The grid was never designed to do this, at this pace
Across the Eastern United States, a technological revolution is colliding with the physical limits of a power grid built for a different era. The rapid expansion of artificial intelligence infrastructure has driven wholesale electricity prices up 75 to 76 percent within PJM's service territory — a region powering 65 million lives — not through natural disaster or geopolitical shock, but through the sheer, relentless appetite of machines that never sleep. It is a reminder that every leap in human capability carries a hidden cost, and that the systems we inherit were built for the world we imagined, not the one we are racing to create.
- Wholesale electricity prices across PJM's Eastern US grid have surged nearly 76 percent, driven not by weather or war but by the insatiable power demands of AI data centers multiplying across the region.
- A single large data center can draw as much electricity as a small city — and dozens are coming online simultaneously, overwhelming infrastructure that took decades to build and cannot be expanded overnight.
- Federal regulators have issued a stark warning that PJM may be 'too big to function,' signaling that the grid's very structure may be unfit for the crisis now unfolding inside it.
- Tech companies are not waiting for the grid to catch up — they are signing long-term power contracts and breaking ground on new facilities even as utilities and energy developers struggle to respond at anything close to the same speed.
- The collision between AI's pace of expansion and the slow, regulated world of energy infrastructure is the fault line where prices are rising, stability is eroding, and the next chapter of both industries will be written.
Somewhere in the server farms stretching across the Eastern United States, the lights never go out. Thousands of processors run through the night, training AI models and consuming electricity at a scale that has begun to reshape the entire regional power market. The bill has come due — and it is staggering.
Wholesale electricity prices across PJM, the grid operator serving 65 million people from New Jersey to Illinois, have surged 75 to 76 percent in recent months. The cause is not a heat wave or a cold snap, but the relentless expansion of artificial intelligence infrastructure. Data centers built to power the next generation of AI are drawing power at rates the existing system was never designed to handle. What began as a technology boom has become an energy crisis.
The problem is not simply that demand has grown — it is that demand has grown faster than anyone planned for. Regulators and utilities built their forecasts on historical patterns that did not account for an industry capable of constructing dozens of massive facilities in a matter of months. The infrastructure that took decades to build cannot be expanded overnight, and the grid is being asked to do something it was never designed to do.
Federal regulators have begun to sound alarms. The chairman of the Federal Energy Regulatory Commission has warned that PJM may be 'too big to function' — a signal that the current structure may be inadequate for the challenges ahead. Either the grid must be rebuilt at unprecedented speed, or data center growth must be constrained, or both.
Meanwhile, the data centers keep coming. Companies racing to build AI infrastructure are not waiting for the grid to catch up, signing long-term power contracts and beginning construction even as utilities scramble to respond. New power plants take years to permit and build. Renewable projects face their own delays. The mismatch between the speed of AI expansion and the pace of energy development is the core tension driving prices higher — and the processors, indifferent to the crisis they have created, continue their work.
Somewhere in the server farms stretching across the Eastern United States, the lights never go out. Thousands of processors hum through the night, training neural networks, running inference engines, consuming electricity at a scale that has begun to reshape the entire regional power market. The bill has come due, and it is staggering.
Wholesale electricity prices across PJM—the grid operator serving 65 million people from New Jersey to Illinois—have surged 75 to 76 percent in recent months. The culprit is not a summer heat wave or a winter freeze, but rather the relentless expansion of artificial intelligence infrastructure. Data centers built to power the next generation of AI models are drawing power at rates that grid planners did not anticipate and the existing system was not designed to handle. What began as a technology boom has become an energy crisis.
The scale of the demand is difficult to overstate. A single large data center can consume as much electricity as a small city. Multiply that across dozens of new facilities coming online across the region, all competing for the same finite supply of power, and the market responds the only way it knows how: prices climb. Wholesale rates have nearly doubled. Consumers and businesses connected to PJM's grid are beginning to feel the weight of those increases in their monthly bills.
The problem is not simply that demand has grown. It is that demand has grown faster than anyone planned for. Regulators, grid operators, and utility companies built their forecasts on historical patterns of electricity consumption. They did not account for an industry that could suddenly decide to build dozens of massive facilities in a matter of months. The infrastructure that took decades to construct—power plants, transmission lines, substations—cannot be expanded overnight. The grid, in other words, is being asked to do something it was never designed to do, at a pace it cannot match.
Federal regulators have begun to sound alarms. The chairman of the Federal Energy Regulatory Commission has suggested that PJM may be "too big to function," a stark warning that the current structure of the grid operator may be inadequate for the challenges ahead. The implication is clear: something has to change. Either the grid must be rebuilt and expanded at unprecedented speed, or the growth of data centers must be constrained, or some combination of both must occur.
Meanwhile, the data centers keep coming. Companies racing to build AI infrastructure are not waiting for the grid to catch up. They are signing long-term power contracts, securing sites, and beginning construction. Each new facility adds to the strain. The wholesale market is sending a price signal—build more power generation, build it fast—but the response from utilities and energy companies has been sluggish. New power plants take years to permit and construct. Renewable energy projects face their own delays. The mismatch between the speed of AI expansion and the speed of energy infrastructure development is the core tension driving prices higher.
What happens next will shape not just the economics of AI development but the stability of the entire Eastern grid. Regulators are watching. Utilities are scrambling. And in data centers across the region, the processors continue their work, indifferent to the crisis they have created.
Notable Quotes
The grid operator may be too big to function under current demand pressures— FERC Chairman Swett
The Hearth Conversation Another angle on the story
So these are new data centers being built specifically for AI training and deployment?
Yes. Companies need massive computing power to train large language models and run inference at scale. A single facility can be enormous—consuming electricity equivalent to a city of tens of thousands of people.
And this happened suddenly? The grid didn't see it coming?
The buildout accelerated much faster than anyone forecast. Grid planners work on decade-long timelines. AI infrastructure expansion happened in months. There's a fundamental mismatch in speed.
The 75 percent price jump—that's wholesale prices, not what consumers pay directly?
Right. But wholesale prices feed through to retail rates eventually. Consumers will see it on their bills, though the full impact takes time to appear.
What does it mean that PJM is "too big to function"?
It suggests the current grid operator structure can't manage the complexity and scale of demand anymore. You might need to break it up, restructure it, or fundamentally change how it operates.
Can they just build more power plants?
In theory, yes. But new plants take years to permit and build. Renewables are faster but still not fast enough to match AI's appetite. The timing is the problem.