For more than a decade, U.S. electricity demand was relatively flat. That era appears to be ending.
The U.S. Energy Information Administration now identifies data-center load as a dominant driver of long-term electricity-demand growth, with national power demand already rising meaningfully after years of stagnation.
At the same time, AI is changing what a “large load” looks like.
Data centers aren’t simply another commercial building connecting to the distribution system. Individual campuses can require hundreds of megawatts, placing them in the same conversation as generation development, transmission planning, substation capacity and regional resource adequacy.
This creates an interesting shift in the energy market:
The competitive advantage for AI infrastructure may not be computing capacity alone. It may be access to electricity.
And that raises a bigger question for utilities.
Should the industry continue treating data centers primarily as loads that need to be served?
Or should we begin designing them as flexible grid participants?
A large data center combining onsite generation, battery storage, backup generation, controllable computing workloads and participation in demand-response or wholesale markets begins to look very different from a traditional customer.
It starts looking like a grid resource.
FERC has already begun addressing this issue. In June, the Commission directed the six regional grid operators under its jurisdiction to examine or reform tariffs governing the interconnection of large loads such as data centers.
The next phase of grid modernization may therefore require more than building generation and transmission.
We will also need better DERMS, demand response, flexible-load orchestration, operating envelopes and real-time coordination between large customers and grid operators.
AI may be creating one of the largest new electricity-demand challenges in decades.
It may also create the economic incentive needed to modernize how the grid manages demand.
The question isn’t only: “Where will we get enough power for AI?”
The more interesting question may be:
“How much of that new AI load can become part of the solution?”
Sources:
U.S. EIA — Annual Energy Outlook 2026: https://www.eia.gov/outlooks/aeo/
FERC — Large Load Integration Action: https://www.ferc.gov/news-events/news/ferc-launches-aggressive-targeted-action-speed-large-load-integration
IEA — Energy and AI: https://www.iea.org/reports/energy-and-ai
#ArtificialIntelligence #AI #DataCenters #Energy #ElectricGrid #GridModernization #DERMS #DemandResponse #VirtualPowerPlants #VPP #EnergyMarkets #Utilities #FERC #EnergyTransition #PowerGrid
Photo courtesy Jerry Wei