NEW YORK, NY — September 23, 2024 — At Climate Week NYC, Anthropocene Institute and Pillsbury Winthrop Shaw Pittman LLP co-sponsored "AI for Advanced Nuclear Energy," moderated by Julie Kozeracki, Director of Strategy, DOE Loan Programs Office. The panel was part of an afternoon event, "Scaling Advanced Nuclear for Net Zero: Harnessing AI Innovation, Policy Mandates and Legal Developments."
Speakers included Kevin Kong, Founder and CEO of Everstar; Madhumitha Ravichandran, CEO and co-founder of N-ERGY; Brooke Daniels, Partner of Pillsbury Winthrop Shaw Pittman LLP; and William Theron, Founder and CEO of Deep Atomic. The panel covered how nuclear energy can support the growth of AI and how we can use AI to enable the successful operation and deployment of new nuclear reactors at scale.
(From left to right: Brooke Daniels, Kevin Kong, Madhumitha Ravichandran, William Theron, Julie Kozeracki)
Kozeracki noted important milestones, including Constellation Energy restarting Three Mile Island Unit 1 and Holtec International restarting the Palisades reactor with support from the DOE Loan Programs Office. "The turning back on of Palisades and Three Mile Island really represents a sea change in the valuation of clean firm power and the role it provides in ensuring that we can have 24/7 power that is disproportionately important for data center operations," she said.
Theron outlined growing demand for nuclear power and small modular reactors (SMRs). "For AI and data centers, there is now an economically viable model to bring in smaller reactors. If we compare it to other sources of energy, nuclear is slightly more expensive, but if we take all the other benefits we have, including colocation of these data centers and nuclear power plants, there are additional factors that make nuclear power very viable," he said.
Ravichandran chimed in that a unique value nuclear brings is being able to power space exploration and missions. Kong added, "What nuclear offers today counters what we've seen become problems of overbuilding on renewables. There needs to be better baseload power sources, and nuclear is the best candidate for propelling us into the future."
Kozeracki underscored that we still need to build wind and solar as much as we can, as fast as we can, but we still need nuclear. She noted that if you overbuild the generation capacity, you also must overbuild the storage and transmission. "System modeling shows that inclusion of 20-40% clean firm capacity like nuclear can reduce the cost of the system by 40%," she said. "So, I would like everyone here to internalize that it is cheaper to decarbonize with nuclear, and it's important that the marginal cost of solar, storage, and other assets are not representative of the whole system cost."
The discussion turned to the many different reactor designs, from large light water reactors to SMRs and MMRs, all suited for various use cases. Fortunately, AI can help streamline the process of interacting with the NRC, regardless of whether it involves reopening sites or deploying new types of reactors. Kozeracki noted that the DOE's updated Advanced Nuclear Commercial Liftoff Report breaks out the value proposition for large reactors, SMRs, and microreactors.
A big part of Theron's role at Deep Atomic is achieving economies of series and partnering with regulators to ensure that designs are well understood and that regulators understand what is required to manufacture reactors in series. "Modularity is something we would love to see in the industry," he said.
Kozeracki then asked the panelists about their perceptions of the NRC and the most significant opportunity for change or improvement. Ravichandran began, "What we have is the chicken and egg problem. To demonstrate a technology's viability, the regulator must see proof, whereas to show proof to a regulator, you actually have to show that the technology is viable. One cannot happen without the other. A real asset to us is the operational history of existing reactors that we can draw from and use to convince regulators about the viability of advanced reactors that are similar in design."
Kong added that the NRC is aware of AI developments for its internal use cases. "Everybody's on the same page now regarding the need to do things drastically differently. Part 53 will help all the private industry players," Kong said.
Kozeracki discussed using AI to address nuclear's credibility gaps and predictably build nuclear reactors on time and on budget. "There's a very high bar for building trust in AI applications for nuclear," she explained. "Walk me through how you think AI could help us with credibility or cognitive bias issues."
Kong stressed the importance of sharing data in the nuclear industry. "More data sharing and availability have to be the first problem we solve here to start cross-pollinating ideas and technologies across industries and corporate walls. That's how you get the next breakthrough," he said.
Added Ravichandran, "We have to accumulate the work that has already been done into knowledge graphs that we all can learn from. For example, coatings for accident-tolerant fuels have been a priority for more than a decade after the Fukushima accident, yet most in the industry have been focusing on the same kinds of coatings and materials because there is no cross-pollination. Something we can do today with AI is share the work being performed by scientists with the industry and with regulators."
Daniels added that AI explainability needs to improve. "We need a heavy emphasis on transparency, explainability, reliability, robustness, making sure that you can get the same answer over and over with accuracy, and establishing guardrails," she said.
Kozeracki asked the panelists, "If you could wave a magic wand, what would you wish AI could solve for nuclear?" Ravichandran wants an AI model that can tell exactly what's wrong when issues arise with a reactor—or an AI model that can translate which technologies can be deployed. Kong wants AI to solve the massive paperwork problem of the never-ending reports and analysis required.
Ultimately, Kozeracki emphasized the need to understand the value of clean, firm power. "A really critical piece around AI is how you are pricing in the decarbonization, reliability, and long-lived asset benefits of nuclear."
French 
English 
Japanese 
Chinese 
Spanish