If you're wondering why the AI revolution might hit some speed bumps, here's a big one: the American power grid wasn't built for this moment. Alphabet Inc. (GOOG) subsidiary Google just highlighted what might be the most unglamorous constraint on artificial intelligence expansion—you can build all the fancy data centers you want, but good luck plugging them in.
When "Please Hold" Means a Decade
Marsden Hanna, Google's Global Head of Sustainability and Climate Policy, delivered some eye-opening numbers at an American Enterprise Institute event this week. According to Reuters, he told attendees that connecting new facilities to the electrical grid isn't just slow—it's absurdly slow.
"Transmission barriers are the number one challenge we're seeing on the grid," Hanna explained. "We had one utility who told us 12 years to study the interconnection timeline, which is sort of wild, but that's what we're seeing."
Twelve years just to study the connection timeline. Not to actually connect anything—just to figure out if and how it might be possible. That's the kind of delay that makes tech companies rethink their entire infrastructure strategy.
Plan B: Set Up Shop Next Door
Google's solution? If you can't connect to the grid in a reasonable timeframe, skip the grid entirely. The company is exploring colocation arrangements, essentially building data centers right next to power plants. This lets them tap directly into electricity generation without navigating the transmission system's bureaucratic maze.
But Hanna made clear this is meant as a bridge, not a permanent fix. "That's the strategy we're pursuing with colocation and our hope is that these can eventually be grid-connected resources," he said. The endgame still involves proper grid integration—it's just that waiting around for a decade isn't an option when you're racing to deploy AI infrastructure.