The intelligence boom is becoming an energy and climate problem.
Global
The rapid expansion of artificial intelligence infrastructure could significantly increase greenhouse gas emissions from the power sector over the next decade. BloombergNEF projects that data centers could consume more than 5 percent of global electricity by 2035, nearly three times their current share. Much of that additional demand may be supplied by new and existing natural gas plants. If that pattern holds, power sector emissions could rise by as much as 6 percent.
The scale of planned infrastructure helps explain the concern. BloombergNEF recently increased its forecast for U.S. data center capacity in 2030 to 118 gigawatts, 52 percent above its estimate from December 2025. The figure remains uncertain because many proposed projects may never be completed. Even so, it reflects how quickly AI computing demand is reshaping electricity planning.
A more extreme scenario shows what could happen if gas becomes the dominant response. BloombergNEF identified 99 proposed gas plants intended to supply data centers directly, outside the conventional power grid. If all were built, U.S. power sector emissions could increase by between 20 and 33 percent under that specific scenario. That projection is conditional and should not be added to the broader 6 percent estimate.
Other analysts reach somewhat more moderate conclusions. Rhodium Group estimates that accelerated data center growth could leave 2035 emissions between 6 and 13 percent above its baseline scenarios, although still below 2024 levels. Both research groups agree on the central uncertainty: the final climate impact will depend on how much of the new electricity demand is supplied by natural gas and how much comes from cleaner sources.
Water use adds another dimension. Data centers consumed an estimated 222 billion liters of water for cooling in 2025, and that figure could approach 644 billion liters by 2030 if conservation measures are not strengthened. Communities hosting these facilities are therefore confronting multiple pressures at once, including electricity costs, water availability and local environmental impacts.
The broader challenge is becoming difficult to separate from the future of artificial intelligence itself. AI infrastructure is no longer only a question of chips, models and computing capacity. It is increasingly a question of what kind of energy system will be built to power them.
Información que anticipa futuros. / Information that anticipates futures.