On a hot afternoon in western Washington, Puget Sound Energy sends its customers a text message, an automated phone call and an email, all asking the same thing: use less power for the next few hours, and we will pay you for it. The utility added a bill credit of 50 cents per kilowatt-hour this summer to make the request land harder. Over three seasons of those events, it has cut as much as 16.6 megawatts of load at a time.
Multiply that kind of program across a continent and the numbers get large. Oracle said Nov. 11 that utilities using its Opower software delivered 1 gigawatt of peak demand reduction in North America so far this year, across 73 winter and summer events, up from 49 events in 2025. The programs engaged 4.3 million customers, 30% more than the 3.3 million of a year earlier, and the average event saved 16,592 kilowatts, which the company says equals the peak demand of about 4,150 homes.
One gigawatt is not an abstraction in this business. Oracle compares it to more than three of the Fredonia generating stations in Washington state, the peaker plants that exist to cover the hours when demand outruns supply. A peaker that is never built is a capital project avoided, and in a year when data centers have tightened reserve margins across the country, avoided capital is the product utilities most want to buy.
The mechanism is deliberately low-tech. Opower runs behavioral demand response, which asks customers to shift usage during an event and then tells them how they did, and peak time rebates, which pay them for the reduction. Neither requires a smart thermostat, a home battery or an electric vehicle charger. That matters because device-based programs reach the customers who already own the devices, while these reach the rest of the service territory.
CPS Energy in San Antonio is the clearest illustration. More than a third of its nearly one million customers take part in the behavioral events, and the utility says it called 12 events this summer and reached nearly 19 megawatts in August, drawing on 350,000 households that are now counted inside its virtual power plant. Puget Sound Energy’s program has produced statistically significant reductions in each of three deployments, according to Oracle.
The timing fits the grid’s problem. Utilities in much of the country are moving toward dual peaking, with winter demand as binding as summer demand, as more heating and transport run on electricity. Oracle expanded both programs to run year-round this year, allowing utilities to call events before winter storms rather than only during heat waves, and added tools to target different customer segments with different messages.
The limits are real and worth stating. Behavioral savings are measured against a counterfactual baseline, and baselines are contested in every jurisdiction that pays for them. A household that raises its thermostat for two hours is not a battery: it cannot discharge on command for a third consecutive day, and its willingness to participate decays when events become frequent. Comparisons to dispatchable capacity should be read as an accounting of avoided peaks, not as firm megawatts.
Regulators have made the arithmetic easier to sell. Several state commissions now allow utilities to count verified demand reduction toward the capacity they are required to hold, which turns a customer-engagement program into an asset on a planning spreadsheet. Where that treatment exists, the software pays for itself in avoided purchases; where it does not, the program has to justify itself on customer relations alone.
The business model is software, which is the point. Oracle sells the platform to utilities, and the revenue arrives as subscription and services fees rather than as anything tied to the price of electricity. In a market where every utility is being asked by its regulators and its largest customers how much load it can absorb, a product that produces a measured reduction and a customer-satisfaction story has a straightforward sales pitch.
Matt O’Keefe, the group vice president who runs Opower, described behavioral demand response and peak time rebates in the company’s release as growing pillars of demand reduction, giving utilities more ways to deepen savings and reward customers. Utility executives tend to agree in public and worry in private about how much of their capacity planning they want to rest on customer behavior.
The number to watch next year is not the gigawatt but the ratio behind it. If event counts keep rising by half while average savings per event hold steady, the programs are scaling. If the events multiply and each one delivers less, the format is meeting the ceiling that every behavioral program eventually finds, where the willing customers have already opted in and the rest have stopped reading the text messages.
One gigawatt is roughly the draw of a single large data center campus, which is the comparison that makes the achievement look small and the challenge look obvious. The grid is adding that much load faster than it is adding the plants to serve it. Whether several million households will adjust their afternoons at the necessary scale is now a capacity question, and not a marketing one.
