Consulting firms say that as AI gets cheaper, power demand could keep climbing

AI is becoming less expensive to use, but greater adoption could add strain to electricity grids, McKinsey says.

Many providers charge by tokens, the text units AI models process. Lower prices could encourage companies to use AI more, while they also work to improve the efficiency of models, chips and data centers.

McKinsey identifies data centers as the fastest-growing source of electricity demand in OECD power markets. In some markets, they are already the main driver of projected demand growth through 2030.

The firm forecasts worldwide data-center electricity use will rise 24% annually through 2030, before growth eases to 5% per year from 2030 to 2040.

The outlook reflects a tradeoff: AI tasks may require less power and cost less, but wider use could push total electricity consumption higher.