RTS Nuclear + AI Briefing · August 2026
Three developments in summer 2026 moved the nuclear AI story from theme to specifics. The Department of Energy put real money behind AI on the critical path. The NRC said its own AI tools cut a four-year review to nine months. And a shelved small modular reactor came back, aimed at data center and campus loads. Funding, regulation, and demand are lining up at the same time, and the announcements are running ahead of the frameworks meant to govern them.
AI gets a budget line: DOE backs Project Prometheus with $60M and 32 partners
The DOE awarded Project Prometheus a $60 million Phase II grant over three years, matched by more than $200 million in industry cost sharing. The consortium, announced July 24, is coordinated by Idaho National Laboratory with Oak Ridge, Argonne, and Sandia, and includes X-energy, TerraPower, Oklo, NVIDIA, AWS, Microsoft, GE Vernova, and Westinghouse. It targets a tenfold cut in design and licensing time, threefold faster manufacturing, and 50 percent lower deployment cost.
The RTS perspective
This is the clearest sign yet that AI in nuclear is a funded program, not a pilot. Watch the workflow claims, not the headline dollars. A tenfold cut in licensing time only holds if the outputs are traceable and defensible to a regulator. The harder problem is the audit trail that has to travel with every result.
Source: Neutron Bytes
The regulator picks up the tool: NRC says AI cut a four-year review to nine months
NRC Chief Data Officer Basia Sall said on June 25 that a licensing review that once took four years was completed in nine months, inside the 18-month target set by the May 2025 executive order on NRC reform. The agency built an internal tool, SimplifAI, on Azure OpenAI to draft regulatory documents and check them against precedent. The NRC is also testing Anthropic’s Claude, Azure OpenAI, and Google Gemini for limited use cases on public data.
The RTS perspective
The regulator adopting AI matters more than any single developer doing so. When the NRC drafts against its own precedent with a tool, the bar for applicants rises. Well-structured, traceable submissions clear faster, and weak ones stand out more than before. The durable advantage goes to organizations that treat the regulator’s AI as a known reviewer and prepare accordingly.
Source: Nextgov
A shelved SMR comes back: Applied Atomics revives BWXT’s mPower
Applied Atomics secured exclusive land-based rights to BWXT’s mPower small modular reactor on June 17 and will re-engage the NRC to resume design certification. The mPower is a 195 MWe integral pressurized water reactor with an integrated steam generator that removes primary coolant piping. Applied Atomics is targeting behind-the-meter and campus-scale deployments.
The RTS perspective
Reviving a design with years of NRC interaction behind it is a different bet than starting clean. The value depends on how much of that documentation record is usable today. Resuming certification is not the same as holding one, and the distance from an exclusive license to a construction permit is measured in years.
Source: GlobeNewswire
Research news
- INL and NVIDIA target Prometheus with generative AI and digital twins, validated on the NRAD and MARVEL research reactors, aiming to cut deployment schedules at least in half.
- Los Alamos used machine learning on neutron-star data to sharpen models of the strong nuclear force.
- Argonne published an overview of AI in reactor operations, covering anomaly detection and predictive maintenance.
What this signals
- Federal dollars now buy AI on the critical path, not around it.
- The regulator is using AI itself, which resets what a competitive submission looks like.
- Proven light-water SMRs are being pointed at data center and campus loads.
- Developers are integrating vertically to control cost and schedule.
Funding, regulation, and demand rarely line up this cleanly. The organizations that win will treat AI adoption as a licensing and documentation discipline, not a technology purchase. Speed without traceability does not clear the NRC. Traceability at speed is the whole game.
Talk to RTS
One conversation. No pitch deck. Schedule a roadmap session or browse RTS research.
Comments are closed