AI needs electricity now; U.S. grid lags by years

The root of the problem is a timing mismatch. Tech companies and utilities operate on different clocks—the former moves fast, the latter must ensure thousands of grid components work together through 'deliberate study and planning.' This is why the grid might not be ready for data centers' demand, even as AI itself grows faster than any prior technology. The grid was already constrained before AI. After 25 years of little demand growth, the industry 'got out of practice on building new infrastructure,' says Gramlich. Now, electric vehicles, electric space heating, new manufacturing, and data centers are all expanding at once, with data centers accounting for about half of the new growth. The result is scarcity. Not everyone will get the full service they want, at least until the system catches up. Utilities are asking data center developers for more money upfront to avoid being 'stranded' if projects don't materialize, and delays are likely. Whether many data centers actually get built remains a question mark, according to Burke.