AtkinsRéalis, NVIDIA partner to design nuclear-powered AI data centres

Key Takeaways:

• AtkinsRéalis and NVIDIA are partnering to design nuclear-powered AI data centres, combining AtkinsRéalis’ nuclear engineering expertise with NVIDIA’s accelerated computing and digital twin technology.
• The collaboration is driven by surging global demand for AI infrastructure that is outpacing power supply, with nuclear energy emerging as a preferred low-carbon baseload solution for large-scale facilities.
• The two companies will focus on three areas together: integrating nuclear power through the CANDU portfolio, applying AI and simulation tools to project delivery, and engineering the power distribution, cooling, and modular construction systems these facilities require.

The Whole Story:

AtkinsRéalis Group Inc. is collaborating with NVIDIA to explore integrating NVIDIA’s technologies—including Omniverse libraries and accelerated compute—into the development and deployment of nuclear-powered, large-scale AI data centres, the company said in a news release Thursday.

The partnership will focus on designing AI-ready facilities that leverage AtkinsRéalis’ expertise in nuclear power infrastructure and complex project delivery alongside NVIDIA’s accelerated computing capabilities. The companies will use digital twin technology and OpenUSD frameworks to design and optimize integrated infrastructure systems before physical construction.

The initiative comes as global demand for AI infrastructure outpaces available power supply, prompting governments and developers to pursue gigawatt-class data centre solutions. Nuclear power is emerging as a leading low-carbon baseload option for AI facilities.

As the original equipment manufacturer and exclusive licence holder of CANDU technology, AtkinsRéalis is positioned to integrate standardized nuclear power solutions with large-scale AI data centre requirements. NVIDIA will contribute strategic leadership in compute, networking and AI, including digital twin and high-fidelity simulation tools.

The collaboration will examine three key areas: nuclear integration within AtkinsRéalis’ CANDU portfolio; accelerated computing and new project delivery methods using NVIDIA Omniverse and agentic AI; and power and data centre engineering, including power distribution, cooling systems and modular construction efficiency.

“AtkinsRéalis brings deep engineering and delivery expertise across complex infrastructure and a 70-year legacy of excellence in the nuclear industry,” said Ian L. Edwards, President and CEO, AtkinsRéalis. “This collaboration enables us to leverage these strengths in energy, infrastructure and complex project delivery to complement NVIDIA’s leadership in accelerated computing to help power critical AI data centres.”

Vladimir Troy, Vice President, AI Infrastructure at NVIDIA, said: “Energy is essential to the AI industrial revolution. Pairing AtkinsRéalis’ deep engineering and nuclear expertise with the NVIDIA DSX Blueprint can help scale the deployment of efficient next-generation AI factories with accelerated computing and digital twins.”

CANDU — short for Canada Deuterium Uranium — is a pressurized heavy-water reactor design that has been generating electricity in Canada for more than 60 years. Development of the technology began in the late 1950s, with the first commercial reactors coming online at Pickering, Ontario, in 1971. Today, the technology powers 19 reactors across Ontario and has become a cornerstone of the province’s electricity grid, currently supplying roughly 30% of Ontario’s power through the Bruce Power facility alone.

Its adoption also enabled Ontario to phase out coal entirely by 2014, which remains the single largest greenhouse gas emissions reduction initiative in North American history. Beyond Canada, CANDU technology has been exported to South Korea, Romania, India, Pakistan, Argentina and China, with 27 CANDU power reactors now operating across seven countries.