MAIN CONTENT

Efficiency gains do not necessarily reduce total demand

Computing history shows that the same amount of processing power can be packed into ever-smaller spaces. From room-sized machines to smartphones, gains in transistor density, chip architecture, software and cooling have steadily reduced the resources required for a unit of computation. Data centers are part of the same trajectory. Higher-density racks, liquid cooling, specialised accelerators and smarter workload management can allow a physical facility to deliver far more AI capability than an earlier generation of infrastructure.

Yet that does not mean total electricity use will fall. The IEA projects global data-center electricity consumption could rise from roughly 460 TWh in 2024 to around 945 TWh in 2030, with the United States accounting for a large share of the increase. This is a modern version of the Jevons effect: when computing becomes cheaper and more efficient, the number of profitable uses can expand so rapidly that overall demand still rises.

The building may age faster than the right to connect to the grid