on-prem

Grid operators could struggle to support new bit barn construction

Energy consumed by datacenters is set to grow 26 percent this year thanks to AI, and grid supply may be unable to keep pace with demand by 2030, Gartner warns.

The research giant expects global datacenter electricity consumption to reach 565 terawatt-hours (TWh) in 2026, as power demand rises from 104 GW in 2025 to 132 GW this year.

This is higher than the 500 TWh per year Gartner estimated two years ago that AI-optimized servers would consume by 2027.

And as everyone knows by now, the culprit is the ballooning requirement for compute power to drive AI workloads, as fear of missing out (FOMO) drives otherwise sensible companies to throw money at AI projects, despite seldom seeing much of any return on their investment.

In fact, Gartner notes that AI-optimized servers are what continue to fuel the increase in datacenter power consumption. This has been reported before, with hyperscalers and other buyers funneling much of their server budgets into heavily configured systems to meet the requirements of AI processing.

Now, the firm estimates that AI-optimized servers will account for 31 percent of all datacenter power consumption this year, and that, by next year, their combined power consumption will surpass that of all conventional servers in operation.

This matches up with earlier forecasts that AI was on track to overtake all other server workloads – such as databases and analytics – and become the top workload by server deployment by 2027.

But this continued expansion points to a worrying forecast. Total datacenter electricity consumption is estimated by Gartner to pass 1,200 TWh by 2030, and it says that grid supply may be insufficient to support additional datacenter capacity.

There have been earlier warnings about the bit barn energy demands outpacing the capacity of the grid to deliver. Goldman Sachs estimated that their combined energy use would more than double by the end of the decade, but if Gartner’s figures are correct, demand is already higher than where that report estimated it would be for 2027.

Energy infrastructure biz Schneider Electric also published four scenarios for future electricity consumption by AI datacenters at the start of last year, but Gartner’s latest estimate for total datacenter electricity demand in 2030 surpasses even Schneider's most aggressive forecast.

Power grid operators and datacenter developers in the US in particular are in a bind, as The Register reported recently, and energy analysts can't see an easy way out. 

“Surging demand for compute-intensive AI workloads is driving unprecedented datacenter power growth, while AI capacity is now constrained by power availability, making datacenter power security the new battle ground for scaling and protecting margins in the global AI race,” commented Gartner director analyst Linglan Wang.

But can anything be done to mitigate this coming power apocalypse?

“Infrastructure and operations (I&O) leaders must prioritize efficiency upgrades and secure grid access. They also need to invest in high-efficiency cooling systems and edge computing to mitigate power constraints and ensure sustainable, scalable growth,” Wang said, helpfully. ®