In 2016, Pacific Gas & Electric planned to wind down operations at California’s Diablo Canyon power plant by 2025, citing concerns from environmental advocates about its waterfront location and perch on multiple fault lines.
But in September 2022, PG&E received unexpected news from the state. Governor Gavin Newsom signed SB 846 into law, extending nuclear power plant operations until October 2030.
“We needed to pivot in order to continue operating,” Maureen Zawalik, the company’s senior vice president and chief risk officer, who has worked at PG&E for 30 years, told Business Insider.
With deadlines for regulatory filings required by federal and state authorities to continue operating Diablo Canyon approaching, and job losses due to previously planned closures, Zawarik said he felt it was worthwhile to look to AI to streamline PG&E’s back-office processes.
In January 2024, Zawalik agreed to meet with Trey Lauderdale, a local resident and CEO and founder of Atomic Canyon, an artificial intelligence startup focused on building a ChatGPT-like search tool for the nuclear energy sector.
In addition to location, Lauderdale said his past expertise in the healthcare sector, which operates with many regulatory burdens, influenced his interest in nuclear power. His family lived 40 miles away from Diablo Canyon, he told Business Insider. He envisioned developing an AI solution that would reduce the administrative costs of the nuclear industry.
PG&E and Atomic Canyon have jointly built the first commercial installation of on-site generation AI tools at a U.S. nuclear power plant. The AI tool, called Neutron, was designed to make it easier for Diablo Canyon employees to find and use the millions of documents they are legally required to keep on file, Zawalik said.
Because the nuclear sector is highly regulated at both the federal and state levels, workers have a large amount of documentation and data to manage. At Diablo Canyon, billions of data points, including engineering records, design updates, and regulatory filings, are distributed and stored on-site across six systems. Because Neutron connects to all of these systems over the internet rather than the cloud, staff can use the AI tools with a simple web-based search on their factory computers.
technology
In December 2023, Lauderdale founded Atomic Canyon and hired an AI engineer he knew from the healthcare field.
His plan was to build a generative AI tool that could understand nuclear terminology better than OpenAI’s ChatGPT or Anthropic’s Claude. To train the AI model on the correct terminology, the Atomic Canyon team downloaded 53 million pages of publicly available industry data from the Nuclear Regulatory Commission, the government agency that regulates commercial nuclear power plants, Lauderdale said.
Then, in May 2024, Atomic Canyon announced a project with the Department of Energy’s Oak Ridge National Laboratory, based in Tennessee. Lauderdale’s team will deploy a supercomputer called Frontier that could train Atomic Canyon’s AI models on a treasure trove of nuclear terminology, the institute said.
In November 2024, PG&E announced that it had selected Atomic Canyon to develop what would become Neutron at Diablo Canyon.
Neutron helps PG&E employees capture documentation of operating and maintenance procedures, past engineering design changes, and state and federal regulatory requirements.
“There are more than 9,000 steps in operating and maintaining a plant,” Zawalik said. The 40-year-old Diablo Canyon has billions of pages of data to track and analyze. “And we’re just one power plant,” she added.
Previously, workers had to crawl through multiple systems to find all the information. With Neutron, you get these details from a single database built and running on AI chipmaker Nvidia’s AI platform.
Zawarik said PG&E piloted Neutron in the first half of 2024 with a small beta test group that included employees from engineering, maintenance and operations teams.
Starting in July 2025, PG&E will roll out Neutron in phases, including training sessions to teach employees how to use the AI-enabled search tool, Zawarik said. By fall 2025, all 1,300 Diablo Canyon employees will have access.
result
As of January 2025, Neutron’s search and generation AI summaries have significantly reduced the total time required to search for documents, a PG&E spokesperson told Business Insider.
Zawarik shared one case study of how this works. When investigating the Diablo Canyon safety valve issue, a team of 12 employees was previously assembled to retrieve all necessary documentation before any valve repair work was performed. Zawarik said the project typically takes about 180 days. With Neutron, teams of the same size can get all the documents they need in less than 40 days.
In the future, Zawalik expects generative AI tools like Neutron will make it easier for utilities to build new solar, wind or nuclear facilities to meet America’s growing energy needs.
“Tools like this are extremely useful for regulatory filings, construction, and engineering,” Zawalik said.