Meta Platforms Had Talks to Acquire Perplexity Before Scale AI Deal, Bloomberg Reports

A handful of tech giants have signed deals with nuclear energy companies in recent months to meet outsized electricity demand amid the artificial intelligence boom, but they have been opting for one of two distinct paths: new and innovative versus old and reliable.

Microsoft kicked off the dealmaking rush in September 2024, announcing a 20-year energy agreement with Constellation Energy that will pave the way for the revamp and restart by 2028 of Three Mile Island Unit 1 in Pennsylvania under the new name Crane Clean Energy Center. Under the terms of the deal, Microsoft will purchase energy from the renewed plant, which is expected to add more than 800 megawatts of electricity to the grid.

That agreement was followed by deals announced in October by Amazon.com and Alphabet's Google with companies producing next-generation nuclear technology in the form of small modular reactors, or SMRs.

Google announced its clean energy agreement with Kairos Power on Oct. 14. Under the terms of the deal, the first of its kind, Google will purchase nuclear energy from multiple SMRs to be developed by Kairos, with the stated goal of bringing the first SMR online by 2030, followed by additional SMR deployments through 2035. The deal will enable up to 500 megawatts of carbon-free power to US electricity grids, Google said.

Two days later, Amazon and Energy Northwest, a public power agency, announced an agreement to fund efforts to deploy SMRs in Washington state developed by X-energy. Under the deal terms, Amazon will have the right to purchase electricity from the first project, consisting of four SMR modules, which is expected to generate 320 MW of energy capacity starting in the early 2030s.

Meta Platforms and Constellation announced a 20-year deal on June 3 in which the tech giant agreed to buy electricity from Constellation's Clinton Clean Energy Center in Illinois. Under the terms of the agreement, the Clinton plant will provide Meta with 1,121 MW of energy starting in June 2027.

Deals like those made by Microsoft and Meta to secure energy from existing nuclear plants may be the more practical option in the nearer term for big tech companies that are engaged in a generative AI arms race requiring significant amounts of electricity to power their data centers, experts said.

"Momentum among experts is shifting back toward the large plants because economies of scale are real and costs matter — large-scale has significant cost advantages over small-scale," said Rob Gramlich, president of Grid Strategies, a Washington, D.C.-based consultancy. "That said, small modular reactors have the advantage of the prospect that you could manufacture them onsite and get all the efficiencies of manufacturing. That's real, and, of course, there's a risk advantage with [SMRs] because the cost of a large unit is arguably too big for any one company to absorb."

Neither revitalizing and restarting existing plants nor a few SMR projects will come close to meeting data-center demand on their own, experts said. Global power demand from data centers will increase by 50% by 2027 and by as much as 165% by the end of the decade compared to 2023, Goldman Sachs Research said in a note. That means many new reactors — big and small — will need to be built, a costly, time-sensitive proposition.

"Almost all of [the play for nuclear] is long term," Gramlich said. "You're not going to get a new plant, conventional or advanced, this decade."

New, large nuclear reactors would be equipped with advanced technology matching that of SMRs, said Stephen Comello, senior vice president of strategic initiatives at the EFI Foundation, in an interview.

"Just because they're big, it doesn't mean they're dumb," he said, citing two new units being added to Plant Vogtle in Georgia that will utilize the Westinghouse AP1000 reactor technology as an example.

Future AP1000 builds will be costly at around $10 billion and take a long time to construct — roughly six to seven years — but they have the institutional knowledge advantage over SMRs, Comello said.

"The good thing with large reactors, we've already built them, there already have been lessons learned ... (we) already have licenses approved and the supply chain is pretty much there," he said. "SMRs may cost less, financing might be less, but the technical risk profile is not the same. We haven't built SMRs before, and we're bound to encounter some in-construction updates no matter how well they're planned for."

SMRs also still need to procure certain licenses certifying they work as intended and meet safety specifications, all steps AP1000s have already achieved. Once the first few SMRs are built, those reactors may become far faster to build and less risky from a cost perspective, Comello said.

Either way, it will be a while before nuclear output scales to the point where it can better meet the needs of data centers. In the meantime, big tech will have to continue to rely heavily on fossil fuels and less-productive renewables.

"They'll get grid power, which, depending on the region, will be a lot of gas, at least a third, sometimes half, a fair amount of solar and wind, and probably some coal," Grid Strategies' Gramlich said.

Meta Platforms said Friday it is collaborating with sports eyewear brand Oakley on performance artificial intelligence glasses, and their first product, Oakley Meta HSTN, is set to be available for preorder starting July 11.

Oakley Meta HSTN features a built-in camera, open-ear speakers, and the Meta AI personal assistant, the company said.

The product will be available in the US, Canada, the UK, Ireland, France, Italy, Spain, Austria, Belgium, Australia, Germany, Sweden, Norway, Finland and Denmark. It is also expected to be available in Mexico, India, and the United Arab Emirates later this year, the company added.

Shares of Meta fell 1.3% in recent trading.

By Adam Clark

Meta Platforms stock has climbed this year in response to the social media company's progress with artificial-intelligence. Now, it is reportedly stepping up its efforts, trying to acquire a major AI start-up and recruiting new AI executives.

Meta looked to buy Safe Superintelligence earlier this year but was rebuffed by its founder Ilya Sutskever, CNBC reported late Thursday, citing people familiar with the matter. Safe Superintelligence was valued at $30 billion in a funding round in March.

Meta and Safe Superintelligence didn't immediately respond to requests for comment.

On the face of it, such an acquisition would have been an odd move. Safe Superintelligence hasn't released any products, as it concentrates on developing supersmart AI. Meta has also been getting along perfectly well on its own, with its stock up 19% so far this year.

The real attraction of such a deal likely would have been to get Sutskever and his key employees on board.

Sutskever was previously chief scientist at OpenAI, where he helped develop the technology behind ChatGPT. He left OpenAI last year following a break with its CEO Sam Altman, and subsequently launched Safe Superintelligence.

Thwarted in his efforts to bring Sutskever on board, Meta CEO Mark Zuckerberg has instead negotiated to recruit Safe Superintelligence's CEO Daniel Gross, as well as former GitHub CEO Nat Friedman, according to CNBC. Gross and Friedman are partners in the investment fund NFDG, which has backed several AI start-ups.

So far, Meta has relied on in-house AI models, as opposed to acquiring or funding an AI start-up as Microsoft has done with OpenAI and Amazon.com has with Anthropic. However, there have been signs that Zuckerberg feels Meta's AI team needs bolstering.

Last week, Meta completed an investment in Scale AI. The Wall Street Journal reported that Meta would pump $14 billion into the data-labeling company in exchange for a 49% stake and that Scale AI founder Alexandr Wang would join Meta.

The bigger picture here is that multiple AI companies have delayed the releases of their next flagship models amid concerns they don't show sufficient improvement. That suggests the industry's "scaling law," the idea that larger and more complex models are automatically more intelligent, is breaking down.

Meta is among those struggling to make a breakthrough. Its "Behemoth" model, originally meant to be released in April, is being delayed until fall or later, according to the Journal.

The response from AI companies has been the development of so-called reasoning models that break down problems step-by-step. However, a recent paper from researchers at Apple found "fundamental limitations" in such models. At tasks beyond a certain level of complexity, these AIs suffered "complete accuracy collapse," according to the researchers.

That suggests the industry will need to adopt new techniques to push AI to the next level of intelligence. Meta will hope that its new recruits can get there first.

Write to Adam Clark at adam.clark@barrons.com

This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.