McKinsey Predicts 24% Annual Increase in Electricity Consumption

McKinsey forecasts an annual increase of 24% in electricity consumption by data centers until 2030, followed by a slowdown between 2030 and 2040. Cheaper tokens could stimulate the use of AI and energy demand. BCG observes that the most advanced companies are managing their token spending. After 2030, the trajectory remains uncertain.

After 2030, uncertain trajectory and cost-energy trade-off

McKinsey indicates that the growth of data centers after 2030 remains uncertain, as companies are still trying to determine whether AI provides sufficient value. The firm anticipates a slowdown in the growth rate to 5% per year between 2030 and 2040. According to McKinsey, as each AI task becomes cheaper and less energy-intensive, AI expands into more areas, and total energy demand increases. Both firms emphasize that lower AI costs do not automatically lead to a decrease in energy consumption and could, on the contrary, lead to new uses.

In the short term, increased usage and strong pressure until 2030

In the short term, McKinsey estimates that cheaper tokens could promote broader use of AI and increase electricity demand, while efforts are being made to improve the efficiency of models, chips, and data centers. McKinsey describes the electricity demand of data centers as the fastest-growing load segment in OECD markets, noting that in several markets, they are already the main reason for the expected increase in demand until 2030. The firm forecasts a global growth of 24% per year in data center consumption until 2030. These findings come as data centers are already consuming more energy and AI billing is often done per token. The two reports published this week indicate that the intensification of usage could pose a problem for the electrical grid.

Managing token spending among advanced companies

According to BCG, companies that effectively control their token spending achieve the best results with AI. The most advanced companies avoid having their usage costs automatically increase. A survey conducted by BCG among 1,300 executives across more than 20 sectors identified a group of "futuristic companies," three-quarters of which have made an explicit decision regarding token spending and manage it based on precise returns. Half of this group actively encourages the use of paid AI tools, compared to 25% of lagging companies, while 22% of advanced companies impose limits or controls to manage costs. In this context, where models and tokens are becoming cheaper, this price decrease could encourage more frequent and larger-scale usage, which could increase pressure on the electrical grid.