The new hotness in data centers is to sign deals with nuclear power operators to supply them with reliable 24/7 electricity. There are two ways to do this. First, the data center connects to the power grid normally and just uses a lot of juice. This is called "in front of the meter." Second, a data center can connect directly to the nuclear plant for a specific amount of power. This is called "behind the meter."
Late last week, behind-the-meter hookups took a hit:
Late on Friday, the Federal Energy Regulatory Commission rejected Talen Energy’s request to increase the amount of power that it could provide directly from its existing Susquehanna nuclear power plant to Amazon’s data center from 300 megawatts to 480 MW. That hamstrings Talen’s ability to step up the amount of power it sells directly to Amazon, whose co-located data center has a potential capacity of up to 960 MW.
Data center owners like behind-the-meter deals because they're quick and provide a reliable amount of power. Nuclear operators like them because they can charge a hefty premium for the power.
The FERC decision isn't final, and it might be revised later after they collect more data about the cost and impact of behind-the-meter deals. For now, though, they say that overall grid reliability for existing businesses and consumers will remain their top priority.
So, what's the argument for not allowing direct from supplier deals? That it means less total power available to everybody else and presumably we don't have enough so therefore it should just all be done through metered service?
That sure sounds to me like a slam-dunk argument for build more high-capacity, non-fossil-fuel power plants already (read: nuclear).
It largely is, we can't reduce fossil fuel consumption through renewables alone and nobody wants to just admit that.
Actually, our experience with our solar array shows we are generating surplus electricity back to the grid for 9-10 months of the year. In the process we rack up credits for power supplied that we can draw down in the winter.
We didn't pay for any actual KwH used till the March/April billing cycle this year- a paltry $14 in charges. Then in the next billing we dropped down to only paying the meter fee - $7.49. And this is with charging a Leaf EV in our garage that we do most of our driving with these days and reducing our NG heating bills because we have a huge Heat Exchanger that means the furnace fires up a lot less frequently.
Now, if we add some battery farms in the mix (which is happening in a number of places - including Texas), I don't see why we can't put a large dent in the need for fossil fuels and keep driving their use down further over time. Note that a stationary battery farm can use lower energy density battery technology, since they can afford to be bulky.
"That sure sounds to me like a slam-dunk argument for build more high-capacity, non-fossil-fuel power plants already (read: nuclear)"
Sure. Except the fine folks at Southern Company have demonstrated that building new nukes is colossaly expensive. Vogtle 3 and 4 total a bit more than 2GW of capacity but cost an absolutely insane more than $35 billion. And it took 18 years to build them (bankrupting Westinghouse in the process, and nearly doing the same to Toshiba), so I'd guess that the investment community is not really likely to say "cut me a slice of that!"
I mean, given even the midpoint levelized cost, you could build nearly 350GW of solar generation for what Vogtle 3&4 cost. And sure, that wouldn't have the ~90% capacity factor that a nuke would, but utility scale solar in this country averages about 25%, so you'd still end up with almost 45x the energy generated.
Now, maybe a utility could argue that it learned from SoCo's experience and won't make quite so many mistakes, but even so, those are some pretty hard numbers to overcome for a utility to convince investment types to finance a new nuke. There's a reason that Amazon wanted a bigger behind-the-meter slice of an existing nuke (and why Microsoft is trying to get the idled 3 Mile Island units up and running) instead of building a new plant that won't pencil out and even if it did won't deliver a kwh for another decade or so.
Nuclear power was much more expensive than hoped. Even the French, who considered nuclear power development a matter of state security, were hit by soaring costs in the 1970s. Meanwhile, solar power keeps getting cheaper. Battery technology is lagging, but it too is getting cheaper.
I don't think there is much interest in building custom large reactors anymore given the cost fiascos.
SMR (small modular reactor) is getting the attention. Unfortunately most of those in existence are still experimental. The only one in "commercial" operation is in Russia, according to this article in powermag.com from 2023. Anyone interested in buying one with the embargoes in place at the moment?
These data center owners investing into SMRs have misplaced priorities. In the time between now and when the first SMRs are actually operational (if they're even built) they could have already covered their flat roofs with solar panels and filled in the gap with the construction of solar farms.
The first SMR to be approved (in 2022) in the US never got built (cancelled in 2023).
Why would this be? Cost and time.
Shocking..
This is just the tech bros trying to rip everyone off. They want the utilities and their customers to take all the risks and make all the big payments but give them dibs on cheap power. Let the tech bros build their own nuclear plants if they're so hot and horny for them.
When the aluminum industry took off in the Pacific Northwest, the deal was the feds would build the dams for hydro power, but the utilities were structured so that the aluminum refiners could get cheap power but they'd have to cut back in response to high power demand. The tech bros want to win both ways.
960 MW for one data center? We are never going to get rid of fossil fuels if we keep building more of these power hogs.