Cheaper AI Use Could Intensify Data-Center Power Demand, McKinsey and BCG Warn
Редакция ЦОД:Главное
New reports from McKinsey and Boston Consulting Group indicate that falling AI model and token costs could make more frequent, higher-volume AI use economically viable, increasing pressure on power systems. McKinsey expects this rebound in usage to lift electricity demand despite efficiency gains in models, chips and data centers.
- Lower AI model and token costs could increase the frequency and scale of enterprise AI use.
- McKinsey expects expanded AI adoption to raise electricity demand despite efficiency improvements.
- Data centers are described as the fastest-growing load segment in OECD power markets.
- Grid interconnection and generation availability could become more material constraints on future data-center deployment.
- The reported findings are forward-looking and do not represent confirmed data-center capacity additions or measured electricity consumption.
McKinsey and Boston Consulting Group have warned that lower-cost AI could broaden adoption and increase electricity demand, according to a Business Insider report on the firms’ new publications.
The central issue is a potential rebound effect: as AI models and tokens become cheaper to use, enterprises may deploy them more frequently and at greater scale. McKinsey said the near-term result could be wider AI use and higher power demand, even as efficiency improves across models, chips and data-center infrastructure.
McKinsey characterises data-center electricity demand as the fastest-growing load segment in OECD power markets. In several markets, data centers are already the principal driver of expected electricity-demand growth through 2030, according to the reported summary of its findings.
For the data-center sector, the analysis suggests that workload growth may be driven not only by larger models but also by a broader base of AI applications and more intensive inference and training activity. This could increase demand for connected capacity, grid interconnections and low-carbon generation. Where network development does not keep pace, access to electricity could become a tighter constraint on data-center siting and delivery.
The reports do not concern a specific data-center project, new commissioned capacity, capital expenditure programme or power-supply agreement. Their conclusions concern expected demand growth rather than confirmed operating load or delivered infrastructure.
The reports reinforce that AI efficiency improvements may not reduce absolute energy consumption if lower costs expand the volume and intensity of workloads. For data-center operators and power-system planners, this could increase the importance of securing grid interconnection capacity and generation ahead of future load growth.
- — / 3 октября 2026 г. · АнглийскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.