AI Data Centers' Environmental Footprint: Power Use Surging, Gas Buildout Growing, Data Scarce
Summary
- • About 90% of Americans are more concerned than excited about AI data centers, per a September poll
- • AI data center electricity use rose 50% last year, but still about 0.5% of global power
- • Data centers could reach up to 12% of US electricity by 2028, per Lawrence Berkeley National Lab
- • Onsite gas turbines for data centers raise air pollution and health concerns
Details
About 90% of Americans are more concerned than excited about AI data centers
Result of a CNN poll by SSRS in September. Public opinion on the infrastructure behind AI is overwhelmingly wary, which could shape local approvals and policy.
Environmental impact depends on location, power source, and technology used
There is no single footprint figure. The same workload can have very different emissions and resource impacts depending on the grid and facility design.
AI providers release little transparent data on energy and environmental impact
This limits independent verification and forces outside analysts to rely on estimates and projections.
A simple text query uses less electricity than a TV running over the same time
Per the International Energy Agency. Video generation can require thousands of times more energy, so the shift toward multimedia AI drives rising demand.
AI data center electricity use surged 50% last year
Per the International Energy Agency. Demand is growing fast even though the absolute share of global power remains small.
AI data centers used about 0.5% of world electricity in 2025, projected 1.5% by 2030
Per Our World in Data. A tripling of share within five years, though modest globally compared with local concentration.
Nearly half of data center consumption is in the US, possibly up to 12% of US electricity by 2028
Per Lawrence Berkeley National Lab. Impact is concentrated in the US, so regional grid effects may be much larger than global averages suggest.
Global data centers are powered mostly by coal (30%), renewables (27%), and gas (26%)
Coal remains the largest single source for data center power worldwide, which drives the carbon footprint.
US has the most gas-fired capacity in development globally, over a third intended for AI
Per Global Energy Monitor's January report. AI demand is driving new fossil fuel infrastructure, not only renewables.
Data centers are building their own gas turbines, which emit NOx and particulates
These turbines are less efficient than utility-scale plants and carry air pollution health costs for nearby communities.
Stat = data/figures, Insight = analysis, Context = background, Infrastructure = power and buildout
What This Means
AI's electricity appetite is small globally today but growing quickly and concentrated in the US, where data centers could take up to 12% of power by 2028. How much harm it does depends heavily on where facilities are built and what powers them, and onsite gas turbines add local air pollution risks. Limited disclosure from AI companies makes it hard to judge the real impact, which helps explain widespread public concern.