Main-Nuke/ai-and-nuclear-energy-summary.md

# AI and Nuclear Energy: Current State of Convergence

*Compiled September 2, 2026*

Based on the IAEA's "The Atom and the Algorithm" statement and additional reporting, this summarizes where AI and nuclear energy currently intersect.

## The core idea

IAEA Director General Rafael Mariano Grossi frames this as a bidirectional "structural alliance" he calls **"Atoms for Algorithms"**: nuclear power is emerging as the leading option to feed AI's massive electricity demand, while AI is simultaneously being used to make nuclear technology work better.

In nuclear power operations, AI supports predictive maintenance, anomaly detection, and optimization of thermal performance. In design, it supports accelerated reactor modelling, fuel-cycle simulation, and materials development. In safety, AI supports accident simulation and emergency-procedure development, and in safeguards it helps analyze surveillance footage and satellite imagery for pattern recognition — work the IAEA says is already happening in its labs and with member states.

This statement was tied to the IAEA's first International Symposium on AI and Nuclear Energy, where the IAEA also signed an agreement with Atomic Canyon, a company building AI tools for the nuclear sector, to explore AI solutions for nuclear information management.

## 1. Nuclear powering AI (data centers)

Big Tech has been signing large nuclear power-purchase deals to secure reliable, carbon-free electricity for data centers:

- **Microsoft** — A 20-year, $16 billion deal to buy all the power from the restarted Three Mile Island Unit 1 (now the Crane Clean Energy Center), expected online in late 2027.
- **Google** — Ordered up to 500 megawatts of small modular reactors (SMRs) from Kairos Power.
- **Amazon** — A 1.9-gigawatt deal through 2042 with Talen Energy's Susquehanna nuclear plant, part of a broader $20 billion AWS investment in Pennsylvania.
- **Meta** — A 20-year power purchase agreement with Constellation Energy for 1.1 gigawatts from the Clinton Clean Energy Center in Illinois, plus a separate RFP targeting 1–4 GW of new nuclear generation.

By one industry tracker's count, hyperscaler nuclear deals now total roughly 9.8 GW of committed capacity — more nuclear capital commitment than any prior decade in US history.

**Counterpoint:** Not everyone is convinced this is substantive. The Bulletin of the Atomic Scientists argues that Amazon's and Google's nuclear commitments are a small fraction of their overall data-center capex, and reads much of the messaging as greenwashing.

## 2. AI improving nuclear plant operations

- **Diablo Canyon (PG&E)** — AI is streamlining document search and retrieval at the plant, marking the first commercial use of generative AI at a US nuclear power plant, via a partnership with **Atomic Canyon**. Its FERMI family of AI models is an open-source embeddings system built with Oak Ridge National Laboratory specifically for nuclear data, trained on tens of millions of pages of documentation. Atomic Canyon also built NIVA (Nuclear Industry Virtual Assistant) with INPO, EPRI, and NEI, now available across the North American commercial nuclear fleet.
- **Westinghouse + Google Cloud** — A custom AI platform combining Google and Westinghouse models with Westinghouse's WNEXUS 3D digital twin, aimed at the reactor construction process (historically ~60% of reactor costs) to predict bottlenecks and optimize scheduling.
- **Microsoft + NVIDIA + Aalo Atomics** — Aalo Atomics says it has cut permitting time by 92% using a Microsoft generative-AI permitting tool, saving an estimated $80 million a year, as part of a broader collaboration spanning NVIDIA Omniverse, digital twins, and AI-assisted licensing workflows.
- **Argonne National Laboratory** — Developed the PRO-AID tool, using generative AI/LLMs for real-time reactor monitoring and diagnostics, and is also applying AI to plant performance data for more accurate nuclear fuel reload planning.
- **DOE-wide funding** — The Department of Energy has committed nearly $900 million to programs like Light Water Reactor Sustainability, which funds explainable AI for predictive maintenance across the aging US reactor fleet.

## 3. IAEA's own institutional role

Beyond the symposium, the IAEA runs an "AI for Atoms" platform and cooperates with the ITU and nearly 40 UN organizations on AI-and-nuclear issues, including a Working Group on AI for Safeguards Verification focused on spent-fuel verification and video surveillance analysis.

## A caveat

This is a fast-moving space with a lot of PR-driven announcements on both sides. Figures like "9.8 GW committed" or specific savings percentages should be read as industry claims rather than independently audited numbers.

---

## Sources

1. [IAEA — "The Atom and the Algorithm: Nuclear Energy and AI are Converging to Shape the Future"](https://www.iaea.org/newscenter/statements/the-atom-and-the-algorithm-nuclear-energy-and-ai-are-converging-to-shape-the-future) — original statement
2. [IAEA — "IAEA Hosts First International Symposium on AI and Nuclear Energy"](https://www.iaea.org/newscenter/news/iaea-hosts-first-international-symposium-on-ai-and-nuclear-energy)
3. [Balkan Green Energy News — symposium coverage](https://balkangreenenergynews.com/iaea-hosts-landmark-symposium-nuclear-ai-to-forge-atoms-for-algorithms-alliance/)
4. [404 Media — critical coverage of DOE's AI-nuclear vision](https://www.404media.co/nuclear-rian-bahran-iaea-international-symposium-on-artificial-intelligence/)
5. [ITU AI for Good — IAEA's "AI for Atoms" platform](https://aiforgood.itu.int/about-us/un-ai-actions/iaea/)
6. [Introl — nuclear/AI data center deal tracker](https://introl.com/blog/nuclear-power-ai-data-centers-microsoft-google-amazon-2025)
7. [Bulletin of the Atomic Scientists — critical take on Big Tech nuclear PR](https://thebulletin.org/2026/07/data-centers-powered-by-next-gen-nuclear-dont-fall-for-big-techs-pr-hype/)
8. [Trellis — "Amazon, Google, Meta and Microsoft go nuclear"](https://trellis.net/article/amazon-google-meta-and-microsoft-go-nuclear/)
9. [SMR Intel — nuclear data center deal tracker](https://smrintel.com/nuclear-data-center-deals/)
10. [Forbes — "The AI Boom Is Making Nuclear Power Bankable Again"](https://www.forbes.com/sites/kensilverstein/2026/07/26/the-ai-boom-is-making-nuclear-power-bankable-again/)
11. [American Nuclear Society (ANS) / Nuclear Newswire — AI in nuclear plant operations](https://www.ans.org/news/2025-01-09/article-6677/article-considers-incorporation-of-ai-into-nuclear-power-plant-operations/)
12. [Utility Dive — "America's AI future hinges on nuclear energy"](https://www.utilitydive.com/news/americas-ai-future-hinges-on-nuclear-energy/747395/)
13. [Microsoft Cloud Blog — AI for nuclear energy / Aalo Atomics collaboration](https://www.microsoft.com/en-us/microsoft-cloud/blog/energy-and-resources/2026/03/24/ai-for-nuclear-energy-powering-an-intelligent-resilient-future/)
14. [Westinghouse — "How Westinghouse is reenergizing nuclear power with — and for — AI"](https://info.westinghousenuclear.com/blog/how-westinghouse-is-reenergizing-nuclear-power-with-and-for-ai)
15. [Atomic Canyon — Solutions page](https://atomic-canyon.com/solutions)
16. [Atomic Canyon — homepage](https://www.atomic-canyon.com/)
17. [ACM (Communications of the ACM) — "Nuclear Power for AI; AI for Nuclear Power?"](https://cacm.acm.org/news/nuclear-power-for-ai-ai-for-nuclear-power)