Recent AI x nuclear moves: INL and ORNL joined Argonne National Laboratory to cohost an exclusive AI x Nuclear Executive Summit held July 17 and 18 at Argonne, for conversations between laboratory, government, and tech company leaders on “AI for nuclear, nuclear for AI.”
On July 23, the White House issued an AI Action Plan and an executive order on “Accelerating Federal Permitting of Data Center Infrastructure,” which in part prioritize new power deployment for AI and data centers. The DOE followed on July 24 with news that the DOE had selected sites on federal land—INL, the Oak Ridge Reservation, the Paducah Gaseous Diffusion Plant, and the Savannah River Site—to host new data centers.
ORNL and Atomic Canyon’s proprietary AI: On July 22 during the Nuclear Opportunities Workshop in Knoxville, ORNL signed a memorandum of understanding with Atomic Canyon on plans to extend the scope of an existing collaboration on using AI to streamline the licensing process for nuclear power plants.
ORNL and Atomic Canyon agreed to use ORNL’s high-performance computing resources “to create high-fidelity simulations that ensure the safety of designs while accelerating licensing with artificial intelligence to automate aspects of the review process.” The agreement would also allow Atomic Canyon to continue to develop a proprietary version of its Neutron AI platform, called Neutron Enterprise, with “exclusive capabilities and enhanced cybersecurity features to protect sensitive nuclear information.”
“ORNL was critical to the development of nuclear energy more than 75 years ago, and we are committed to advancing the technologies needed to sustain and grow the nation’s nuclear capacity today,” said ORNL Director Stephen Streiffer.
ORNL’s Frontier supercomputer is central to the plan. Atomic Canyon has already used Frontier to develop an AI model it calls Fermi to power the company’s Neutron AI platform. According an ORNL press release, the Fermi model allows users to quickly locate relevant documentation from the over 53 million pages of historical documentation in the Nuclear Regulatory Commission’s ADAMS public library and “has the potential to save the industry vast amounts of time and labor by streamlining reporting requirements and accelerating licensing and regulatory compliance processes.”
“Our mission at Atomic Canyon is to build the most advanced generative AI platform for the nuclear industry,” said Atomic Canyon CEO Trey Lauderdale. “ORNL’s expertise in nuclear science and high-performance computing was critical for us to be able to build AI in a reliable format. We want to double down on that relationship to build AI that can be used to help every reactor in America’s nuclear fleet.”
INL twin on Amazon’s cloud: Unlike Argonne or ORNL, the two other labs at the AI x Nuclear Executive Summit, INL doesn’t have an on-site exascale supercomputer packed with GPUs. But INL has partners. The lab announced a collaboration with Microsoft on July 16 to streamline nuclear permitting and licensing using Microsoft’s Azure cloud and AI technologies, as previously covered on Nuclear Newswire.
One week later, INL announced it will work with Amazon Web Services (AWS) to develop a digital twin of an unspecified small modular reactor to enable advanced modeling and simulation.
INL expects the tools it develops with Amazon “to reduce the costs and timeframes of designing, licensing, building, and operating nuclear facilities,” adding that “ultimately, the tools could be used for safe and reliable autonomous operation of nuclear reactors and accelerating deployment of new advanced reactors.”
By providing INL access to its cloud computing and AI capabilities, AWS is enabling “nuclear energy AI at scale,” said Chris Ritter, division director of scientific computing and AI at INL. “Through this collaboration with AWS, we have access to AI models, GPUs (graphical processing units), and specialized cloud services, including Amazon’s Bedrock service, which will enable INL researchers to use many leading foundation models to build nuclear energy applications,” Ritter said.
