Microsoft and Nvidia claim AI can speed approval of new atomic plants
Effort includes permitting and planning
Microsoft is working with Nvidia on nuclear power. Not to build it, but to offer AI-driven tools to deal with all the red tape, help with the design work, and optimize operations for nuclear projects.
Announcing the move on social media site X, Microsoft President Brad Smith said this latest AI collaboration covers "the full lifecycle from permitting and design to construction and operations."
Nuclear has long held promise as a stable source of carbon-free power, Smith added, and the aim of this particular initiative is to bring more of it online sooner.
While building a nuclear plant is a highly complex operation, designing one and navigating the Byzantine regulations governing atomic energy can take years, cost hundreds of millions of dollars, and involve an immense amount of data processing and reporting, Microsoft claims.
Of course, the Trump administration is seeking to solve this in a different way: by gutting the safety rules and skipping full environmental reviews for new reactors.
AI, we're told, is expected to help by making highly complex work repeatable and predictable, and slashing development timelines without sacrificing safety. The system ensures there is a "paper trail," so regulators can verify everything, while each engineering decision is digitally linked to the evidence and regulations for auditing.
Microsoft claims its tool, Generative AI for Permitting, reduced the time-intensive legwork to get approvals by 92 percent for Aalo Atomics, a firm working on mass-production of modular atomic reactors.
Another, Southern Nuclear, has developed and deployed agents using Microsoft's Copilot to improve consistency in engineering and licensing.
For design and engineering, digital twin technology and high-fidelity simulations enable faster iteration, while generative AI handles the drudge work of documentation drafting.
Nvidia isn't just aiming at atomic plants with its digital twin technology; it also unveiled Omniverse DSX last year, a blueprint for designing and operating gigawatt-scale AI datacenters – one of the reasons why there is a pressing need for more nuclear power in the first place.
- SoftBank to build massive AI datacenter on former US nuclear weapons site
- Google plugs AI into nuclear reactor biz – what could possibly go wrong?
- US is moving ahead with colocated nukes and datacenters
- Amazon, Meta, Google sign pledge to triple nuclear power capacity by 2050
More atomic power is seen as a solution for the burgeoning energy demands caused by the AI-driven datacenter building boom, but a nuclear plant typically takes at least five years to construct, while AI's energy demands mean more power is needed now.
Microsoft has itself invested in nuke-generated electricity, and has a 20-year power purchase agreement (PPA) with Constellation Energy to restart the infamous Three Mile Island nuclear facility, but even this is not expected to be online before 2028.
On the back of all the server farms it is building, Microsoft has seen its greenhouse gas emissions heading in the wrong direction. As The Register reported previously, its emissions have risen by nearly 30 percent since 2020, despite its much-publicized goal of becoming carbon-negative by 2030.
The new AI for nuclear operations initiative brings together Nvidia's Omniverse, Earth 2, CUDA-X, AI Enterprise, PhysicsNeMo, Isaac Sim, and Metropolis, with Microsoft's Generative AI for Permitting and Planetary Computer. The combination represents an AI-powered digital ecosystem for nuclear energy on Azure, Microsoft said. ®
