo facility, and earlier this year, they completed their first in-house build of their commercial-scale reactor vessel--in a matter of months--without benefit of their planned mass-production capabilities. Considering they've also been having to push multiple simultaneous permit and license applications through a notoriously slow NRC while doing all this, I would rate their accomplishments and speed so far as quite impressive.
In a cold-start race between a bicycle and an SR-71 Blackbird, the cyclist will easily jump out into the lead right off the line, and in a few seconds the cyclist will be sprinting away at near top speed while the Blackbird is still sitting motionless. But just because the Bird hasn't moved doesn't mean nothing is happening, and it doesn't mean the Bird is "too slow". Once the engines are spooled up and making thrust, the race dynamics will change rapidly.
Remind me what the CO2 contributions of the bicycle are compared to the SR-71 Blackbird?
For that matter, how many passenger-miles did we get last year from Lockheed's (retired) strategic reconnaissance marvel? How does that compare to the humble bicycle?
Yes, I'm being deliberately obtuse. And I'm actually an advocate for responsible nuclear power. But the AI bollocks and it's attendant energy race do not and will not contribute to the sum of human happiness. And we would be better off if we accepted that sooner rather than later.
This post has been deleted by its author
"Remind me what the CO2 contributions of the bicycle are compared to the SR-71 Blackbird?"
The analogy was physical speed to production speed. But you knew that. I couldn't think of an analogy which covered both production speed and CO2 profile, but hotter reactors using some of their heat to power CO2 extraction from the air could have very large negative-CO2 profiles--way better than wind, solar, hydro, and bicycles. And we are going to need a lot of that if we are going to have reasonable hope of removing about a trillion tonnes of CO2 from the air fast enough to avoid catastrophic warming or ocean acidification collapse.
"...the AI bollocks and it's attendant energy race do not and will not contribute to the sum of human happiness."
The enormous sums of money connected to AI development could launch multiple kinds of next-gen reactors, and once launched, they could take off on their own. The clean energy potential is huge, the fuel is abundant, and it is the only energy source we know of capable of driving industrial-scale CO2 drawdown. It could help to shut down some toxic and deadly forms of pollution, offset highly destructive kinds of mining, could lift billions out of energy poverty, and could help shrink the human footprint on the natural world. And I think all of those factors, and more, have the potential to contribute to human health and happiness--even if the AI itself never does.
Replanting a large chunk (i.e. a trillion) of the roughly 3 trillion trees humanity has felled will remove about that much CO2 from the atmosphere, reduce local and possibly global temperatures (greenery has a local cooling effect), filter more pollutants from the air (trees act as air natural filters), and even increase oxygenation slighty (increased photosynthesis), for far less cost and on about the same time scale as building a load of nuclear reactors.
I am very pro-nuclear, but if your goal is removal of existing CO2, there are far easier and cheaper ways to do it. Of course we aren't allowed to consider simple and cheap solutions. We can only consider solutions that force massive lifestyle changes on the entire planet, at eyewatering cost.
A 2022 IPCC report estimated the potential for improved biological CO2 capture means (reforestation, wetland and peat bog restoration, changes in agriculture, biochar, etc.) and it projected a potential to sequester 860 million tonnes of CO2 per year worldwide by 2030, and as much as 4.19 Gt per year by 2100. And that would be great, but right now, we are making almost no progress towards implementation. And even if we accomplish it, it's not enough. Even if we were at zero CO2 emissions by 2100 for a net negative of 4.19 Gt per year, that rate of capture would take over 200 years to draw down a trillion tonnes (and our current excess is a trillion tonnes, so the excess could be considerably higher by 2100). We need something more like 20 - 30 net-negative Gt CO2 per year, and we need it a lot sooner than 2100.
We don't know at this point how much industrial CO2 drawdown can help, but there are some intriguing possibilities. We have figured out ways to pull CO2 out of the air using common and re-usable materials at an energy expenditure of around 1.5 to 2 MWh per tonne CO2--with most of that needed energy being in the form of heat, and most of the non-heat energy going to fans. Coincidentally, some of the advanced reactor developers are planning for smaller, hotter power plants which can be air-cooled--meaning they will be running fans to move a lot of air anyway, and will be throwing away a lot of reject heat--in a range that CO2 capture can use. Piggyback CO2 capture onto a 200MW(e) nextgen nuke, and on top of generating the electricity, it could also sequester around a million tonnes of CO2 per year with the reject heat. If we can get the cost of nukes down into an attractive range, we'll easily have enough energy demand to support building tens of thousands of nukes at that scale--potentially tens of gigatonnes of CO2 removal per year. This won't replace the need for the biological methods. We'll still need every scrap of those we can achieve, and every other CO2 capture method we can develop as well. But nuclear is in a unique position to help because it is, and will always be, the largest source of clean energy which produces its primary energy in the form of heat, rather than electricity.
But the AI bollocks and it's attendant energy race do not and will not contribute to the sum of human happiness. And we would be better off if we accepted that sooner rather than later.
But they might. So techbros spaff a few billion for say, 20GW power to feed AI bollocks. AI bubble inevitably bursts and there could still be 20GW power. When supply exceeds demand, prices should fall. Give or take the need to keep electrons shuffling at 50 or 60 beats per second. Downside is if electricity prices fall below operating costs, but in the US, that's what Ch.11 is for.
So (in theory) datacentres are a good match for someone looking to do a FOAK design.
On the upside Kiaros is low pressure/high temperature (as in near-conventional FPP temperature levels) power plant so can use COTS steam turbines available from half a dozen mfgs.
On the down side MSR have historically run high enrichment IE near-bomb-grade levels.
*Fun fact about boiler designs. While the pressure in the tubes is about the same as in a PWR or BWR the pressure in the boiler is more like 10atm. This makes FPP boilers much easier (or though not necessarily faster to construct) by welding sheet. OTOH all of the reactor pressure vessel has to operate around 175, hence needing the 200mm/8" thick forged shell (about 600tonnes of billet) that only about 6 forges in the world can make. Not a good design if you want to roll out by the 100, as you will need to if you want to make a dent in the climate crisis.
I missed out the unit of pressure for the PWR.
It is also atmospheres.
So that's 10atm inside the boiler (the "pressure shell" as boiler makers call it, with around 2500psi inside the tubes inside the boiler) Vs 174atm inside all of the PWR pressure vessel.
The difference in diameters is why the hydraulic tubing on a car is quite thin (and it's running at c3000psi) while the wall on a PWR is 200mm or 8" thick.
10atm by consumer measures is pretty high but by industrial practice it's actually very comfortable
BTW those wall thincknesses have a massive impact on the amount of steel you need, the size of the containment walls, the size of the foundations etc.
> The difference in diameters is why the hydraulic tubing on a car is quite thin (and it's running at c3000psi) while the wall on a PWR is 200mm or 8" thick.
For a given working pressure, wall thickness scales directly as tank diameter. This is the core concept behind "SCH" ratings on steam pipes.
> 10atm by consumer measures is pretty high but by industrial practice it's actually very comfortable
10 atm is 150 psi. I got 600 pounds of flammable gas next to my house, there's a 50 foot tank at the depot, all sized for 150psi on a hot summer day. If I had steam heat, that's what the pipes would be rated (home heat takes a huge margin of safety). To blast dirt and paint off my house I use a 150+psi pressure washer. When real work happens: the Big Boy steam locomotive runs 300psi (20atm), steam turbines have run much higher for many decades to gain efficiency.
"On the down side MSR have historically run high enrichment IE near-bomb-grade levels."
I think there is no historical precedent for the Kairos strategy of running Triso fuel in molten salt coolant. The Kairos fuel will be enriched to slightly under 20% U-235. Weapons-grade U-235 is 90%. (Some U.S. Navy reactors have used fuel enriched to 97% U-235, though now 93% is more common.) And once formed into Triso fuel balls, it would be very tough to dig out the tiny particles of fuel--a lot more work than just enriching natural uranium up to the same level. So far as I know, all of the other enriched-U molten salt developers are also planning on using below-20% enrichment.
The bomb-hazard molten salt reactors are the ones that will run on U-233 bred from thorium--the kind that China is developing. Base-purity without improvement is around 99.87% U-233. With 54 days of decay segregation, that can be reduced to one part contaminant per 1589 trillion parts U-233. (i.e. 99.99999999999994% pure U-233) Gun-detonator grade is anything below 1 part contaminant per million parts U-233. And with an unreflected critical mass of only 15 kg. (as opposed to 52 kg. for U-235), that kind of purity would make it possible--even easy-- to build a briefcase-sized, hand-carried U-233 nuke that uses a very simple gun detonator. And if we don't like the idea of China selling a bunch of those reactors to their friends, we'd better develop something which can outcompete it quickly. They've already got their thorium test reactor up and running.
Not really.
Crush the pellets, burn off the graphite, dissolve the alumina with one of several reagents and you'll have a fresh supply of uranium to go.
But IMHO the real killer is the enrichment higher enrichment --> higher cost. This stuff is roughly 4 1/3x the enrichment of PWR fuel and may 10-15x that of CANDU fuel.
Then of course there's the Beryllium which is not radioactive (to begin with) but is very toxic.
It's a pity because I liked both the MSR and the pebble bed reactor. This thing looks like someone said "ChatGPT come up with a new kind of high temperature low pressure reacto that hasn't been pattened yet."
IMHO ow pressure is good. High (USC FPP) temperature is good. High enrichment (actually any enrichment, given we know how to build natural U reactors, and have since the 1940's) is bad. High pressure is very bad.
Guess time will tell if this gets done or what.
"Crush the pellets, burn off the graphite..(etc)... and you'll have a fresh supply of uranium to go."
But it would still be a long ways from weapons-grade, so it would still need a lot of additional enrichment to make a bomb. But if you already have the enrichment equipment to do that, it's easier to just start from natural uranium than to deal with the Triso mess. Even easier is to get ultra-high purity bomb fuel from a thorium MSR.
"But IMHO the real killer is the enrichment ... higher enrichment --> higher cost."
It will cost more; I don't think that will be fatal. Early cost indicators look like around 35 - 40,000 $ per kg. U in Haleu Triso. And each kg. U could produce around 4250 MWh of heat, so that works out to around $9.40 per MWh of heat, or around $28 per MWh of electricity (or a bit less with improved thermal efficiency). That's not horrible compared to the fuel cost of combined-cycle natural gas or coal, at around $35 per MWh(e). Plus, with experience, Triso fabrication cost should come down. Plus there's the MOX option, replacing some of that enrichment with reactor-grade plutonium. Thorium blending is another possibility. And every Kairos reactor will be able to benefit from future fuel improvements.
On the flipside, there should be very large capital cost savings relative to old nuclear. And cheaper turbines relative to combined-cycle gas. The tough one to beat will be coal, but only because coal offloads its worst costs onto the environment and the public.
"Then of course there's the Beryllium ... very toxic."
Primarily in dust form--which won't be a thing for beryllium in molten salt. We use beryllium in many aerospace applications, so we have some experience in how to handle it.
"It's a pity because I liked both the MSR and the pebble bed reactor."
Both are good--in different ways. The hybrid creates a new category with its own features. Advantage over liquid-fuel MSR is it's not nearly as corrosive without the tellurium, so the reactor vessel can be stamped out of ordinary 316 stainless (and recycled after about a year out of service). And Triso fuel won't need any on-site chemical processing like liquid fuel will. Disadvantage is that it adds ball-handling equipment, and reactor control will have to be done by control rods. Advantage over gas-cooled pebble bed is easier ball handling (pulling floating balls off the top rather than from the bottom of a heavy stack), hot-spots within the stacks are eliminated, stack and shock loading of the balls is eliminated (much less risk of ball damage--which has been a problem), much easier insertion of control rods, and the salt aggressively grabs and retains any cesium, strontium or iodine which might escape from any damaged balls, adding another robust layer of containment. This would be a particular advantage for maritime reactors--where even a core breach directly into the ocean would not result in a significant release of contaminants. Gas-cooled pebble bed has the advantage for very high-temperature applications.
"High enrichment (actually any enrichment, given we know how to build natural U reactors, and have since the 1940's) is bad."
In moderated reactors, enrichment greatly reduces the amount of uranium that goes through the reactors, which greatly shrinks the amount of high-level waste (used fuel) produced. Molten salt fast reactors could solve the waste problem by utilizing close to 100% of their uranium fuel, but those are a new concept and we likely have another 10-15 years of development work yet to do on those. But that is the direction we'll need to go eventually.
"Guess time will tell if this gets done or what."
Advanced reactors will definitely get developed somewhere. China is hoping we'll drop the ball here, like we've done in so many other places where we could have led.
Funny you should say that.
there's a table of information supplied to aerospace engineering students on relative costs of different materials.
IIRC Be is 200x the cost of Al alloy.
Nothing says "Cost-plus govt contract" quite like using Be on a project.
The combination of melting point, brittleness* and high toxicity all contribute to this cost.
*Both Britain (who planned to use it for cladding on the AGR fuel) and France (who had a nuclear centre at Grenoble) solved the brittleness problem. Control of "Tramp" elements is critical. The French did IIRC a 5 pass zone refining run and came up with a bar they could literally tie in a knot. Pretty amazing IMHO.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
