Acentric Research is produced with AI tools under human review. The authors are not financial professionals, and nothing here is investment advice or a recommendation to buy or sell any security. AI analysis can be wrong; check every claim against its cited source.
This Brief is an initial scan, not a full investigation. The method and scores behind it are in AI × Energy in the Landscape Report.
Overview
Building a nuclear reactor takes years of design, safety analysis and paperwork, and each licence application is reviewed by a regulator. The U.S. Department of Energy (DOE) states that nuclear plants have historically been challenged by long development timelines and rising costs. [1] This Brief covers artificial intelligence (AI) used in designing, licensing and operating fission reactors, including advanced and small modular designs.
What stands out is that nearly all of the public evidence found is about licensing paperwork. DOE has reported a test case in which AI drafted sections of a licence application in one day, a national laboratory has signed an agreement on AI for licensing, [2] and a news report relays an NRC official’s remarks on faster reviews. [3] In the sources opened, evidence for AI in reactor design and operation is mostly stated aims; other demonstrations may exist that this Brief did not open. Drafting documents is a general use of AI, so how much it changes nuclear energy itself is still an open question.
Current state
A test case. In March 2026 DOE reported that, with Idaho National Laboratory (INL), Argonne National Laboratory, Microsoft and the company Everstar, it had used AI to convert a safety analysis for a generic high-temperature gas reactor into sections equivalent to a licence application to the U.S. Nuclear Regulatory Commission (NRC). The 208-page document took one day; DOE says a team typically needs four to six weeks. A subject-matter expert reviewed it. [4] A trade report adds that the work used open-source documents, that the safety analyses it started from were preliminary and incomplete, and that the expert described the result as “consistent with what would be expected of a Revision 0 document”, a first draft. A pilot using approved safety documents is planned for a later phase. [5]
The companies, in their own words. Everstar describes “Gordian” as the “AI backbone for nuclear work”, and states that it “builds the intelligence to decouple rigor from speed”. [6] A trade report describes Gordian as using retrieval of source documents so that its output is cited. [5] Atomic Canyon is described by Oak Ridge National Laboratory as having developed AI models for the nuclear industry, called FERMI, which power its “Neutron AI” platform and enable search across technical documents. [2]
A laboratory agreement. In July 2025 Oak Ridge National Laboratory and Atomic Canyon signed a memorandum of understanding to use high-performance computing for simulation and AI to automate aspects of licence application reviews. An agreement states intentions; it is not a result. [2]
The regulator. In June 2026 an NRC official said at a conference that a type of licensing review that used to take four years had been done in nine months in its first round, against a target of 18 months. The news report adds that not all of this is strictly due to AI, because a May 2025 executive order set an 18-month deadline. The official also said that the NRC uses AI to help draft documents and is testing several AI tools on public data. [3]
Federal aims for design and operation. DOE’s Genesis Mission opportunity, issued in March 2026, includes a challenge on nuclear energy that states an aim of using AI to design, license, manufacture, construct and operate reactors, “enabling at least 2x schedule acceleration and greater than 50 percent operational cost reductions”, through models, digital twins and autonomous laboratories. These are stated targets for funded work. [1]
How far it has come. A proof-of-concept demonstration, not routine use in applications. The regulator describes AI use inside its own review work; the sources opened do not show a licence granted on an application drafted with AI. [4][3]
Opportunities
DOE cites a study by its National Reactor Innovation Center that found AI has the potential to reduce both document development time and regulatory review cycles by as much as 50 percent. [4] If so, shorter paperwork steps could shorten the path from design to a licence, which matters because the Genesis call links nuclear timelines to demand from AI data centers. [1] Next steps listed by DOE’s office include a review tool to document and resolve discrepancies, a formal NRC working group on AI-assisted licensing, and extending the approach to more safety report chapters and to fuel facility licensing. [5]
Challenges
- Regulation. Licences are decided by the NRC. The sources opened give no NRC acceptance of an AI-drafted application, and the news report separates AI from the executive order’s deadline. [3]
- Engineering and evidence. A DOE spokesperson said the tool showed areas for refinement in telling a parameter’s safety limits from its operating range, a distinction that matters in licensing. The demonstration also used open-source, preliminary documents, and the sources opened do not report errors found in a full review. [5]
- Scope. AI for drafting and searching documents helps the process around a reactor more than the reactor itself; for design and operation the sources opened give aims rather than results. [1]
Data sources
- NRC dockets and ADAMS for licensing applications, review schedules and decisions. AI use is not a coded field. The NRC site refused scripted access in this run and was not opened. [7]
- DOE Office of Nuclear Energy and national-laboratory announcements for demonstrations and agreements. [4][2]
- DOE funding opportunities and award lists, including the Genesis Mission, for federal programs. [1]
- USAspending.gov for federal awards. [8]
- Company announcements, as company claims. No registry lists AI licensing tools or their users.
Changelog
- 2026-11-18: First published.
Sources
- US Department of Energy, “The Genesis Mission: Transforming Science and Energy with AI”, Notice of Request for Application DE-FOA-0003612, issued 2026-03-17, Amendment 000003 of 2026-04-20; challenge 4. science.osti.gov (accessed 2026-11-18)
- Oak Ridge National Laboratory, “Oak Ridge National Laboratory, Atomic Canyon to accelerate nuclear licensing with AI”, 2025-07-22. ornl.gov (accessed 2026-11-18)
- Edward Graham, “AI has helped slash nuclear licensing review times, NRC official says”, Nextgov/FCW, 2026-06-25 (reports remarks by an NRC official). nextgov.com (accessed 2026-11-18)
- US Department of Energy, Office of Nuclear Energy, “Department of Energy Unleashes AI to Reduce Reactor Licensing Timelines”, 2026-03-26. energy.gov (accessed 2026-11-18)
- “The DOE’s plan for AI in NRC licensing”, Nuclear News (American Nuclear Society), 2026-04-02 (includes statements by a DOE Office of Nuclear Energy spokesperson). ans.org (accessed 2026-11-18)
- Everstar, company home page (company claims). everstar.ai (accessed 2026-11-18)
- US Nuclear Regulatory Commission, ADAMS. nrc.gov (listed, not opened: HTTP 403 to a scripted request on 2026-11-18)
- USAspending.gov, home page. usaspending.gov (accessed 2026-11-18)