The glossy, almost fantastical vision of an artificial-intelligence-driven future has always been powered by very real, very present-day energy infrastructure. AI may feel weightless and abstract—lines of code floating in the cloud—but the machines behind it are anything but. Increasingly, it’s becoming clear that this supposed leap into the future is not only fueled by today’s power grid, but is also leaning heavily on the energy sources of the past. Chief among them is coal: an aging, highly polluting fuel whose climate impacts rival those of other fossil fuels, but whose localized environmental and human health consequences are especially severe and immediate.
Coal has long been understood as a climate liability. But its danger doesn’t stop at carbon emissions. Coal plants release mercury, sulfur dioxide, fine particulate matter, and other toxins that directly poison air, water, and soil—often in low-income or politically marginalized communities that have little power to resist. As AI data centers multiply and energy demand spikes, these risks are being reintroduced and intensified under the banner of technological progress.
According to a recent Politico report—drawing from new analysis by the Frontier Group—utilities across the United States have postponed the planned retirement of at least 30 coal-burning power units. These plants were slated for closure, not because of environmental pressure, but because coal had simply become economically obsolete. Now, those shutdowns are being delayed for a single reason: to provide electricity to data centers, many of them built specifically to support AI training, inference, and cloud-scale computing.
The political framing of this pivot has been blunt. In a Fox News op-ed quoted by Politico, EPA Administrator Lee Zeldin wrote that “to meet growing demand from domestic manufacturing and the rapid expansion of artificial intelligence data centers, America’s coal-powered power plants must remain in operation.” That statement makes explicit what had previously been implicit: AI’s energy appetite is being used as justification to extend the life of America’s dirtiest power source.
This is a stunning reversal when viewed against recent history. Throughout most of the Obama administration and the early years of Donald Trump’s first term—from roughly 2010 through early 2019—U.S. utility companies announced plans to shut down 546 coal-fired generating units, amounting to approximately 102 gigawatts of capacity. To put that in perspective, this represented a decommissioning effort large enough to power New York City at peak summer demand more than ten times over.
Even then, coal’s decline was rarely about environmental idealism. As Grist noted back in 2012, the collapse of coal was “primarily due to changing market conditions,” not tougher environmental rules. In fact, many regulatory requirements for coal plants were becoming more lenient over time. Coal was simply losing to cheaper, cleaner alternatives like natural gas and renewables.
Now, those market conditions are shifting again—but not because coal has become cleaner or safer. Instead, AI has arrived as a power-hungry wildcard, warping energy demand curves and creating incentives to keep otherwise outdated infrastructure running. The difference is that this revival carries a measurable human cost.
A report cited by the National Institutes of Health found that between 1999 and 2020, an estimated 460,000 deaths would not have occurred in the absence of emissions from coal-fired power plants. These aren’t abstract or disputed numbers. They represent heart attacks, strokes, chronic lung disease, cancers, and asthma—conditions linked directly to coal pollution. Coal doesn’t merely warm the planet; it kills people in the present tense.
There was, briefly, an attempt to curb this. The Obama administration eventually imposed emissions limits squarely targeting coal pollution. But when Trump returned to office, those rules were swiftly dismantled. By 2019, many had been eliminated entirely. In his current administration, Trump has gone even further, not just scrapping protections but actively weakening what remains.
According to Politico, nearly 70 coal-fired power plants have now been granted permission to ignore EPA mercury and soot regulations until at least 2027, while the agency works on crafting new, more permissive standards. This effectively creates a multi-year grace period during which known toxic emissions are allowed to continue unchecked.
The rollback isn’t limited to power generation. As Gizmodo previously reported, Trump has repeatedly praised what he calls “beautiful, clean coal” while eroding safety standards for the workers who mine it. Labor unions are currently suing the administration over its decision to pause silica dust regulations—rules the Department of Labor estimated would save 1,000 lives by preventing black lung disease and other fatal conditions.
In this context, Trump’s self-portrayal as the savior of the coal industry becomes more than political theater. He didn’t create AI’s exploding demand for electricity, but he has actively reshaped regulatory and economic conditions to ensure coal benefits from it. The gains flow upward—to shareholders, executives, and energy companies kept profitable by regulatory retreat. The costs, as usual, are absorbed downward—by miners, nearby communities, children breathing polluted air, and patients filling hospitals with preventable illnesses.
So yes, AI may represent the future. But right now, that future is increasingly being propped up by an energy source that belongs firmly—and dangerously—in the past. If artificial intelligence is meant to symbolize human progress, it’s worth asking why its growth depends so heavily on technologies that shorten human lives.