Data centers could ease pressure on electricity grids more effectively by using on-site generation and storage than by moving computing workloads away from peak periods, according to an Ember analysis published in October 2026. The study examines flexibility options in nine markets across the United States, Europe and Southeast Asia.

In Ireland, data centers accounted for an estimated 1.1 GW of demand during a 6 GW system peak in January 2025. Ember projects that by 2030 they could represent 24% of peak demand, reaching 1.6 GW against a projected system peak of 6.5 GW.

The analysis estimates that shifting workloads could lower Ireland’s peak demand by around 100 MW in 2030, or about 1.5% of the projected peak. Across the other markets studied, the potential reduction ranges from 0.1% to 0.6%.

The scope for shifting is constrained by the type of computing involved. Cloud services, enterprise workloads and AI inference require low latency, while AI model training can more readily be paused or moved. Ember estimates that, after accounting for workload flexibility and average utilization, about 3% of global data center capacity could effectively be shifted during periods of grid stress.

A different approach, known as bring-your-own-capacity, allows data centers to keep computing while temporarily drawing less electricity from the grid. Facilities can use on-site generation or storage, including batteries. Ember says this could provide nearly five times as much grid relief as workload shifting.

In an example involving two Texas data centers with a combined projected load of 5.6 GW, the study estimates that shifting workloads could cut grid demand by about 0.55 GW during periods of stress. On-site generation could reduce it by up to 2.7 GW. The report says the example illustrates the potential of behind-the-meter resources when grid connections are constrained.

Ireland has also introduced a regulatory requirement for data centers of 10 MW or more to provide generation or storage capacity. The measure followed public consultation on earlier proposals to require flexibility from data centers.

Ember cautions that on-site generation does not necessarily reduce emissions: fossil-fuel-based systems could lock in emissions and infrastructure, while facilities’ private investments could duplicate capacity procured by system operators. The analysis recommends accounting for behind-the-meter resources in grid planning and connection requirements, and says renewable generation and storage can help reduce grid demand during critical periods.

The study concludes that flexibility options vary by facility. AI training offers more scope for shifting workloads, while traditional computing and AI inference are less flexible. For those data centers, renewable generation and behind-the-meter storage could be the main ways to reduce reliance on the grid during peak periods.

Sources: pv magazine.