Big Tech turns to nuclear power as AI's electricity demand outpaces supply
Amazon and Google have signed 20-year nuclear contracts while data centre demand pushes up capacity costs for households, an Al Jazeera opinion piece argues.
At a glance
- Amazon and Google signed 20-year nuclear contracts, buying output from existing reactors rather than funding new ones.
- Data centre demand is pushing up PJM capacity costs, with about $29.4bn attributed to it across four auctions.
- Gulf states including Qatar, Abu Dhabi and Saudi Arabia are planning gigawatt-scale AI data centre capacity.
Amazon and Google have each signed 20-year contracts to buy nuclear electricity, deals that underscore how the artificial intelligence boom is straining power supplies, according to an opinion piece published by Al Jazeera.
Amazon has secured 690 megawatts from Constellation's Calvert Cliffs plant in Maryland, including 190 megawatts of additional generating capacity expected between 2030 and 2032. Google's agreement will bring 890 megawatts of additional nuclear capacity into the United States. Constellation is investing more than $4.3bn to upgrade 11 reactors across Illinois, Pennsylvania and New Jersey.
Neither deal involves building a new reactor. The companies are paying for more output from reactors that already exist, some of which are older than the companies themselves.
Nuclear's return
The agreements would have been hard to imagine a few years ago. Nuclear power fell out of favour after accidents at Three Mile Island in 1979, Chornobyl in 1986 and Fukushima in 2011. Germany phased out nuclear power, and Japan shut down its reactors before beginning to restart them.
Today, an undamaged reactor at Three Mile Island is being brought back into service, with Microsoft contracted to buy its output for 20 years. The US Department of Energy is providing up to $4.4bn in loans to restart three shuttered plants, aiming to restore about 2.25 gigawatts. Palisades in Michigan, closed in 2022, is contracted to generate again by March 2027, which would make it the first US commercial nuclear plant to restart after entering decommissioning.
Efficiency did not cut demand
The industry previously answered AI's energy problem with efficiency: better chips, smaller models and smarter cooling. Those gains did not reduce overall electricity demand. Even when DeepSeek showed a competitive model could be trained at a fraction of the expected computing cost, spending did not fall. Laboratories kept planning longer training runs, reasoning models, agents working continuously in the background, and video generation.
The article compares this to William Jevons' 1865 observation that more efficient steam engines made Britain burn more coal, not less. The International Energy Agency estimates global data centre electricity demand will more than double to around 945 terawatt-hours by 2030.
New large reactors take a decade or more to deliver power, and small modular reactors are mostly not expected to produce electricity until the early 2030s, with very few built anywhere so far. Restarts and upgrades are the quicker option, but that still means 2027, 2028 or 2029, while AI capital spending moves on a far shorter cycle.
Who pays
Most of these deals sit inside PJM, the grid operator serving roughly 67 million people across 13 US states and Washington, DC. Its capacity auctions pass costs to consumers, and prices have reached the regulatory cap in three consecutive auctions. PJM's independent market monitor attributes about $29.4bn in capacity costs across the four most recent auctions to data centre demand, including roughly $6.3bn in the latest auction for facilities that have not yet been built.
The Natural Resources Defense Council estimates the typical PJM household could face an additional $70 a month by 2028 if current trends continue. Virginia has created a separate electricity rate class for data centres, Ohio has introduced minimum billing requirements and Illinois is debating legislation.
The article notes that in October 2026 US federal regulators rejected most of PJM's proposal to require large new electricity consumers to provide their own generation capacity, leaving such commitments voluntary.
For Gulf states, the piece says the US experience offers a warning. Qatar's sovereign fund has committed to a $20bn joint venture with Brookfield for AI infrastructure, starting at about 300 megawatts and potentially reaching near 1.5 gigawatts. Abu Dhabi is pursuing up to five gigawatts of AI data centre capacity, and Saudi Arabia is targeting almost two gigawatts through HUMAIN.
Source: Al Jazeera.
Questions and answers
- Why are tech companies buying nuclear power for AI data centres?
- AI data centres need large amounts of electricity, and efficiency gains have not reduced overall demand. Companies are signing long-term contracts to secure power from reactors that already exist.
- Who pays for the extra electricity capacity that data centres require?
- Under PJM's capacity auctions, costs are passed to consumers through electricity bills. The Natural Resources Defense Council estimates the typical PJM household could pay an extra $70 a month by 2028 if current trends continue.
- Which Gulf countries are investing in AI data centre capacity?
- Qatar's sovereign fund committed to a $20bn joint venture with Brookfield, Abu Dhabi is pursuing up to five gigawatts, and Saudi Arabia is targeting almost two gigawatts through HUMAIN.
