A generational crisis is brewing

The rise of AI is consuming energy faster than the power grid can supply.

By 2028, U.S. data centers will use up to 12% of the nation’s electricity — enough to power 24 million homes.

Nuclear energy is the only realistic path forward, yet building a single plant takes an entire decade.
Since each reactor produces about 8,000 GWh per year, roughly 70 new ones would be needed just to meet AI’s demand.
The U.S. has built only two in the past 30 years and has none under construction, while China already has 29 underway.

Meanwhile, companies are signing multi-billion-dollar AI deals that demand as much power as large cities. The grid was never designed for this kind of technological arms race: every AI model, every image, every response burns electricity.

The smarter our machines become, the hungrier they get.

Every query, every upgrade, every training run consumes more energy than the last. The new factories of 2025 are data centers — massive, relentless, always running. The reality is simple: we’ve built infinite digital demand on top of a physical grid never meant to handle it. We’ll have to get used to blackouts, brownouts, and inevitable price chaos — as our digital ambition collides with our physical limits.

We taught machines to think, but forgot how to keep the lights on. AI may indeed reshape the world — but only if the world can keep the lights on.