NuScale Power Corporation (NYSE: SMR) is deploying nuclear-specific AI across its advanced reactor programme, tying its NRC-certified NuScale Power Module (NPM) to Nuclearn's AtomAssist knowledge-management platform and NPX's nuclear-project delivery stack. The three-way partnership, announced on August 25, drew immediate market approval — NuScale shares jumped roughly 9.8% after the announcement, following an internal proof of concept that showed information-retrieval times cut by as much as 80%.
What the tools actually do
AtomAssist wraps nuclear-specific language models around NuScale's controlled engineering, licensing and standards content, preserving traceability from every AI-generated answer back to its source document. NPX's team layers on new-build project management, licensing coordination and vendor integration experience, and hardens the deployment discipline needed to bring the tools into safety-critical workflows. According to NuScale's press release, the goal is faster decisions without loosening safety, quality or regulatory rigour.
The economic argument for nuclear AI
"AI is only as valuable as the knowledge it can be trusted to deliver," said Jerrold Vincent, CFO and co-founder of Nuclearn. "In a highly regulated, safety-critical industry like nuclear, every answer must be grounded in controlled engineering and licensing information." NPX's Joshua Bell framed it as "not AI for AI's sake" but as a way to close the gap between promising SMR technology and the day-to-day project execution needed to actually build reactors.
Fits a wider SMR-AI wave
The deployment lands as the US nuclear industry commercialises SMRs on multiple fronts. NuScale has been busy with a fuel-cycle MOU with Curio and Framatome, the Paragon Highly Integrated Protection System contract, and the addition of former NRC chair Dale Klein to its board. In parallel, Holtec's four-unit Oyster Creek plan and the IAEA's new ATLAS maritime-SMR initiative show the sector's pace picking up. Atomic Canyon's new US nuclear-fleet assistant, and rising AI budgets among utilities, suggest AtomAssist-style tools are moving from pilots into production knowledge stacks.
What to watch
Two metrics will show whether nuclear-specific AI keeps its momentum: how quickly NRC and CNSC reviewers accept AI-assisted licensing evidence submissions, and how vendors like NuScale and Holtec cite AI-driven design or licensing throughput in their next earnings updates. If the 80% search-time reduction generalises across engineering teams, it will become a competitive edge in reactor delivery — one every SMR vendor now has to match.
Reporting based on coverage from NuScale, Seeking Alpha and World Nuclear News.