The demand for generative artificial intelligence (GenAI) is skyrocketing, and it’s putting a serious strain on AI datacentres. Gartner predicts that by 2027, 40% of these facilities will face power supply issues. The reason? The rapid increase in energy needs as more server farms pop up to support AI workloads.
Gartner’s numbers are eye-popping. By 2027, datacentres will require 500 terawatt-hours (TWh) of power annually just to run AI-optimized servers. That’s more than double what was needed in 2023. Bob Johnson, a vice-president analyst at Gartner, points out that the growth of hyperscale datacentres for GenAI is driving an insatiable need for energy. Utility providers simply can’t keep up with the demand.
This situation threatens to disrupt energy availability and could stifle the growth of new datacentres for GenAI and other applications by 2026. Major cloud providers like Microsoft and Amazon are already aware of this trend, leaning heavily into building new datacentres to meet this exploding demand.
But here’s the catch: these new facilities might not launch as quickly as expected due to ongoing power supply challenges. Johnson highlights that larger datacentres are in the works to accommodate the vast data needed for training large language models, but until new power transmission and generation capabilities come online, businesses will face short-term energy shortages.
Enterprises looking to leverage GenAI for future growth need to keep this in mind. Energy costs are likely going to rise. Johnson mentions that big power users are already negotiating with major energy producers for long-term, guaranteed power sources, cutting through the noise of grid demands.
In the short term, the operational costs for datacentres will spike as operators scramble to secure necessary power. That’s going to trickle down to AI and GenAI service providers, impacting their pricing.
Another layer to this issue is the likely reliance on non-renewable energy sources to meet escalating energy demands from GenAI. Johnson warns that the increased use of datacentres will lead to greater carbon emissions in the near term. This complicates sustainability goals for both datacentre operators and their clients, making it harder to achieve net-zero targets.