Argonne’s Aurora sets the stage for AI and nuclear energy executive summit
Leaders from private companies, government, and national laboratories gathered at Argonne National Laboratory on July 17 and 18 for an exclusive AI x Nuclear Energy Executive Summit that the Department of Energy called a first-of-its-kind forum to “align next-generation nuclear systems with the needs of digital infrastructure.”
Energy Secretary Chris Wright is on a tour of all 17 national laboratories, and the summit coincided with Wright’s visits to Argonne, in Lemont, Ill., and nearby Fermi National Accelerator Laboratory—and with a July 16 ribbon-cutting ceremony for Argonne’s exascale supercomputer Aurora.
Argonne lab director Paul Kearns said at that ribbon-cutting event that for the next days’ summit—or “what we call the AI and nuclear energy smashup”—organizers were “pulling the two communities together to think about how we can use nuclear power to really enable artificial intelligence, but also how we can use artificial intelligence to accelerate the deployment of advanced reactor technologies.”
Motivations: In his own comments at the ribbon-cutting event, Wright drew a parallel between the development of AI and the Manhattan Project. “There’s a national security implication of AI that means we can’t get second,” he said.
Wright said he is “very passionate” about trying to drive nuclear energy forward to power AI. “Nuclear was 6 percent of global energy in the year 2000. It’s 4 percent today,” he said. He attributes the backward step to politics and fear. “But we need to push back and win that tug of war and really launch a nuclear renaissance,” Wright said. “In the next few years I believe we will do that.”
During a media appearance after the dedication, Wright said the DOE is doing “a lot” to accelerate nuclear energy deployment.
“We’re offering up to developers of data centers and also to developers of nuclear facilities to use our land, use a different permitting process through the DOE,” he said. “We’ve released requests for proposals on building U.S. domestic enrichment capacity, and today, we had [a request for applications] go out on fuel fabrication. . . . So this is trying to use government funds to spur commercial businesses to grow their capabilities so we have the supply chain here to bring the cost of nuclear down and build faster.”
Aurora’s high-performance computing: Aurora is the third-most-powerful GPU-powered supercomputer in the world (after two other DOE machines—Lawrence Livermore National Laboratory’s El Capitan and Oak Ridge National Laboratory’s Frontier) and covers a floor space in the Argonne Leadership Computing Facility (ALCF) equal to four tennis courts. The machine was officially deployed for open scientific research in January.
Aurora’s 63,744 graphics processing units (GPUs) allow it to perform more than 1 quintillion (1 × 1018) calculations per second—qualifying it as an “exascale” machine. It’s a valuable research tool not only because it can run large-scale simulations to advance nuclear energy and the DOE’s research portfolio across a range of scientific fields, but also because, like commercial AI data centers, Aurora has high power needs that can fluctuate, providing a unique look into how to manage the large demand from data centers.
“We brought 60 megawatts of power to the building to support Aurora,” Michael Papka, deputy associate laboratory director for computing, environment, and life sciences and division director of the ALCF, told Nuclear News after the ribbon-cutting event. “Right now I would say we’re probably sitting somewhere right around 40 megawatts when everybody’s on the machine [and] it’s getting full use.
“We do pull off the commercial grid,” Papka said, adding that the local utility, ComEd, has been “a great partner.” The lab and ComEd have a strategy for “those peak summer days,” though. According to Papka, “We’ve never shut the machine down over this.” Part of the reason is that the power demands of GPU-based high-performance computing have researchers at Argonne and university and industry partners looking at innovative ways to ease power demand from large computers.
“We didn’t worry about power as much in the past,” Papka said.
“Let’s get nuclear technology rolling again”: Wright doubled down on nuclear for AI during his remarks to open the AI x Nuclear summit, which was cohosted by Argonne, Idaho, and Oak Ridge National Laboratories. “To power AI and to power our world, the most promising source, alongside our backbone hydrocarbon system, is nuclear,” he said.
“I view AI as just this incredible tool that’s going to leverage the power that . . . all humans have to simply make our world better,” he said. Wright exhorted attendees to “use the tools of AI to get nuclear technology rolling again.” He added, “It’s the people in this room who have been major drivers of those technologies—nuclear, AI, and computation.
“I’m all in on that as your partner in D.C.,” Wright said. “We have programs, we have funding. . . . Our fantastic lab partners are doing everything we can to offer up those lands, to offer up our licensing authorities, to get test reactors on the ground at Idaho National Lab going critical by next summer, July 4. We’re working across the whole government, including the NRC.
“How do we get regulation thoughtful and reasonable and not smothering?” Wright asked rhetorically before stating, “I think there is unanimity that that’s what we need to do.”
Summit conversations: Kirsten Laurin-Kovitz, associate lab director for nuclear technologies and national security at Argonne, described the AI x Nuclear summit as “AI for nuclear, nuclear for AI” when speaking with NN after the ribbon-cutting ceremony.
“Nuclear is high-density, scalable, reliable power. That is what AI needs,” she said. Conversations at the intersection of AI and nuclear include what nuclear energy can offer, she said, “but also how we can accelerate the deployment of nuclear to align it with the timelines of AI."
She added, “Nuclear doesn’t have a history of being a fast-moving industry. So the idea is to have AI companies and nuclear companies share perspectives on how they can better align.”
Discussions at the summit focused on defining priorities, identifying opportunities for joint action, and shaping future energy and AI policies, according to Argonne. Participants highlighted how AI can streamline reactor design processes, accelerate licensing, and optimize plant maintenance.
“Convening national leaders from across AI, nuclear energy, and the utilities sector reflects the urgency and opportunity of this moment,” said Kearns. “The collaborations strengthened here will help guide the future of AI and energy innovation.”
