AWS says Britain needs more nuclear power to feed AI datacenter surge

The UK needs more nuclear energy generation just to power all the AI datacenters that are going to be built, according to the head of Amazon Web Services (AWS). In an interview with the BBC, AWS chief executive Matt Garman said the world is going to have to build new technologies to cope with the projected energy demands of …

[Ken G]

Yes, but then again, No.

Nuclear power is a useful contributor to carbon neutral power generation. Europe (and it's semi-detached neighbour, the UK) would benefit from more of that.

Kicking it off to serve the current AI bubble is pointless since there's a 99% chance* that will have burst before the first reactor is designed, let along got through planning approval.

* with a 0.99% chance that AI will work as intended and provide us with smarter and smarter machines designing portable fusion generators and a 0.01% chance that it will work not as intended and make us it's duracell bunny bitches ala The Matrix

[Wellyboot]

Re: Yes, but then again, No.

Given that we're already importing near continuously 3-6GW of leccy, a couple of extra full fat reactor sites won't be short of customers regardless of the bit barns existing or not.

[Anonymous Coward]

Re: Yes, but then again, No.

I'm not disagreeing with you, but why are we also exporting electricity when we are in a deficit?

<quote>

n 2018, the annual production of electricity in Wales was 30.2 TWh and consumed 14.9 TWh, which means that Wales generates twice as much electricity as it consumes and is a net exporter of electricity to England, Ireland and Europe.</quote> - https://www.gov.wales/sites/default/files/publications/2019-10/energy-generation-in-wales-2018.pdf

[munnoch]

Re: Yes, but then again, No.

Mostly due to grid constraints. If we can't move it within our own country to where its needed we might be able to sell it to Europe via the interconnects. Its not often and pretty piddling amounts compared to what we regularly import. Ireland does take a pretty steady supply from the UK however. How many of those electrons speak with a Welsh accent is difficult to measure.

An interesting case is when our windmills are going full tilt and the wholesale price craters the Norway interconnect consumes from us to refill their reservoirs for free which they can then sell back to us later. Most days it runs full tilt towards us. Must be a nice little earner...

https://grid.iamkate.com/

[Wellyboot]

Re: Yes, but then again, No.

We could dam up a few Scottish glens and do the same trick, that'd eventually save a bundle.

[munnoch]

Re: Yes, but then again, No.

Unfortunately, like just about everything in this country, the geography is a bit crap and bland.

There were lots of hydro schemes built in Scotland in the mid 20th century but they are all really quite small in today's terms. There are a couple of larger schemes under construction but by comparison the amount of hydro storage in Norway is enormously more. Hence why they can afford to send us a near-constant 1.4GW almost 450 miles across the North Sea.

Something to do with the crinkly bits...

[Wellyboot]

Re: Yes, but then again, No.

Slartybardfast really did them a good turn.

[Anonymous Coward]

Re: Yes, but then again, No.

Could Holyrood be considered a glen?

[Alan Brown]

Re: Yes, but then again, No.

"Ireland does take a pretty steady supply from the UK however"

Other way around. Those Irish windfarms are usually exporting energy to the UK and will continue to do so until the Celtic Connector projects provide a link to France

[Anonymous Coward]

Re: Yes, but then again, No.

We get all our water and power from Wales, that's your job. Like it or not Wales is part of the UK.

[LionelB]

Re: Yes, but then again, No.

> Kicking it off to serve the current AI bubble is pointless since there's a 99% chance* that will have burst before the first reactor is designed

The truly scary prospect is that the bubble may not burst.

The current business model is to foist AI on you - it is rapidly becoming harder to opt out - and to fund and monetise it by slurping your data and using it to hurl advertising back in your face. Who's to say that this is not, in fact, a sustainable model?

> with a 0.99% chance that AI will work as intended

Perhaps it already is - it's just that "work" and "intended" do not mean what you thought they did (see above).

[Alan Brown]

Re: Yes, but then again, No.

"The truly scary prospect is that the bubble may not burst."

A human brain does vastly more with 75W than AI can do with a couple of gigawatts

Some efficiency drives would help a lot even if the bubble continues

[Anonymous Coward]

Re: Yes, but then again, No.

Maybe that's why we are soft and squidgy biological organisms. Perhaps it is impossible to usefully achieve GAI any other way. Evolution and a long time are quite efficient.

[cyberdemon]

Re: Yes, but then again, No.

> Kicking it off to serve the current AI bubble is pointless since there's a 99% chance* that will have burst before the first reactor is designed, let along got through planning approval.

Shhh. Don't tell them.

We'll build the nukes, the AI bubble pops, and then finally maybe we'll have cheap leccy

[Andy The Hat]

Re: Yes, but then again, No.

But the electricity price will still be artificially tied to oil price and the Government refuses to remove that tie even though green energy costs production are dropping.

As "net zero" plans and electricity usage is enforced, oil usage will drop, it's barrel cost will naturally balloon because of production efficiency losses and, surprise surprise, the cost of electricity *that everyone will now be forced to buy* will go through the roof.

Maximum profit for a few electricity moguls ... I wonder how many politicians will get directorships and fat brown envelopes?

[Anonymous Coward]

Re: Yes, but then again, No.

Oil use will not fall significantly unless a better or cheaper alternative emerges. Many large nations are not as dumb as us. Electricity has to be real cheap and battery technology much better, perhaps one day but right now it is bleeding us of money and moving it eastwards. After a little skimming on the way.

[Alan Brown]

Re: Yes, but then again, No.

Perhaps not "cheap", but I'll settle for "not having rolling blackouts in the 2030-40s"

[Anonymous Coward]

Re: Yes, but then again, No.

If you want NetZero you are going to have a disaster one way or another. Unless there is a massive shift to nuclear and the nirvana of fusion arrives or something we haven't thought of. More could be done with solar but not on farmland we need food security and solar only works in summer and as they are trying to dim the sun with chemicals that might well be causing problems with insects and plants it is a major problem.

[Anonymous Coward]

Re: Yes, but then again, No.

You can never have enough energy and diverse energy if you want to move up the Kardashev rating.

[firstnamebunchofnumbers]

Alternatively..

... just less AI?

[munnoch]

Re: Alternatively..

But then how will we be a powerhouse???

... whatever the fuck that even means...

[Lon24]

Re: Alternatively..

Alternatively if AI can teach us how to deliver a nuclear powerhouse on time and on budget it may just have a future ... Just sayin'

EDIT: AI appears to have taken over administration of this forum or the SysAdmin has had one too many at lunchtime and shared their instability .. hence double posts, missing pages ....

[Alan Brown]

Re: Alternatively..

"if AI can teach us how to deliver a nuclear powerhouse on time and on budget it may just have a future"

Or we could just pay attention to what's happening in the Gobi desert at Wu Wei (SINAP TMSR-LF1)

Our very own hack - the late lamented Lester - was a strong advocate of this technology. It's a pity he didn't live long enough to see the revived prototype being built (let alone going critical)

It's also a pity that the USA continues to operate under laws which make reviving their own MSR program illegal, thanks to Richard Milhaus Nixon in 1972, under pressure from a military which was afraid that divorcing civil nuclear power from its complete dependency on the unwanted waste products(*) of weaponsmaking would expose that weaponsmaking program's costs and also expose uranium separation facilities to nuclear limitation treaties

(*) Enriched uranium was used for the first reactor because it was known it would work and because there were tens of tons of the stuff that uranium separation plants wanted to get rid of. Depleted uranium (9kg for every 1kg of 3% enriched uranium) is the feedstock for making weapons-grade plutonium and nobody's bothered with an enriched uranium bomb since the 1950s because they cost so much that it's cheaper to BUY your enemy

[Anonymous Coward]

Re: Alternatively..

Unless we get leaders that are i) more interested in serving the nation's people and ii) able to separate chaff from wheat, we are not going be any sort of powerhouse.

[ecofeco]

Re: Alternatively..

Less AI? But we must close the AI gap with the commies!

Apologies to Dr. Strangelove. ------------------------------>>>>>>>>>>>

[Sam not the Viking]

Re: Alternatively..

Perhaps we could ask AI how to use less energy?

It does seem perverse that as energy-intensive, productive, industries fade away, low-energy "solutions" start to consume even more.

[Adair]

AWS is welcome to start spending on nuclear build

... just not in my backyard, or anywhere else on the planet.

Nuclear may be a very pleasant and interesting technical exercise (and even eventually delivers—up to a point), but as a socially responsible exercise in affordable financing and practical sustainability: FOAD.

I listened to the interview on this morning's 'Today' and the guy came across as just another self-absorbed money-grubber, i.e. "You (the people) take all the risk, and we'll take all the money. Cheers".

[Roland6]

Re: AWS is welcome to start spending on nuclear build

Can enforce this today: just include the building of a nuclear power station as a Sectin 106 requirement that must be delivered before construction can start on the data centre…

[Alan Brown]

Re: AWS is welcome to start spending on nuclear build

Nuclear suffers from a number of issues

- water cored designs aren't hot enough to be economic on the non-nuclear (power generation) side as they produce wet steam which damages turbines AND they're essentially giant radioactive steam boilers with all the engineering stresses that come with large boilers (This is what drives the long build times and high costs. Boiler explosions are a serious risk) AND high pressure/temperature water is _extremely_ corrosive to the pipework

- solid fuel designs produce high(ish) levels of waste regardless of the coolant/moderator (and can be misused to make weapons grade plutonium or are simply modifications of designs originally intended to produce plutonium)

- enriched fuel designs are a figleaf to cover up for making depleted uranium in a weapons program

- other coolants have various issues (Eg: Santa Suzannah)

Alvin Weinberg built the original Nautilus design as a proof of concept. He was unhappy with it being scaled up to Rube-Goldberg sizes and built a better mousetrap - MSRE - only to get kicked out of the Nuclear industry

None of the western companies pushing molten salt designs have yet to build a working one in the 25 years since the MSRE notes were rediscovered. China has had one working for nearly 2 years with (so far) good results and the indications are that scalability is as straightforward as Weinberg predicted in 1968, as well as having proven that after kickstarting, raw thorium can be used as fuel without needing a Th232-U233 process outside the reactor loop.

Global power and influence has long been closely related to control of access to resources (including energy). In the 2030s China's likely to become the World Hyperpower by way of selling Molten salt designs to power 5+ billion people and weaning them off carbon in the process.

[Anonymous Coward]

Re: AWS is welcome to start spending on nuclear build

There no reason to turn away from nuclear especially SMRs because it might be dangerous, the point is to make it safe. Bit like EV batteries, just because occasionally they burst into flames doesn't mean we should stop. If we behaved like that we would never advance. Imagine when the first cars arrived carrying a tank full of highly flammable liquid. Having once pulled someone out of an upside down wrecked car with fuel spilling out, they're still dangerous. All our advances have come with risk. The job is to localise and minimise.

[ecofeco]

Well see...

.. .it goes to eleven, right?

[jake]

Shirley one should turn that around ...

Britain demands AWS prove that AI datacenter surge will provide enough benefit to justify more spending on nuclear power.

_{I mean, have YOU ever seen a real, old-fashioned business plan out of any AI company?}

[Steve Davies 3]

If AWS want Nuclear power then

for their AI DC, they can pay for it. Otherwise FSCK OFF.

AI is a bubble that will burst in 12-18 months. People will get fed up with the shite coming out of the LLM's

Garbage In, Garbage Out.

[Doctor Syntax]

Re: If AWS want Nuclear power then

I think LLMs have created a new version: Good stuff in, Garbage out.

AWS are welcome to pay for the generators provided we get to keep them.

[Anonymous Coward]

Re: Shirley one should turn that around ...

It's not up to AWS to prove what's safe. It's up to the British authorities to deteremine what is and what is in the UK's best interest. Unfortunately, I don't trust our government.

[codejunky]

Hmm

Almost as though the money that has been thrown into energy generation should have gone into... energy generation. We could have had reliable and affordable energy. But with econuts like Clegg and seemingly most people opposing nuclear and the green madness opposing energy generation in general we can be proud of our high energy costs and being close to blackouts at winter.

[Excused Boots]

Re: Hmm

"and seemingly most people opposing nuclear and the green madness opposing energy generation in general"

Actually I don't think that is necessarily the case, I think (maybe naively perhaps) that the majority of people are quite intelligent, albeit ignorant (that isn't the same thing) about the subject. Which is fine, that's where explanations come in.

Now the extreme 'green' position is that any sort of fossil fuel or nuclear is 'bad' and will result in the death of millions. How about counter with, 'OK fine, your mother is in hospital undergoing an operation and it's night, no solar, the wind decides not to blow, that night, the tidal barrage is suffering from maintenance issues, your mother dies on the table because the power goes out - you OK with that'

It's hideous, horrible to think about no? But maybe, just maybe, that's what people need to be confronted with and made to think about?

So what do we want to do? Burning fossil fuels, not a good idea? Renewables, yes OK sometimes expensive, hydroelectric often needs flooded valleys, relocation of people, a certain geography, not applicable everywhere; solar, uses a considerable ground area, not to mention the obvious problem with nighttime; wind, generally good, offshore is very expensive and onshore tends to not always work when you need it.

And that leaves us with what?

Now I'm a bit of an optimist, but there have been quite substantial advances in fusion research in the last few years; I suspect that a demonstration commercial-grade working fusion power plant is 20-25 years away (probably in China), give it another 30 years to replicate that technology round the world, and we'll be fine.

But in the meantime....?

[Adair]

Re: Hmm

Nuclear fission (and fusion at present) is basically a gigantic moneysink, with a lethally toxic legacy to hand on to many generations to come.

Fossil fuels aren't much different in terms of consequences.

Which really leaves us with learning to make the best of energy sources that really don't stuff things up for people (and creatures) down the line who had no say in the selfishness, ignorance, and stupidity of their ancestors.

[Boris the Cockroach]

Re: Hmm

Quote

"Fossil fuels aren't much different in terms of consequences."

I'll add another quote here

"Whats worse? maybe 10 000 dead from a nuclear accident every 50 years or 500 million dead from global warming?"

James Lovelock

If this A.I. bollocks takes off, theres no way to power it reliably from wind/solar. pumped storage hydro... our biggest one of those lasts for 8 hrs (6 if you run it at full power), so its either fossil fuels or nuclear.

And while you get your knickers in a twist over building nuclear here in blighty, please remember that 60% of the power coming from France via the interconnects is generated ....by nuclear.

[Adair]

Re: Hmm

Just because we can do something doesn't make doing it sensible.

Nuclear fission as a source of power is entirely dependent on a 'hi-tech' society able to pay the enormous costs in construction, maintenance, decommissioning, and safety all down the line for generations afterwards.

There is absolutely no guarantee whatsoever that such a society, let alone civilisation, will exist in 200 years, let alone 500 years plus.

Basically a very expensive, irresponsible technological, positivist wet dream, for those who can afford it - for a while.

Burning stuff to generate power on a global level is, in its own way, clearly just as dumb and selfish.

But never underestimate human capacity for actions that are dumb or selfish, and frequently both.

Good things also happen. ;-)

[Alan Brown]

Re: Hmm

Existing nuclear power plants designs produce about an olympic swimming pool volume worth of waste per 1GW reactor, per 60 year lifespan

That waste is safe enough to handle with gloves on in about 150 years and less radioactive than the original fuel in 450-500

Molten salt liquid fuel designs (LFTRs) should reduce this volume by 95% (and be essentially inert in 450 years) if not 99% AND they can be fuelled on thorium - the primary waste product of rare earth mining AND they can be fuelled on depleted uranium(*) AND they can be fuelled on current high level nuclear waste

A coal fired power station produces an ash lake 6-10 metres deep covering 20 hecatres or more in the same period

The thing to remember is that highly radioative stuff becomes inert in a short period of time whilst long lived stuff ("20,000 years!") emits energies so low that they can be stopped with a sheet of paper. The chemical properties of plutonium/uranium are a vastly greater concern than their radioactivity.

Fusion has a big potential issue inasmuch as the neutrons are likely to produce radioactive iron from the countainment walls, which has a 60 year halflife and is problematic to handle. It may well be that fission ends up cleaner overall

(*) Being able to burn depleted uranium is critically important as this is the primary feedstock for making nuclear weapons and needs to be destroyed to reduce proliferation risks

[Anonymous Coward]

Re: Hmm

"make the best of energy sources that really don't stuff things up"

There aren't any that satisfy the perfect criteria so like Turkeys voting for Xmas we encourage our own demise in line with the Malthus and The Club of Rome's policy suggestions.

[Roland6]

Re: Hmm

>” How about counter with, 'OK fine, your mother is in hospital undergoing an operation and it's night, no solar, the wind decides not to blow”

Doesn’t need to be night…

With an increasing proportion of grid supply being sourced directly from solar “farms”, clouds become a huge problem…

I see this with my own panels, fortunately, at this time of year my batteries have sufficient chargers to handle matters so I’m not rapidly switching between panels and grid supply…

[Anonymous Coward]

Re: Hmm

100% correct!

[Evil Scot]

Re: Hmm

Torch the substations on the outskirts of London. Her life is more important than the output of a PC on LSD.

[Alan Brown]

Re: Hmm

We don't HAVE a "meantime"

It really is that simple. Waiting around for the holy grail of fusion is not an option. We have to roll with what we have NOW and upgrade to fusion later - IF it's ever commercially viable

[Alan Brown]

Re: Hmm

Attitudes of the greenies won't hold much water if rolling blackouts start becoming a winter norm

Right now the general public assumption is that energy planners know what they're doing and handwave the idea away. They don't realise that hands are tied and anyone with authority to issue warnings about critical infrastrcture shortages is being gagged by way of the threat of being sacked

It's the same issue with climate change. Politicians refused to accept science predictions in the 1990s and insisted they be watered down. What's actually happened is that things have been progressing EVEN FASTER than worst case predictions and the political structure has doubled down on the cognitive dissonance of denying what's sitting in plain view

[Anonymous Coward]

Re: Hmm

So what is this climate catastrophy that is happening? It's passing me by. Every year is a little different but not in the same direction. Satellites show increased greening. Some areas have increased ice pack. Why are we so sure some melting of ice is bad. The sea levels have barely changed - we would really know if they had. The ocean refuses to boil away. Climate has always changed due to complexity and the wonky orbit and rotational axis changes. The antarctic used to be ice free at times, I think the last was ~15000 years ago. The Nile Valley a massive fertile moderate climate much bigger than today just 5000 years BC. On the scale of life on earth including modern humans these are not long periods. The earth undergoes orbital changes about every ~ 24,000 years a cycle.

https://www.nasa.gov/centers-and-facilities/ames/human-activity-in-china-and-india-dominates-the-greening-of-earth-nasa-study-shows/

[Adair]

Re: Hmm

You expecting 'catastrophe' to happen at your convenience?

In the geological scheme of things 100 years, or 1000, are pretty much the same—a mere breath. And within that time you and I will have been and gone.

So, how quick do you want ecological catastrophe to be?

If it happens, whether it's all over by tomorrow, or still rolling out in a thousand years time, it will still be 'ecological catastrophe'.

Better stock up on the popcorn—could be a long wait (for one human lifetime).

[Anonymous Coward]

How about just fewer AI datacentres?

[Excused Boots]

"How about just fewer AI datacentres?"

Heresy! "We've caught a witch, may we burn her/him?"

[EvilDrSmith]

"We've caught a witch, may we burn her/him?"

Only if you plant a dozen trees to offset the carbon.

[Marcelo Rodrigues]

Be it AI, industries or households

Someone will use it.

I may be wrong (after all, I don't live there), but from here looks like Britain is already in serious need of electricity.

Couple that with AI and electric cars, and I would say the odds of overbuilding on the next 10 years are very slim.

Of course, the grid must be upgraded accordingly. And taking 5 years for nuclear isn't an excuse: it will take 5 years to build anyway - be it from today or from 3 years in the future.

[EvilDrSmith]

Re: Be it AI, industries or households

Fully agree.

Nick Clegg (Deputy prime minister in the Tory / Liberal coalition, for those that had managed to wipe that from their memory), (in)famously blocked new nuclear in 2010 because it wouldn't be ready for over 10 years.

Is it just UK politicians that seem to have no concept of there still being existence after the next General Election?

[Roland6]

Re: Be it AI, industries or households

But supported the investment in HS2, which also would no be ready within 10 years and also would require new power generating capacity… politicians…

[Alan Brown]

Re: Be it AI, industries or households

There's a big elephant that's being ignored - the phaseout of gas for domestic/industrial heating systems

We'll need to roughly double annual electricity production for this alone

[Roland6]

Re: Be it AI, industries or households

Hence why covering every roof with solar panels (and other mini generation technologies such as geo thermal and mini wind turbines) is so important. It means we get a grid upgrade without running lots of new power lines. Which means it can be delivered quicker than building a nuclear power station and the necessary grid interconnects.

[Roland6]

Re: Be it AI, industries or households

I think many are possibly missing my point. By having solar panels (and batteries) on most houses, we enable existing grid scale producers to be redirected to industry.

[richardnpaul]

FFS always with the nuclear

We need to re-educate people so that when they think about a consistent 24hr energy supply they think of geothermal, not nuclear. It's likely a damn site quicker and cheaper to power up new geothermal sites than it is to build and deal with nuclear power projects, and they would likely come in closer to being on budget and even maybe provide useful by-products; well to be fair nuclear provides us with Plutonium which I'd like us to turn in the phalus shaped bombs with "get reamed" emblazoned on the side. We could then send photos to Putler et al.

[Alan Brown]

Re: FFS always with the nuclear

Yeah.... good luck with geothermal. It's not particularly hot (wet steam), has major environmental effects (usually toxic runoffs), the water is extremely corrosive and because rock is a GREAT insulator they almost always take more energy out of the ground than is being replenished from below

"Yes, but, if you go deep you can get hotter and more sustainable sources" - except you're now talking about needing to frack at depths of 3-5+ miles. with even more problems than oil/gas fracking

Geothermal is one of those "nice ideas, but practicalities intrude" solutions that generally only works reasonably well over an active magma chamber or hotspot - not many of those in Britain

[Alan Brown]

Hmmm

For full decarbonisation, the UK will need to at least quadruple (more likely 6-8 times) its current annual TWh production (gas is being switched off(*)) for starters)

If the AI bubble which pushed builds of new nuclear capacity crashes just as demand starts ramping up, that would save a lot of rolling blackouts

(*) No, hydrogen is not a viable replacement. Why would anyone buy hydrogen at three times the cost per joule of electricity - which is 3-5 times the price per joule of existing natural gas supplies - Faced with a 9-15x higher gas bill most users will simply dump it in favour of heat pumps

[Scene it all]

Maybe get some Thorium tech from China? Much simpler than current Uranium reactors.

[charlieboywoof]

cooling tower graphic?

Nice try

[Anonymous Coward]

No problem

Don't worry Mad Ed Milliband has it covered. We just convert all the agri land to solar farms, create more slave labour and big open mines everywhere in the 3rd world and you'll have all the power you'll need, in summer. And you'll need loads of panels because he'll be dimming the sun as well. And no point worrying about farming because even if there was still enough land for it the sun would be so dim yields would be awful. But hey once people start starving we don't need loads of datacenters.

[Anonymous Coward]

"This might be true, but new atomic capacity simply can't be delivered fast enough to meet near-term demand"..."typically takes at least five years to construct"

From earlier in the article, they're looking 10 years out. A five year build time gives us plenty of time.

And Hinkley C was never primarily intended to generate power, it's there to keep money flowing to politicians' mates.

[stevebp]

Shame the Government's AI Energy Council is just inviting the hyperscalers to feed the agenda that more nuclear is needed for them - they don't care about anyone else - whereas in the US, they are funding the new nuclear facilities themselves. The Government should put its efforts into clearing out the regulatory hurdles to installing more smaller nuclear power facilities, so that SMR and vSMR can take off. It's the only way currently to build energy resilience and also meet Net Zero commitments. Thorium salt reactors are the sustainable future