In the last year, virtually every major tech company has signaled their interest in advanced nuclear power to fuel their data center ambitions. Google signed a deal with small modular reactor company Kairos Power. Meta issued a request for proposals targeting up to four gigawatts of new nuclear capacity, including SMRs. And late last year, Amazon both invested $500 million in the advanced reactor company X-energy and partnered with Energy Northwest to deploy a nuclear project in Washington.
Given the breadth of the advanced nuclear field today and the immense energy needs of the artificial intelligence era, a lot goes into inking deals like the ones Amazon signed in October, said Daniel Gross, a director at Amazon’s Climate Fund, which initially identified X-energy as a potential investment for the company.
Amazon competitor Microsoft embraced nuclear via a deal to restart a conventional nuclear reactor at the notorious Three Mile Island in Pennsylvania. But Amazon has instead opted for advanced nuclear, in part because of the safety characteristics, Gross said — and because they were looking for “something that was small and modular and able to turn on and commission in a way that better matches the way in which we turn on and commission [data centers.]”
The tech sector has a long history of helping to commercialize energy technologies. But getting the advanced nuclear sector off the ground is especially tricky, Gross said. And the distinction between signing a power purchase agreement and investing in the technology itself is significant, he added.
“A PPA for a solar project or a wind project has, for the last decade, been sufficient to get that project built and financed,” he said. “If an independent power producer or renewable energy developer is able to sign a power purchase agreement with a hyperscaler, the credit quality is so good, the certainty around the revenue is so good that they’re able to borrow money, raise tax equity, and raise infrastructure money in order to build.”
That’s also the case with other “proven” technologies, like natural gas, for example, he added. But SMRs are complicated in part because there are none currently operating in the U.S.
“So regardless of which SMR tech you look to, none of those technologies would be viewed as being proven by infrastructure investors,” Gross said. “And therefore just having a PPA…isn’t enough to raise the money that’s necessary to take the risk of construction, or even the earlier-stage risk of getting a license through the [Nuclear Regulatory Commission].”
So while some PPAs for advanced nuclear have been signed, they haven’t catalyzed investment activity from investors who would ordinarily pay for development and construction. A PPA “is not an orderbook,” said Gross.
And therein is the gap: no parties are coming up with the capital to actually build the advanced nuclear power. Utilities, which might seem like the obvious choice, mostly aren’t taking the lead, thanks in part to fears of delay and cost overruns, which were made worse by the case of the Vogtle plant in Georgia.
“And then the independent power producers, who have [historically] built combined cycle gas plants and these days are building a lot of wind and solar, they’re not stepping up to take that risk either,” Gross said. “They will take development and construction risk on wind and solar, but not on SMRs.”
That’s the observation that led Amazon to the deals it ultimately signed with X-energy and Energy Northwest: “If we want the power, we need to address the shortfall in the market, which is capital for development, licensing, and construction,” Gross explained.
Energy Northwest lacked the capital to build its X-energy project until October 2024, when Amazon agreed to fund development, licensing, and construction. That type of funding, he said, is ultimately what enables an orderbook.
The stock decision
Another key element of Amazon’s SMR investment, Gross said, is its decision to purchase stock in X-energy. That’s again different from a PPA in terms of its purpose and impact, he added.
“The challenge for X-energy is as an original equipment designer,” he said, “they still need capital in order to finish their design and take it through a licensing process, and signing a PPA doesn’t help them raise the capital in order to do that; investing the capital does.”
The project X-energy will build in Washington to supply power to Energy Northwest — while an early iteration — isn’t technically first-of-a-kind. The SMR developer is deploying its first SMRs in Texas, for a Dow Chemical manufacturing site. Just this week, the Dow subsidiary Long Mott Energy applied to the NRC for its construction permit, saying the project would nearly zero out the Seadrift, Texas, facility’s roughly 500,000 metric tons of greenhouse gas emissions each year.
The Dow project was influential in Amazon’s selection of X-energy for its SMR investment, Gross said. “We have a lot of respect for their capabilities in managing meta projects,” he explained. X-energy also received a vote of confidence from the Department of Energy, which signed a cooperative agreement in 2021 to support X-energy through the Advanced Reactor Demonstration Program.
Of course, geography was also key. “The Pacific Northwest is a very important geographic region for our data center operations,” Gross added. And Energy Northwest, the only utility operating a nuclear power plant in the Pacific Northwest, had already vetted and decided to go with X-energy.
“So there were all of these other points of endorsement that came to the same conclusion that we were coming to that helped us get comfortable,” Gross said.
Amazon is hoping its partnership with X-energy will deploy as much as 5 GW of SMR capacity by 2039. That first project in Washington state will be 320 megawatts initially, with plans to ultimately scale it to 960 MW.
But Gross said the company is clear-eyed about just how far advanced nuclear still has to go, even with Amazon acting as an infrastructure financier for the Energy Northwest project. After all, one project doesn’t mean traditional investors and utilities will suddenly be lining up to fund SMR development. Even with second- and third-of-a-kind projects, he added, “financing challenges persist.”
