Nuclear energy should be expanded globally to meet increasing AI-driven energy demands and climate objectives

What's this about?

People disagree about whether the world should build more nuclear power.

The goal is to meet AI’s rising power needs while cutting heat-trapping gases.

What supporters say

• AI may soon need much more power for data centers, which run AI tools.
• Nuclear plants make steady power with few heat-trapping gases, while coal and gas make much more.
• Nuclear power can help when wind and solar make less power because of night or bad weather.
• Nuclear power causes fewer deaths than coal and gas, mainly because those fuels dirty the air.

What critics say

• New nuclear plants can take many years to plan, build, and start.
• Nuclear plants cost a lot to build, so other clean choices may cut heat-trapping gases for less money.
• Experts have not shown that the world needs nuclear power to meet AI’s needs or climate goals.
• Nuclear waste needs safe care for a very long time, and strong rules must keep people safe.

How to read this

The number of points on each side does not show who is right; the strength of each point matters more.

The bottom line

Nuclear power could help meet new power needs and cut heat-trapping gases.

But the evidence does not show that the world must expand nuclear power everywhere.

The claim is that the world should expand nuclear power to meet rising electricity demand from artificial intelligence while cutting greenhouse-gas emissions. The evidence supports nuclear as a useful option, but does not show that a global expansion is essential.

The case for

Nuclear power provides reliable electricity with low lifecycle greenhouse-gas emissions, making it one of the low-carbon tools identified by the IPCC and the International Energy Agency. Unlike wind and solar, reactors can produce power continuously, helping replace coal and gas while electricity demand grows.1

That demand is becoming more important. The IEA projects substantial growth in data-center electricity use through 2030, with AI-focused facilities among the fastest-growing sources. U.S. Department of Energy analysis also points to rapid potential growth and new challenges for grid planning.2 In fossil-fuel-heavy regions, nuclear could meet some of this demand without requiring more coal or gas.

Nuclear may also help balance systems built around weather-dependent renewables. Wind and solar vary by time of day and weather, creating a need for storage, stronger transmission networks, flexible demand or other dependable power sources. The IEA treats nuclear as one possible complement in such systems, rather than a replacement for renewables (see Figure 2).3

Another supporting point is public health. Studies generally find fewer deaths per unit of electricity from nuclear power than from coal and gas, largely because fossil fuels create major air-pollution harms (see Figure 3).4 This strengthens the case for nuclear where the alternative is continued fossil-fuel generation, though it does not prove nuclear is better than every combination of renewables, efficiency, storage and grid investment.

The case against

The central objection is that nuclear has not been shown to be necessary. IPCC pathways and IEA analyses describe several possible low-emissions portfolios, combining renewables, efficiency, storage, transmission, demand management and, in some cases, nuclear. Rising AI demand demonstrates the need for credible clean-power planning, but not that one particular technology is indispensable.5

New reactors may also arrive too late for near-term AI growth. The IEA identifies shortages of skilled workers, manufacturing capacity, financing and fuel-supply capability. Research has also linked delays to regulatory complexity, design changes, supply-chain problems and limited specialist expertise.6 If data-center demand grows before reactors are completed, other measures may be needed in the meantime.

Cost is another major uncertainty. Nuclear projects are highly sensitive to financing costs, construction conditions, regulation and delays. The IEA, the OECD Nuclear Energy Agency and an MIT study all find that these factors can substantially change the price of nuclear electricity. In places where renewables, efficiency, storage or grid upgrades can be built more quickly and cheaply, directing scarce climate funding toward new reactors may be less effective.7

Radioactive waste creates a long-term governance responsibility. The IAEA says established containment and disposal methods can manage the problem, but high-level waste still requires lasting technical, political and institutional commitments.8 AI demand forecasts also remain uncertain: adoption, more efficient models, hardware use, facility locations and grid-management choices could all change the scale and timing of electricity needs.

The bottom line

The evidence favours selective nuclear expansion as part of a varied low-carbon strategy, but only weakly supports the stronger claim that global nuclear expansion is necessary. Nuclear’s low-carbon and dependable output are well supported, as is the expectation that data-center demand will grow. But the comparative evidence does not establish that nuclear will consistently beat renewables, storage, efficiency, transmission or demand management on speed, cost and reliability.

Nuclear is most defensible where fossil fuels dominate, reliable clean power is scarce, financing is workable and projects can be completed on schedule. Regions with rapidly deployable renewables, strong grids or high nuclear construction risks may cut emissions and serve new demand sooner through other approaches. The prudent policy is therefore to keep nuclear available where it is competitive, while rejecting a universal requirement unsupported by current evidence.