On January 9, the hyperscaler announced that it would be continuing to invest in the nuclear business. The announcement went far beyond its deal last year to buy power from a single existing plant in Illinois and embraced a smorgasbord of financial and operational approaches to nukes. Meta will buy the output for 20 years from two nuclear plants in Ohio, it said, including additional power from increased capacity that will be installed at the plants (as well as additional power from a nuclear plant in Pennsylvania), plus work on developing new, so-far commercially unproven designs from nuclear startups Oklo and TerraPower. All told, this could add up to 6.6 gigawatts of clean, firm power.
Sounds good, right?
Well, the question is how exactly to count that power. Over 2 gigawatts of that capacity is already on the grid from the two existing power plants, operated by Vistra. There will also be an “additional 433 megawatts of combined power output increases” from the existing power plants, known as “uprates,” Vistra said, plus another 3 gigawatts at least from the TerraPower and Oklo projects, which are aiming to come online in the 2030s
Princeton professor and Heatmap contributor Jesse Jenkins cried foul in a series of posts on X and LinkedIn responding to the deal, describing it as “DEEPLY PROBLEMATIC.”
“Additionality” means that new demand should be met with new supply from renewable or clean power. Assuming that Meta wants to use that power to serve additional new demand from data centers, Jenkins argued that “the purchase of 2.1 gigawatts of power … from two EXISTING nuclear power plants … will do nothing but increase emissions AND electricity rates” for customers in the area who are “already grappling with huge bill increases, all while establishing a very dangerous precedent for the whole industry.”
Data center demand is already driving up electricity prices — especially in the area where Meta is signing these deals. Customers in the PJM Interconnection electricity grid, which includes Ohio, have paid $47 billion to ensure they have reliable power over the grid operator’s last three capacity auctions. At least $23 billion of that is attributable to data center usage, according to the market’s independent monitor.
“When a huge gigawatt-scale data center connects to the grid,” Jenkins wrote, “it's like connecting a whole new city, akin to plopping down a Pittsburgh or even Chicago. If you add massive new demand WITHOUT paying for enough new supply to meet that growth, power prices spike! It's the simple law of supply & demand.”
And Meta is investing heavily in data centers within the PJM service area, including its Prometheus “supercluster” in New Albany, Ohio. The company called out this facility in its latest announcement, saying that the suite of projects “will deliver power to the grids that support our operations, including our Prometheus supercluster in New Albany, Ohio.”
The Ohio project has been in the news before and is planning on using 400 megawatts of behind-the-meter gas power. The Ohio Power Siting Board approved 200 megawatts of new gas-fired generation in June.
This is the crux of the issue for Jenkins: “Data centers must pay directly for enough NEW electricity capacity and energy to meet their round-the-clock needs,” he wrote. This power should be clean, both to mitigate the emissions impact of new demand and to meet the goals of hyperscalers, including Meta, to run on 100% clean power (although how to account for that is a whole other debate).
While hyperscalers like Meta still have clean power goals, they have been more sotto voce recently as the Trump administration wages war on solar and wind. (Nuclear, on the other hand, is very much administration approved — Secretary of Energy Chris Wright was at Meta’s event announcing the new nuclear deal.)
Microsoft, for example, mentioned the word “clean” just once in its Trump-approved “Building Community-First AI Infrastructure” manifesto, released Tuesday, which largely concerned how it sought to avoid electricity price hikes for retail customers and conserve water.
It’s not entirely clear that Meta views the entirety of these deals — the power purchase agreements, the uprates, financially supporting the development of new plants — as extra headroom to expand data center development right now. For one, Meta at least publicly claims to care about additionality. Meta’s own public-facing materials describing its clean energy commitments say that a “fundamental tenet of our approach to clean and renewable energy is the concept of additionality: partnering with utilities and developers to add new projects to the grid.”
And it’s already made substantial deals for new clean energy in Ohio. Last summer, Meta announced a deal with renewable developer Invenergy to procure some 440 megawatts of solar power in the state by 2027, for a total of 740 megawatts of renewables in Ohio. So Meta and Jenkins may be less far apart than they seem.
There may well be value in these deals from a sustainability and decarbonization standpoint — not to mention a financial standpoint. Some energy experts questioned Jenkins’ contention that Meta was harming the grid by contracting with existing nuclear plants.
“Based on what I know about these arrangements, they don’t see harm to the market,” Jeff Dennis, a former Department of Energy official who’s now executive director of the Electricity Customer Alliance, an energy buyers’ group that includes Meta, told me.
In power purchase agreements, he said, “the parties are contracting for price and revenue certainty, but then the generator continues to offer its supply into the energy and capacity markets. So the contracting party isn’t siphoning off the output for itself and creating or exacerbating a scarcity situation.”
The Meta deal stands in contrast to the proposed (and later scotched) deal between Amazon and Talen Energy, which would have co-located a data center at the existing Susquehanna nuclear plant and sucked capacity out of PJM.
Dennis said he didn’t think Meta’s new deals would have “any negative impact on prices in PJM” because the plants would be staying in the market and on the grid.
Jenkins praised the parts of the Meta announcement that were both clean and additional — that is, the deals with TerraPower and Oklo, plus the uprates from existing nuclear plants.
“That is a huge purchase of NEW clean supply, and is EXACTLY what hyperscalars [sic] and other large new electricity users should be doing,” Jenkins wrote. “Pay to bring new clean energy online to match their growing demand. That avoids raising rates for other electricity users and ensures new demand is met by new clean supply. Bravo!”
But Dennis argued that you can’t neatly separate out the power purchase agreement for the existing output of the plants and the uprates. It is “reasonable to assume that without an agreement that shores up revenues for their existing output and for maintenance and operation of that existing infrastructure, you simply wouldn't get those upgrades and 500 megawatts of upgrades,” he told me.
There’s also an argument that there’s real value — to the grid, to Meta, to the climate — to giving these plants 20 years of financial certainty. While investment is flooding into expanding and even reviving existing nuclear plants, they don’t always fare well in wholesale power markets like PJM, and saw a rash of plant retirements in the 2010s due to persistently low capacity and energy prices. While the market conditions are now quite different, who knows what the next 20 years might bring.
“From a pure first order principle, I agree with the additionality criticism,” Ethan Paterno, a partner at PA Consulting, an innovation advisory firm, told me. “But from a second or third derivative in the Six Degrees of Kevin Bacon, you can make the argument that the hyperscalers are keeping around nukes that perhaps might otherwise be retired due to economic pressure.”.
Ashley Settle, a Meta spokesperson, told me that the deals “enable the extension of the operational lifespan and increase of the energy production at three facilities.” Settle did not respond, however, when asked how Facebook would factor the deals into its own emissions accounting.
“The only way I see this deal as acceptable,” Jenkins wrote, “is if @Meta signed a PPA with the existing reactors only as a financial hedge & to help unlock the incremental capacity & clean energy from uprates at those plants, and they are NOT counting the capacity or energy attributes from the existing capacity to cover new data center demand.”
There’s some hint that Meta may preserve the additionality concept of matching only new supply with demand, as the announcement refers to “new additional uprate capacity,” and says that “consumers will benefit from a larger supply of reliable, always-ready power through Meta-supported uprates to the Vistra facilities.” The text also refers to “additional 20-year nuclear energy agreements,” however, which would likely not meet strict definitions of additionality as it refers to extending the lifetime and maintaining the output of already existing plants.
Swiss Re
