The rise of data centers dedicated to training and supporting artificial intelligence (A.I.) is forcing utility forecasters back to their drawing boards as demand continues climbing for these power-hungry facilities. Data centers, along with an uptick in advanced manufacturing, are pushing anticipated growth up to 3% per year—rates that haven’t been seen since the 1980s. Utilities are beginning to use A.I. to help meet this challenge—and no new data centers are required for one of the most promising technologies now coming to market.

Meter manufacturers are providing this solution, equipping today’s latest-generation products with A.I.-powered chips. This added computing power enables real-time monitoring and throttling of delivered voltage. While savings for individual customers might be small, they can add up across a utility’s service territory, making A.I. a driver of demand and a way to address it.

A.I. boom driving growth

A.I. use has climbed dramatically in the last couple of years. Almost 80% of corporate respondents to a 2024 McKinsey survey noted their organizations used A.I. in at least one business function—a jump from 55% in 2023. Big names such as ChatGPT and Claude that can do everything from researching (and sometimes writing) a student’s term paper to generating digital code get most of the attention in the daily news cycle.

These programs represent billions of dollars in corporate development and can play across a range of industries, with user interfaces that can make internet searches more like a conversation with a friend. Alphabet, Amazon, Meta and Microsoft are among the biggest current developers of A.I. data centers.

A.I.’s ability to find patterns in massive data collections and make near-instantaneous control decisions based on current conditions have made it a natural aide for routine tasks in our daily lives, from filtering out junk in your email to aiding with online product recommendations.

Getting edge-y

These latter applications are examples of what’s called edge A.I. or A.I. at the edge. While the large A.I. engines such as ChatGPT depend on enormous cloud-based databases housed in those massive data centers, edge A.I. performs its computing functions directly on a device. This approach, using onboard chips similar to those powering larger operations, allows for minimal delays and rapid responses, improved privacy and greater energy efficiency, since remote servers aren’t involved.

Meter manufacturers now are adding new edge A.I. functionality into their latest products. Utility insiders are calling this the beginning of a new generation of advanced metering infrastructure (AMI). The smart meter utilities began installing a couple decades ago introduced remote meter reading and on/off capabilities, but with little opportunity to add new features over time. In utility lingo, this was AMI 1.0.

More recently, though, meters have become more like smartphones that can be updated with new apps pushed out by utilities. This is AMI 2.0, and it’s looking to be the future for U.S. distribution systems. 

National standards allow utilities to deliver standard 120V power between a range of 114V to 126V. Maintaining delivery close to the lower end of that spread means individual customers are using less electricity. The energy savings aren’t huge, but they can add up across a utility’s service territory. A.I.-equipped smart meters are creating new opportunities for utilities to explore these savings.

These meters are capable of load disaggregation—that is, they can track the individual power use of appliances within a home or business and recognize if delivered voltages are meeting the equipment’s needs. The resulting data can be fed back to utility substations where, when combined with data from other meters on the same circuit, it informs millisecond-fast operations to keep voltage balanced.

Savings in action

Dominion Energy has introduced voltage optimization with a smart meter rollout plan over the last several years in its Virginia service territory. Its aim sounds modest: a system-wide average of 1% energy savings for customers. But multiplied across all participating customers, that small reduction is helping the company meet the energy-efficiency standards of Virginia’s Clean Economy Act.

The program has been successful enough that the utility has patented its approach and now is marketing it to other utilities interested in making similar improvements. Dominion’s marketing materials suggest utilities adopting it could save 3%–5% of total delivered electricity when combined with other elements in its Edge Application Suite. 

cloud7days/STOCK.ADOBE.COM