AI data centers are driving a surprising energy solution: the revival of nuclear power plants that would otherwise remain economically unviable. Microsoft, Amazon, Google, and OpenAI are investing billions in nuclear infrastructure to power their AI operations, fundamentally reshaping how the tech industry approaches its massive electricity demands. This shift reveals a paradox at the heart of artificial intelligence's climate story: the same technology that consumes enormous amounts of energy is also creating financial incentives to build carbon-free power sources.

Why Are Tech Companies Turning to Nuclear Power?

The answer lies in a fundamental mismatch between how AI data centers operate and how renewable energy works. Solar panels and wind turbines are intermittent; the sun does not always shine, and the wind does not always blow. AI data centers, by contrast, need reliable electricity 24 hours a day, seven days a week. Battery storage at grid scale remains expensive and limited, leaving nuclear power as the most practical carbon-free alternative.

This is not altruism. Tech companies want reliable, carbon-free electricity that does not undermine their public renewable energy commitments. But the practical effect is that AI demand is creating financial incentives to build and restart nuclear plants that would otherwise remain economically marginal. Consider the concrete examples already underway:

• Three Mile Island Restart: Microsoft and Constellation Energy signed an agreement to restart Three Mile Island Unit 1 in Pennsylvania to power Microsoft's AI data centers. The plant, shut down in 2019 for economic reasons, reopened in late 2024, marking the first US nuclear plant to be restarted rather than decommissioned.
• Small Modular Reactors: Amazon has invested in multiple Small Modular Reactor (SMR) companies, including X-energy, and signed agreements to purchase nuclear power from next-generation plants once operational.
• Molten-Salt Technology: Google announced agreements to purchase nuclear power from Kairos Power's molten-salt reactor, expected to come online in the 2030s.
• Private Investment: Sam Altman, OpenAI's CEO, has separately invested in Oklo, an SMR startup seeking regulatory approval for advanced fission reactors.

What Does This Mean for Climate Change?

Whether this nuclear expansion is genuinely good for the climate depends partly on your views about nuclear energy itself. The carbon lifecycle of nuclear power is among the lowest of any energy source, producing roughly 12 grams of CO2 per kilowatt-hour over its full lifecycle. That compares favorably to natural gas at roughly 490 grams per kilowatt-hour and coal at roughly 820 grams per kilowatt-hour.

If AI's insatiable power demand accelerates the transition to nuclear alongside renewables, that could represent a significant climate positive. However, construction timelines for Small Modular Reactors remain measured in decades, not years, meaning this solution addresses a future problem rather than today's energy crisis.

How to Evaluate AI's True Climate Impact?

Understanding whether AI is ultimately good or bad for the climate requires looking at both sides of the equation simultaneously. The energy footprint is real and substantial. Training a frontier AI model is a massive, one-time computation that runs around the clock for weeks or months. GPT-3, with 175 billion parameters, generated roughly 552 tonnes of CO2 equivalent during training, equivalent to about 330 round-trip flights from New York to London. GPT-4 produced similar emissions at roughly 500 tonnes CO2 equivalent. Llama 3 70B generated between 300 and 500 tonnes CO2 equivalent.

But training is only part of the story. Inference, running models at scale to answer billions of queries, is where most long-term emissions come from. A single ChatGPT-style query consumes roughly 0.001 to 0.01 kilowatt-hours of electricity, compared to roughly 0.0003 kilowatt-hours for a standard Google search. That puts a ChatGPT query at approximately 3 to 33 times the energy of a Google search, with roughly 10 times as the most commonly cited estimate. At over 100 million daily active users, those queries add up fast.

The International Energy Agency (IEA) projected that global data center electricity consumption could reach 1,000 terawatt-hours per year by 2026, roughly double 2022 levels, with AI workloads as the primary driver of that growth. To put that in perspective, 1,000 terawatt-hours is approximately 3 to 4 percent of total US electricity consumption, comparable to the entire electricity consumption of Japan in some years, and roughly five times the electricity consumed by all electric vehicles in the world.

Beyond electricity, data centers consume enormous amounts of water for cooling. Microsoft disclosed that its global data centers used 6.4 million cubic meters of water in 2022, up 34 percent from 2021. Google reported 5.6 billion gallons of water consumption in the same year. In water-stressed regions like the US Southwest, parts of India, and the Middle East, data center water consumption competes directly with agriculture and municipal use.

Is AI Actually Helping Solve Climate Problems?

The energy footprint debate can obscure the other half of the story: AI is already delivering meaningful climate benefits across several domains. Traditional weather forecasting runs physics simulations on supercomputers. A 10-day global forecast from the European Centre for Medium-Range Weather Forecasts (ECMWF) takes roughly 12 hours of compute time. Google DeepMind's GraphCast, trained on 40 years of weather data, produces the same forecast in under one minute, running on a single Google TPU v4 processor. In benchmarks published in Science in 2023, GraphCast matched or exceeded the accuracy of the best operational weather prediction systems for most metrics beyond day 7.

The practical benefits of faster weather forecasting are substantial. Faster hurricane track forecasting means more evacuation time. Better extreme heat event prediction 7 to 10 days out gives cities time to open cooling centers. Agricultural frost warnings with greater accuracy save crops. Renewable energy output forecasting helps grid operators balance supply and demand.

Wildfire prediction represents another critical application. Wildfire burned area has grown dramatically across the western US, Australia, Canada, and the Mediterranean. Traditional fire spread models run on high-performance computing clusters and can take hours to generate a forecast. AI models trained on satellite imagery, topography, fuel moisture data, and historical fire behavior can predict fire spread in near real-time, with outputs running in minutes rather than hours. This matters directly for evacuation timing; a difference of two hours in evacuation order timing can be the difference between orderly departure and gridlock.

The US Forest Service and several state fire agencies are piloting AI-assisted fire behavior prediction. Australia's Bureau of Meteorology began integrating machine learning-based fire risk models after catastrophic fire seasons, demonstrating that the technology is moving from research into operational use.

What Should We Take Away From This?

The story of AI and climate change is not a simple contradiction between energy consumption and climate solutions. It is the same story told from opposite ends. AI is simultaneously one of the fastest-growing sources of new energy demand on Earth and one of the most powerful tools humanity has ever developed for understanding and managing its climate. Ignoring either half of that sentence leads to bad conclusions. Cheerleaders who cite only the climate benefits ignore real emissions. Critics who cite only the energy cost ignore tools that could genuinely help avoid catastrophe.

The nuclear power revival driven by AI data centers represents a pragmatic, if imperfect, response to this tension. Whether it ultimately proves beneficial depends on whether the construction timelines accelerate, whether renewable energy continues to grow alongside nuclear capacity, and whether the climate benefits from AI-powered weather and fire prediction materialize at scale. For now, the outcome remains uncertain, but the stakes are undeniably high.