AI Data Centers Push Power Grids Back to Coal and Gas

A Surging Appetite for Power

The explosive growth of artificial intelligence is colliding with the nation's clean energy ambitions in ways few anticipated just a few years ago. As AI training and inference workloads multiply, the data centers that host them are consuming electricity at a scale that is forcing utilities and state regulators to rethink how they will meet demand — and increasingly, the answer is not renewable 12.

Reporting on the trend shows a consistent theme: the sheer speed of data center expansion is outpacing the buildout of wind, solar and battery storage, leaving grid operators to lean on whatever generation is fastest to bring online 12. In many cases, that means fossil fuels, undercutting years of progress toward decarbonizing the power sector.

Coal Makes an Unwelcome Comeback

A federal Energy Information Administration report cited by public radio coverage found that coal-fired power generation rose 13% last year, a shift directly linked to data centers reshaping electricity demand across the U.S. grid 3. That increase helped drive a roughly 4% rise in carbon emissions from the power sector — the first such increase after years of steady decline 3. The finding underscores that the AI boom is not just a hypothetical risk to climate goals; it is already showing up in emissions data.

Gas Plants Fill the Gap

Alongside coal, natural gas is emerging as the go-to resource for utilities racing to serve new data center campuses. According to reporting from the Associated Press, renewable energy advocates are actively lobbying to ensure that this new wave of gas-plant construction doesn't lock in decades of fossil-fuel dependence, pushing instead for cleaner alternatives to be paired with or substituted for gas as the primary power source for AI infrastructure 5. Their concern is that once gas plants are built to meet urgent, near-term demand, they will remain in place long after cleaner options could have been scaled up.

The Grid Strain Cuts Both Ways

The pressure isn't only external. Fortune reports that AI's own power demand has become erratic enough to damage the very data centers it fuels, with equipment failures tied to volatile training loads delaying projects and straining regional grids 4. The disruptions have been serious enough to trigger a rare level-three regulatory alert, highlighting that grid instability from AI workloads is now a two-way problem — threatening both climate targets and the reliability of the infrastructure AI companies depend on 4.

Why It Matters

Taken together, the coverage suggests a structural tension: state and utility clean energy targets were built around gradual demand growth, not the sudden, concentrated load spikes that hyperscale AI facilities create 12. Whether coal and gas prove to be temporary bridges or long-term fixtures may depend on how quickly renewable and storage capacity can catch up — and how forcefully advocates and regulators push back in the meantime 35.