“AI HAS long ceased to be just technology; it is now infrastructure,” I said in a recent talk. “If it’s so important, why isn’t it making money?” somebody asked. Fair question. On paper, the investment in AI looks irrational. Billions are going in, with no clear profits coming out. Even Meta’s Reality Labs lost nearly $20bn last year alone.

So I told them a story. There was a time when railways looked like a terrible investment. Massive costs and slow returns. If you judged them by early balance sheets, you’d call it a failure. Rail was never about selling tickets but controlling movement, connecting economies, and unlocking industries. The money came later. The power came first. The people who understood that were concerned about control, not early profits.

Artificial intelligence feels eerily similar. Still framed as software and models, it is something far more physical. It runs on data centres that consume electricity at the scale of cities. It is an infrastructure that looks less like code and more like power grids, mines and industrial systems. And increasingly, executives at some of the largest technology companies are now acknowledging that the real constraint on AI is not innovation, but energy. Google’s leadership has warned that the US must significantly expand energy production to keep pace with AI demand, calling for the use of “all sources of energy”.

Tech companies are securing massive, long-term power deals while exploring private generation and reshaping where data centres are built based on access to electricity. They are projected to spend hundreds of........

© Dawn