AtkinsRéalis on Nuclear Powered AI Factories and the Future of Infrastructure

Why It Matters

Nuclear power offers a secure, carbon‑free energy source that can keep AI data centres running continuously, addressing a critical bottleneck for hyperscalers and boosting ESG credentials. This model could reshape global data‑center siting and investment strategies.

Key Takeaways

• AtkinsRéalis partners with Nvidia to build nuclear‑powered AI factories
• CANDU reactors offer 24/7 low‑carbon power for data centers
• On‑site nuclear increases AI campus footprint by 20‑30%
• Digital twins in Nvidia Omniverse accelerate nuclear permitting and design
• Industry expects nuclear AI factories to become standard model

Pulse Analysis

The surge in artificial‑intelligence workloads has outpaced the power capacity of traditional data‑center hubs, prompting developers to look beyond fossil fuels. By marrying Canada’s proven CANDU reactor platform with Nvidia’s Omniverse DSX Blueprint, AtkinsRéalis is positioning nuclear energy as a scalable, baseload solution for AI compute. The collaboration promises 1 GW‑class AI campuses that can deliver uninterrupted, low‑carbon electricity, while leveraging closed‑loop water cooling and waste‑heat recovery to cut operating costs. This approach directly tackles the volatility of fossil‑fuel pricing and the emissions pressures facing hyperscalers.

From an engineering perspective, on‑site nuclear generation reshapes campus design. A 1 GW AI factory, according to Nvidia’s reference model, expands land use by roughly 20‑30%, but the added footprint enables integrated power, cooling, and grid infrastructure. CANDU’s online refuelling and use of unenriched uranium provide a reliable, long‑term energy supply, while digital twins built in Omniverse simulate construction, operation, and decommissioning phases. AI‑driven risk analytics further compress permitting timelines, delivering projects faster and with fewer regulatory hurdles.

The broader market impact could be profound. As investors prioritize ESG outcomes, nuclear‑backed AI facilities offer a compelling narrative of carbon neutrality and cost certainty. Geographic diversification of AI infrastructure—moving beyond the United States toward regions with existing nuclear expertise—may accelerate sovereign AI initiatives and reduce latency for end‑users. If early pilots succeed, the industry could see a shift toward co‑located nuclear‑data‑center complexes, redefining the economics of AI compute and setting a new benchmark for sustainable, high‑performance infrastructure.