Rethinking nuclear
Grid, AI, industry, and security

Batteries and demand response can manage intraday volatility. What they cannot do – and what no grid-scale storage technology can yet do – is replace wind and solar for days at a time. Interconnection helps: for example, price spikes pull in hydro from Norway and nuclear from France. But interconnector capacity is finite and regional weather systems often cover vast areas at once. If your storage challenge is seasonal rather than diurnal, a reasonable proportion of firm low-carbon capacity is worth having. That logic is one of the reasons why Sweden, Belgium, Denmark and the Netherlands have reopened the nuclear question.

Critics point to the levelised cost of energy (LCOE) comparisons in which nuclear can look expensive. But LCOE measures generation cost in isolation. It puts no price on the system-wide cost of intermittency, transmission build-out and seasonal backup that a renewables-only grid requires. Once those are counted, nuclear’s value in high-renewable grids shifts materially. The UK’s value for money assessment for Sizewell C put the savings at £2 billion a year from reduced network, interconnector and balancing costs against a renewables-led alternative.

The US Department of Energy reached a similar conclusion for California: including nuclear lowered generation and transmission costs by around 37% compared with renewables and storage alone. Even California has winter periods where wind and solar output falls to 30-40% of maximum summer production.

While traditional reactors are not designed for flexibility, they are capable of it. EDF in France has in recent years used its reactors’ ability to vary output to help manage the impact of renewables. Average daily swings in the French nuclear fleet’s generation have grown from 1.5 GW in 2022 to 6 GW in 2025.

In the US alone, big tech has contracted well over 10 GW of nuclear capacity since late 2024. Microsoft has signed a 20-year power purchase agreement with Constellation Energy to restart Unit 1 at Three Mile Island. Google has a 500 MW development agreement with Kairos Power. Amazon is putting over US$20 billion into two Pennsylvania data centre campuses – including one next to the Susquehanna nuclear plant –  and has anchored a US$700 million funding round into SMR developer X-energy. Meta has agreements with Constellation, Vistra, Oklo and TerraPower for up to 7.7 GW of existing and new capacity by 2035. Oracle has announced plans for a gigawatt-scale, SMR-powered data centre.

Nuclear is the obvious solution. AI training and inference need constant firm power. US reactors run at capacity factors above 90% and nuclear offers exceptional energy density, using roughly one to two orders of magnitude less land per MWh than utility-scale solar or onshore wind. Meanwhile, co-location behind the meter can remove transmission constraints; ease grid-connection queues; deliver the frequency and voltage stability that sensitive GPU clusters require; and shield operators from wholesale price volatility.

Timing is key. AI demand is likely to be met by utility-scale solar/battery storage and natural gas peaking plants in the near term, natural gas combined cycle in the medium term, and nuclear in the longer term. The "long term" in this context means going operational in the 2030s, meaning we are in a decisive stage for nuclear pipelines. Recent US reform has been explicitly driven by this imperative. The Biden administration’s January 2025 executive order backed AI infrastructure on federal lands with new clean energy including nuclear, and the ADVANCE Act of 2024 updated licensing processes to accelerate advanced reactor deployment.

Unlike renewables, on nuclear the former and current administrations are moving in the same direction. President Trump’s second-term executive orders aim to expand US nuclear capacity to approximately 400 GW by 2050, with streamlined permitting for advanced reactors co-located with AI facilities on federal sites. Lastly, under the White House’s Ratepayer Protection Pledge of March 2026, major technology companies committed to funding the power generation and grid infrastructure their data centres require.

How can nuclear decarbonise industry?

Steel, cement and chemicals are among the largest industrial sources of CO₂; between them they account for close to a fifth of global emissions. Three overlapping problems make them hard to abate. Many of the processes need temperatures above 500°C, sometimes above 1,500°C, which electrification struggles to deliver technically and often cannot deliver economically. Fossil fuels often play a chemical role as well as a thermal one, most obviously coking coal as the reducing agent in primary steelmaking, where substitution needs green hydrogen at industrial scale and price. And cement’s emissions come mostly from the calcination of the limestone itself rather than the fuel burned to heat the kiln, so even fully decarbonised heat leaves roughly 60% of the CO₂ untouched without carbon capture, utilisation and storage (CCUS).

Advanced designs, particularly fourth generation high-temperature gas reactors (HTGRs) and molten-salt reactors (MSRs), offer one of the few low-carbon routes to industrial-scale high-temperature heat and clean hydrogen production. They can supply process heat directly to adjacent facilities and support efficient hydrogen production through high-temperature electrolysis or thermochemical cycles. Microreactors add something else again: transportable, reliable units that can sit alongside industrial sites, supplying firm process heat and power at remote or off-grid locations where a grid connection is impractical.

China already has a demonstration HTGR and an experimental MSR in operation.  Meanwhile, for countries such as France, the Netherlands, Japan, South Korea, the UAE and the UK, getting firm electricity, process heat and clean hydrogen from a single nuclear asset is increasingly attractive. This is also true for emerging industrial hubs, which must somehow reconcile heavy industry, export exposure and net-zero commitments.

What is the strategic need for nuclear power?

Russia’s full-scale invasion of Ukraine showed what dependence on a single gas supplier costs. Before the invasion, over 40% of Europe’s imported natural gas came from Russia. The price shock that followed pushed European wholesale gas to historic highs, triggered demand-reduction mandates across the EU, and forced governments into politically fraught choices between industrial rationing and a return to fossil fuels. The Atlantic Council, writing on the two-year anniversary of the invasion, drew the obvious lesson: Europe “must never again allow such dependence”.