If you listen to the news, you’ll know that A.I. data centers are demanding more power every day and utilities are struggling to keep up. The latest data center designs can demand as much power as an entire generating station might be able to produce. As this trend develops, however, utilities and their tech company customers are considering new ways to address these concerns. 

A.I. delivers real value in our data-heavy environment. Companies already are turning to the technology to help address a range of needs, from vegetation management and risk reduction to real-time grid monitoring. This adoption is also beginning to answer the question of how utilities can best use their massive data collections to improve service and maximize the lifespans of existing assets.

Growing pains

While we often hear what A.I. will do in the future, it is already in use, running in the background as we search and shop online, work through customer service phone menus, and use assistive driving features such as lane control. This growing reliance, along with a recognition that we’ve only scratched the surface of A.I.’s potential, is leading to rapid growth in data center construction to house all the servers needed to support its operation.

These centers can be exponentially bigger and more energy-hungry than predecessors built just a decade ago. According to the U.S. Department of Energy (DOE), A.I.-specific data centers built in 2023 regularly required more than 100 megawatts (MW) of power capacity, versus 5–15 MW in 2014. And larger facilities can require significantly more—just the first phase of Microsoft’s $3.3 billion Mount Pleasant, Wis., data center will be drawing 450 MW when completed in late 2026, enough to power more than 300,000 homes.

Utilities are scrambling to keep up with this rapid demand growth. As a point of comparison, the 450 MW Microsoft will be adding to utility We Energies’ load is about half the size of an average combined­-cycle gas turbine plant, which would typically carry a price tag of about $1 billion. Additionally, the greater load can mean more wear on existing lines and switchgear.

As an added challenge, utilities’ capital investments generally are paid by all ratepayers. As a result, boosting capacity to meet one data center’s needs could raise rates for all consumers in the service territory.

A.I. to the rescue?

While utilities are scrambling to meet the power demand of these new facilities, they’re also turning to A.I. technology to make the most of assets they already have. Local distribution systems are awash in data, thanks to modernization efforts undertaken over the last two decades. The smart meters now connecting to most customer premises, as well as many of the sensors, switches and other devices installed throughout their service territories provide real-time status at local and system-wide levels.

A.I. is providing a range of applications to help utilities make the most of this data.

Predictive maintenance: Drawing on data from power line sensors, transformers and other components, A.I. can identify subtle patterns in temperature, vibration and electrical signatures to flag equipment for inspection or replacement before it fails.

Grid optimization: A.I. can model power flow across distribution networks to identify stress points, helping utilities plan infrastructure upgrades where they’ll make the most impact.

Load forecasting: By analyzing historical usage, weather forecasts and demographic trends, A.I. can help predict demand patterns to help ensure adequate generation capacity.

Vegetation management: Using satellite imagery and LiDAR data, A.I. can identify trees and other vegetation that might threaten power lines to help prioritize trimming programs and prevent outages.

Adapting for the future

Tech companies are also using A.I. to support local grids. While the internet is a 24/7 operation, that doesn’t mean traffic to A.I. servers is a constant—instead, as with business activity in general, there are slower and busier times. Companies have flexibility to respond to conditions with the local utility grid. This can include practices such as grid-aware computing, which scales back server activity during peak-demand periods and ramps back up when power is plentiful. 

The technology also can optimize its cooling efforts, dynamically adjusting levels based on workload, server temperature and outside conditions to help lower facility demand. And as many data centers are adding on-site generation and battery storage, A.I. can coordinate local backup power systems with utility needs, allowing facilities to serve as virtual power plants that support grid stability during peak periods.

pingingz / stock.adobe.com