Data centers powered by artificial intelligence are consuming electricity so rapidly that utilities across the United States are postponing coal plant closures they once considered final decisions. The International Energy Agency (IEA) projects global data center electricity demand will more than double from 415 terawatt-hours in 2024 to 945 terawatt-hours by 2030, with AI-focused facilities growing roughly three times faster than the data center sector overall.

How Much Electricity Are AI Data Centers Actually Using?

The numbers reveal why grid operators are sounding alarms. A single ChatGPT query consumes roughly 2.9 watt-hours of electricity, compared to just 0.3 watt-hours for a typical Google search, meaning each AI interaction uses nearly ten times more power. The IEA found that global electricity demand from data centers grew 17% in a single year, while AI-focused data centers surged by 50% over the same period.

Training large language models requires clusters of thousands of specialized graphics processing units (GPUs) running continuously for weeks or months. Once trained, every single query, image generation, and automated workflow demands additional computing power called "inference." Unlike traditional web traffic that ebbs and flows predictably, AI inference demand compounds continuously as companies layer AI capabilities into search engines, productivity software, customer service systems, and industrial processes simultaneously.

Cooling systems add another significant burden. Modern AI chips generate enormous heat, and keeping server racks within safe operating temperatures can account for a substantial portion of a facility's total electricity draw. The industry has shifted toward liquid and immersion cooling, which is more efficient than traditional air cooling but still requires substantial water and power resources at scale.

Why Are Tech Companies Missing Their Climate Commitments?

Major technology companies made some of the most ambitious corporate climate pledges on record. Amazon committed to net-zero emissions by 2040. Google set a target of net-zero emissions by 2030. Microsoft pledged to be carbon negative by 2030. But the AI buildout is now straining those very commitments.

Microsoft disclosed that its total greenhouse gas emissions rose 25% year-over-year, driven primarily by data center construction and expanded electricity use. The company's Scope 2 emissions, those tied to purchased electricity, jumped from about 2% of its total footprint to 13% in a single fiscal year. Google reported a 25% increase in supply-chain-related emissions, and Amazon logged a 34% rise in Scope 2 emissions.

"AI infrastructure is driving demand for energy, water, land and materials, though sustainability solutions are not scaling fast enough to meet demand," stated Microsoft President Brad Smith and Chief Sustainability Officer Melanie Nakagawa in the company's 2026 Environmental Sustainability Report.

Brad Smith, President at Microsoft; Melanie Nakagawa, Chief Sustainability Officer at Microsoft

Part of this emissions rise reflects an accounting shift as much as a physical one. Microsoft stopped counting certain low-quality renewable energy certificates (RECs) toward its clean energy goals, a type of credit that doesn't necessarily fund new clean power generation. That change made the company's reported emissions look worse on paper, even as it pushed the industry toward more meaningful clean energy investment. Still, the underlying trend is real: electricity demand from AI data centers is growing faster than the supply of new clean power that utilities and developers can bring online.

How Are Data Centers Affecting Coal Plant Retirements?

Perhaps the most tangible way data centers are impacting clean energy goals is by keeping fossil fuel power plants running longer than planned. Across the United States, utilities facing surging data center demand have delayed or canceled dozens of coal plant retirements that were once considered settled decisions.

In Kansas and Missouri, utility Evergy has asked regulators for permission to delay retiring or converting roughly 2.8 gigawatts of coal capacity by at least five years, citing data center growth in the state. In Wyoming and Georgia, aging units at plants like the Robert W. Scherer Electric Generating Plant, one of the largest coal-fired facilities in North America, have had retirement dates pushed back.

Steps Utilities Are Taking to Address Data Center Demand

• Delaying Coal Retirements: Utilities are postponing planned closures of coal-fired power plants to maintain baseload capacity for data center growth, extending the operational life of aging fossil fuel infrastructure by five or more years.
• Investing in Renewable Energy: Companies like Microsoft and Google are pursuing direct contracts with renewable energy developers and exploring nuclear power partnerships to meet AI infrastructure electricity needs without relying on fossil fuels.
• Improving Cooling Efficiency: Data center operators are transitioning from traditional air cooling to liquid and immersion cooling systems that reduce overall electricity consumption while managing the extreme heat generated by AI chips.

The collision between AI infrastructure growth and climate commitments represents one of the defining energy challenges of 2026. Utilities must balance immediate data center demand with long-term decarbonization goals, while technology companies face pressure to deliver AI capabilities without abandoning their sustainability pledges. The outcome will shape not only corporate climate credibility but also the pace at which the electricity grid can transition away from fossil fuels.