AI is getting cheaper to use, but as companies use it more, we’re bound to see more pressure on the electric grid.
In two new reports published this week, McKinsey and Boston Consulting Group said that as companies move beyond experimenting with chatbots and start using AI across more parts of their businesses, cheaper models and falling token prices could make it easier to justify more frequent, higher-volume AI use — and that could pose a problem for the grid.
Companies often pay AI providers based on tokens, the small units of text that models read and generate. In the short term, McKinsey said that cheaper tokens could make AI more widely used and raise electricity demand, while companies push to make models, chips, and data centers more efficient.
That matters because data centers are already using more power. McKinsey called data-center electricity demand “the fastest-growing load segment in OECD power markets.” In several markets, it said, data centers are already the biggest reason electricity demand is expected to grow through 2030. The firm expects global data-center electricity demand to grow by 24% a year through 2030, then slow to 5% a year between 2030 and 2040.
Even as companies scale their use of AI, BCG found that the companies getting furthest with the technology are not simply letting token spending rise by default. Three-quarters of its “future-built” companies — the most AI-mature group in its survey of 1,300 C-Suite and senior executives across 20-plus sectors — had made an explicit call on token spending and managed it toward a specific return. Half of that group actively encouraged employees to use paid AI tools to maximize adoption, compared with 25% of those companies lagging with using the technology, while another 22% of future-built companies set limits or controls on usage to manage costs.
There is a tradeoff. McKinsey said that as companies make each AI task cheaper and less power-intensive, AI will inevitably become useful in more places, and total power demand will rise.
McKinsey said the broader picture for data-center growth after 2030 remains uncertain because companies are still working out whether AI delivers enough value. But the reports suggest that cheaper AI will not automatically mean less energy use. It could mean businesses find many more ways to use it.
