Tech giants set to pay through the nose for nuclear power that's still years away
- Reference: 1729256411
- News link: https://www.theregister.co.uk/2024/10/18/big_tech_nuclear_power/
- Source link:
This week alone, Google and Amazon both confirmed more agreements. Google said it had lined up a deal to [1]purchase nuclear energy from Kairos Power , even though the prospectve supplier hasn't yet developed the small modular reactors (SMRs) it proposes to fulfill the contract with.
Microsoft's carbon emissions up nearly 30% thanks to AI [2]READ MORE
In a similar arrangement, Amazon said it was [3]handing half a billion dollars to three companies to develop SMRs to provide power generation in Virginia and Washington. In this case, the energy isn't expected to start emerging until the 2030s.
And Microsoft is also dabbling in nuclear, recently establishing a power purchase agreement (PPA) with Constellation Energy to [4]restart the Three Mile Island power plant .
The key part, at least in Google's case, is that it is seeking additional clean power, not repurposing existing power, and requires electricity that will be available 24/7, according to Jefferies Research, part of the [5]Jefferies financial services group .
[6]
The report, shared privately with The Register , states that SMRs are likely to be "significantly more expensive" than existing pressurized water reactor designs, which can compete with natural gas or renewables but take six to eleven years to construct – too slow to meet the rapidly growing demands of datacenters.
[7]
[8]
"There are two key concepts here: additionality, i.e. new generation to match new load; and matching clean energy generation with each hour of load," the Jefferies report states. Existing nuclear fails the "additionality" test despite offering 24/7 clean energy, whereas intermittent renewables lack the reliable capacity factor datacenters require.
Some of Mountain View's new bit barns have been indirectly tied to increased coal and natural gas generation, including in Omaha and South Carolina, as the extra electricity required is not associated with new clean energy sources, the banker's report says.
[9]
Also notable is that the Chocolate Factory is "not prioritizing speed to market," with Kairos Power's first SMR not expected to be online before 2030, followed by additional reactor deployments through to 2035. This could be seen as Google taking a long-term view on energy supply, or it may simply reflect the reality that SMRs just aren't ready yet.
"On the positive side, Google's interest in purchasing presumably very expensive nuclear power shows the inherent value of 24/7 clean energy," the report states, but it also adds that the emphasis on new clean power rather than existing clean power somewhat downplays the validity of power purchase agreements with existing nuclear sources, such as Amazon's with [10]Talen Energy earlier this year.
[11]Amazon makes $500M bet on itty-bitty nuclear reactors to fuel cloud empire
[12]Viable fusion power in a decade? Tokamak Energy dares to dream
[13]Google hopes to spark chain reaction with nuclear energy investment
[14]DoE awards next-gen nuclear fuel contracts backwards
Looking at existing nuclear capacity in the US, Jefferies estimates there is about 97 GW, split roughly 60/40 percent between regulated and merchant/unregulated generation, able to produce 792 TWh average annual output. This compares with estimates for datacenter energy demand of 178 TWh right now, growing to somewhere in the region of 600 TWh by 2030.
As far as merchant or unregulated nuclear generation goes, there is only enough capacity to meet expected growth in datacenter power demand through 2026, with a likely shortfall in the following years. Even if only hyperscale companies are considered, their estimated 425 TWh demand by 2030 will be 25 percent higher than the forecast merchant nuclear plants' output.
The analysis by the report's authors is that regulated nuclear plants are unlikely to have their output dedicated towards supplying bit barns, but that there is "a real possibility of regulated utilities creating new tariffs at premium pricing to dedicate nuclear output to a datacenter."
So much for green Google ... Emissions up 48% since 2019 [15]READ MORE
According to Jefferies, the tech companies face paying upwards of $85 per MWh for co-located nuclear, or else trying regulated states such as South Carolina for prices potentially as low as $60 per MWh. Google's actions show it is focused on low price carbon power when available with new clean energy as a mitigating element, it says.
The report is primarily aimed at investors, and so is looking at the situation from the angle of what Google's move means for the assumption that there will be profit in selling existing nuclear power to hyperscale customers for well above market prices. If Google is hypothetically not interested in existing nuclear generation, this still leaves Amazon, Microsoft, and Oracle as potential buyers, it says.
[16]
An earlier [17]report from Morgan Stanley concluded that there is investment potential in decarbonization efforts as datacenter operators strive to meet 2030 carbon neutrality goals. ®
Get our [18]Tech Resources
[1] https://www.theregister.com/2024/10/15/google_kairos_smr_nuclear_investment/
[2] https://www.theregister.com/2024/05/16/microsoft_co2_emissions/
[3] https://www.theregister.com/2024/10/16/amazon_nuclear_smr/
[4] https://www.theregister.com/2024/09/20/three_mile_island_nuclear_plant_microsoft_ai/
[5] https://www.jefferies.com/
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZxKGJdFJjItPH3TcefCgcQAAAM8&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZxKGJdFJjItPH3TcefCgcQAAAM8&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZxKGJdFJjItPH3TcefCgcQAAAM8&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZxKGJdFJjItPH3TcefCgcQAAAM8&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[10] https://www.theregister.com/2024/03/04/amazon_acquires_cumulus_nuclear_datacenter/
[11] https://www.theregister.com/2024/10/16/amazon_nuclear_smr/
[12] https://www.theregister.com/2024/10/16/tokamak_fusion_pilot/
[13] https://www.theregister.com/2024/10/15/google_kairos_smr_nuclear_investment/
[14] https://www.theregister.com/2024/10/08/does_awards_nextgen_nuclear_fuel/
[15] https://www.theregister.com/2024/07/02/google_datacenter_emissions/
[16] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33ZxKGJdFJjItPH3TcefCgcQAAAM8&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[17] https://www.theregister.com/2024/09/06/datacenters_set_to_emit_3x/
[18] https://whitepapers.theregister.com/
Re: They can afford it, and we need them to...
@John Robson
"So I don't care that they're likely paying over the odds"
The question is if they are gonna pay over the odds, or if they expect energy prices to continue to be screwed upwards by unreliables.
Re: They can afford it, and we need them to...
Yeah the fixed rate that will be paid for power from Hinkle Point C that people were complaining was too high... is now lower than current market rates.
Re: They can afford it, and we need them to...
So people are complaining that SV companies are making long term investments in infrastructure ?
Re: They can afford it, and we need them to...
"Yeah the fixed rate that will be paid for power from Hinkle Point C that people were complaining was too high... is now lower than current market rates"
How so? Wholesale day ahead is around £65-75 MWh over recent months, allowing for the indexation Hinkley will get £130 MW/h based on CPI index since the original rotten deal was struck (and that £130 will keep on being uplifted every year).
The UK energy "market" is thoroughly broken. Market forces don't dictate what the market does, government policy specifies everything, and the result is that we have amongst the highest energy prices in the world. And when Hinkley comes on stream it'll get higher.
So, money is being transferred from entities that have way too much of it, to entities that work on clean energy. Sounds good to me.
Except that the money is mostly going to commercial companies who are competing with each other in the SMR field. That means they're each working on first of a kind prototypes, most are going to go bust because they won't have the resources, and those that succeed will find they can build a modest number of units which will then have very high unit costs.
SMR (or any other fission or fusion technology) simply isn't suited to commercial development. Risks are big, all the costs are in development, certification and specialised skills and parts, so government needs to intervene to ensure that there's common standards and that the overheads and fixed costs can be recovered across lots of similar plants. Absent that, all this money gets is a pile of vastly expensive dissimilar plant that will become a nuclear liability in future years.
China, I regret to say, is at least five if not ten years ahead of the West on SMR, simply because they've been able to use state sponsorship to deliver results, and they've not been bedevilled by startup businesses claiming they can magic up an SMR from an old washing up liquid bottle, a shoe box, and sticky backed plastic.
Unregulated Nuclear Generation ?
That fills me with confidence...
Re: Unregulated Nuclear Generation ?
Yeah cos the EPA, DoE and the other regulatory agencies are just going to say, "Oh, you're a tech company not a utility? Well help yourself, move fast and break things - bro"
Meanwhile in silicone valley…
Buy all the power you want, you will all be my subjugates after my plan goes into effect. Ha, ha, ha..
Muahaaaaa haaa
-Max Zorin, Zorin Industries
Is it genuine?
If they really truly want to buy product, why are they messing about with random startups instead of going to suppliers with a more established background?
At least some of the other people looking at SMRs have already been in the game of building small reactors and/or traditional nuclear facilities. Some of the newer startups look more like dreamers or vapourware with little prospect of product even if they find a backer.
The years away/startup combo for datacentre reads more like a cheap promise they don't truly plan to deliver. If they were wanting to actually have a pile of kit to run AI or whatever they'd surely be looking at closer timescales?
Re: Is it genuine?
Whilst I'd agree that working with people who have actually built small reactors is the obvious option, those companies tend to be big listed entities that will be a lot more measured and realistic than the hopefuls.
BWXT, Fluor, Rolls Royce have the expertise and could do all of this easily, but aren't looking to splurge lots of speculative cash without some confidence that there's a viable return on investment at the end. With government and regulators the primary obstacles to SMR, it's in government's hands - do they want this or not? If they do they need to simplify the regulation and planning, and make energy policy accommodating. Or they can do what they're currently doing, of saying nice words but doing precisely nothing other than pushing wind and solar plants.
Re: Is it genuine?
> BWXT, Fluor, Rolls Royce
One of those (BWXT) gave up developing an SMR, the second (Fluor) appears to be a construction/management company that partners with other SMR manufacturers, and the third (RR) is very busy building lots of nuclear reactors for new submarines using pressurised water technology and probably don't have the capacity to expand into SMRs at the same time.
Submarines and ships
Who builds the reactors for submarines? Are they in this? They've had reactors on subs for decades. Maybe not as small and modular as the hyperscalers want, but surely they have useful experience here.
Re: Submarines and ships
Yeah but, it's likely that those subs are under National Security rules and you're not going to be able to just copy/paste their reactors to the civilian world because The Man still doesn't want anyone to know their abilities.
The Cold War may be dead, the paranoia isn't.
Re: Submarines and ships
What you're saying then is that really there's an opportunity for the US Navy to turn itself into a power company, and get paid loads of dosh to build and run lots of nuclear subs?
Re: Submarines and ships
The thing is with reactors on subs and on aircraft carriers is that you have a large and skilled crew available 24/7 constantly monitoring and repairing the reactors, plus as much cooling water as you could ever need to keep them operating correctly.
A land-based SMR needs to be able to operate with fewer personnel (anyone running one of these will staff it as minimally as possible), operate in a range of temperatures (from deserts to arctic tundras) with limited water supplies (to varying degrees of course) and with probably longer maintenance and repair cycles (again to reduce cost).
So the designs probably aren't directly comparable - the military ones probably also use older tech as it's more developed/reliable than some of the newer stuff which may be more efficient or easier to maintain etc.
They can afford it, and we need them to...
So I don't care that they're likely paying over the odds - the benefit to society at least exists here, unlike the rest of their tax avoidance schemes.