Federal funds are helping utilities experiment with AI at the grid edge — but scale remains small.
Photo credit: Utilidata
Photo credit: Utilidata
In mid-October, Michigan’s Consumers Energy became the fourth utility to receive Grid Resilience and Innovation Partnerships Program funding to support the deployment of Utilidata’s distributed artificial intelligence platform to electrical meters in its territory.
Consumers Energy serves around two-thirds of Michigan’s population, and received $20 million in the second round of GRIP funding. The utility will use the money to get its first iteration of a managed charging program off the ground via 18,000 Nvidia-powered modules.
The funding may also help the utility assess just how much value edge computing adds to such a program — and whether the type of technical overhaul edge computing would require is really necessary for managed charging.
DOE’s grant, which the utility will match with $20 million of its own, is helping Consumers Energy get a head start on answering that question before EVs become a significant factor on the state’s grid.
According to Ryan Jackson, who leads corporate strategy and strategic projects at Consumers Energy, the 18,000 modules are “tranche one” of the utility’s foray into edge computing. That first phase is geared toward gaining more insights into when and where managed charging could be an appropriate alternative to grid upgrades. But the utility isn’t yet sure just how widely it will deploy AI at the grid’s edge.
“The GRIP award is a great down payment on what could end up being a large- scale deployment once we understand how to get as much value out of it [as possible] and make sure it’s adding the right value for our customers,” Jackson told Latitude Media.
The rollout of a new generation of AI-powered grid-edge technologies is no small feat, and Consumers Energy isn’t alone in moving cautiously. The high upfront costs and integration challenges continue to be a concern for many utility companies.
Utilidata and its distributed AI platform Karman have been at the forefront of the activity, having teamed up with Nvidia to create a custom module to power Karman. It uses a combination of on-chip and cloud-based software, plus distributed AI, to collect real-time data and manage distributed energy resources.
When the pair first announced their partnership in March 2023, they debuted both a meter-embedded version of their module as well as a “meter collar,” which is essentially an adapter for existing meters.
Utilidata hopes that the embedded module will be the version that most utilities ultimately purchase in large quantities. A smart meter created via their partnership with Aclara will be ready early next year.
Brought to you by Uplight: Learn how virtual power plants differ from traditional demand response programs and how utilities can unlock grid flexibility.
Brought to you by Uplight: Learn how virtual power plants differ from traditional demand response programs and how utilities can unlock grid flexibility.
Brought to you by Uplight: Learn how virtual power plants differ from traditional demand response programs and how utilities can unlock grid flexibility.
Brought to you by Uplight: Learn how virtual power plants differ from traditional demand response programs and how utilities can unlock grid flexibility.
That said, the collars, which can be added to existing meters, are a more flexible entrypoint for utilities, and all current programs are using that form of the module. The number of collars in the field today number in the hundreds, Utilidata confirmed.
Consumers Energy, for its part, hasn’t reached a final decision about which form of the chip it will roll out, though Jackson said in an interview with Latitude Media that “concentrations of collars” is one of the elements they’re considering in their deployment plans, suggesting that collars are likely to be incorporated.
Portland General Electric, ComEd, and Duquesne Light Company, which received awards from the first round of GRIP funding to support edge computing programs, also hadn’t decided between meters and collars by the time their funding announcement came down. Those projects are slated to begin deployment next year, however, so have the option of using the embedded form of Utilidata’s platform.
With additional funding in hand, more detailed planning work now begins for Consumers Energy’s 2026 Karman deployment, Jackson said.
Given where Michigan is on the EV adoption curve, there aren’t all that many meters to choose from. As of 2023 the state had just over 50,000 registered EVs, accounting for about 0.6% of market share. It’s a far cry from Governor Gretchen Whitmer’s 2030 target of 2 million EVs on Michigan roads, supported by 100,000 chargers — but there isn’t really a time crunch to implement managed charging (though a recent utility study tried to pinpoint the right “tipping point” for infrastructure investment for a utility in nearby Indiana).
“There’s ample distribution capacity to support EV charging today,” Jackson said. ”And we’ve had a lot of success so far encouraging customers to charge off-peak, taking advantage of time-of-use rates.”
The utility has already invested in advanced distribution management systems, but Jackson added that the added layer of data will help it to get more granular in its preparations for end-of-decade EV goals: “It’s really about the timeliness of the data, the accuracy of the data and the location of the data, so that we can match local interventions with local grid needs.”
Jackson said he isn’t sure yet what the “end game” will be for Consumers Energy when it comes to edge computing — it might not be necessary, for example, to put AI modules on every meter. “It could actually be that there’s other hardware that we would want to add this type of metering technology to,” he said.
The overall number of modules being deployed via GRIP funding are relatively small compared to the some 50 million meters Utilidata estimates are nearing the end of their useful lifetimes and could be replaced with Karman-embedded upgrades. But Jess Melanson, president and COO of Utilidata, said the key benefit of that funding lies in “democratizing” the learnings of various small programs. Once AI modules are up and running in various GRIP-backed programs around the country, in different scenarios of EV deployment and different circuit and grid conditions, each deployment will benefit from those learnings.
“When it comes to an AI investment, you're not only accelerating the adoption of the hardware, but you're accelerating the learning, and that learning can be shared,” Melanson said. “And that's one of the great benefits of an approach like this.”