Listen to the article
By Sean Rayment
A leading US tech company is developing AI-driven software to accelerate the construction of nuclear power reactors.
Palantir and The Nuclear Company, also US-based, will jointly create the nuclear operating system, which will simplify construction, allowing nuclear plants to be built faster and at lower cost.
Palantir said the AI software will help build plants more cheaply, safely, and quickly, with the goal of outpacing China.
The US has not built a new nuclear plant in 30 years, but The Nuclear Company, with Palantir’s software, aims to deliver electricity by the 2030s and meet President Donald Trump’s goal of 400 GW of nuclear power by 2050, up from around 97 GW today.
The US needs to increase its nuclear power to compete with China, according to Mike Gallagher, head of defence at Palantir Technologies Inc.
The collaboration aims to transform nuclear construction into a data-driven, predictable process, addressing the industry’s persistent challenges of cost overruns and scheduling delays.
The Nuclear Company will pay approximately $100m (£72m) over five years for the development of the platform, according to a Palantir spokesperson.
The partnership will involve a dedicated engineering team working alongside The Nuclear Company’s construction and engineering staff.
“The future of energy security and sovereignty will be shaped by our ability to deploy advanced technologies at scale,” said Mike Gallagher, head of defence at Palantir Technologies. “This partnership marks the first time Palantir’s software will be used to help power the next generation of nuclear energy infrastructure.”
The nuclear operating system will provide schedule certainty and cost savings via comprehensive supply chain tracking, problem prevention using sensors and digital twin technology, and regulatory confidence through AI-powered document review and validation.
According to International Atomic Energy Agency data, of the 62 nuclear plants currently under construction around the world, 29 are in China.
In the last 30 years, the US has built only two plants — the Vogtle-3 and -4 units in Georgia — both of which saw cost overruns and delays due to supply chain issues, shortages of skilled workers, startup problems, and the Covid-19 pandemic.
When Georgia approved the Vogtle project in 2009, the two reactors were expected to cost about $14bn and enter service in 2016 and 2017. However, along with the delays, the costs ballooned to $30bn, and the units eventually began commercial operation in 2023 and 2024.
In May, Trump signed executive orders calling for 10 large-scale nuclear reactors to be under construction by 2030 and for 400 GW of total nuclear power capacity by 2050 — over four times the country’s current output.
Earlier this week, New York Governor Kathy Hochul announced plans to build a nuclear power plant in New York.
Meanwhile, Siemens Digital has recently published a white paper examining the crucial role of high-fidelity digital twins in advancing the development of Generation IV nuclear reactors.
The report stated that digital twins must be precise replicas of the reactor to effectively replace costly demonstration models in satisfying regulatory requirements in the US.
In collaboration with Dutch researchers, Siemens’ digital twin software was rigorously tested and demonstrated exceptional accuracy in predicting the behaviour of complex systems.
The report concluded that the digital twin — a comprehensive virtual model capable of forecasting reactor performance throughout design, licensing, and operation — is essential for delivering next-generation nuclear reactors in time to combat climate change.
It added that developing digital models is considerably more cost-effective than constructing expensive physical demonstrators, even though further work is needed to fully capture all aspects of reactor performance.
