Residential customers of Portland General Electric stand to pay less on their monthly electricity bills at a moment when utilities across the country are asking households to absorb the cost of new power infrastructure driven by artificial intelligence workloads. Under a framework taking shape in Oregon, data centers drawing 20 megawatts or more would instead shoulder a 29 percent rate increase, while household rates drop by 1.3 percent. The arrangement flips the usual script, where surging commercial demand raises costs for everyone on the grid.
How Oregon’s POWER Act separates data-center costs from household bills
The mechanism behind this split traces to House Bill 3546, a measure from the 2025 Regular Session known as the POWER Act. According to the engrossed legislation, a “large energy use facility” is defined as one drawing 20 megawatts or more and operating primarily under NAICS code 518210, the federal classification for data processing, hosting, and related services. By carving out these high-consumption operations into their own tariff class, the legislation requires that the incremental grid costs they create stay on their books rather than spreading across all ratepayers.
That distinction matters because data centers consume electricity at a scale that dwarfs typical commercial or residential loads. A single large facility can draw as much power as tens of thousands of homes. Without a separate rate structure, the capital investments needed to serve that demand, including new generation, transmission upgrades, and grid reinforcement, would be folded into the general revenue requirement. Every residential customer would then pay a fraction of those costs through higher monthly bills. The POWER Act’s classification rules are designed to block that cost shift before it starts.
The bill also gives regulators a clearer legal hook to scrutinize new large-load interconnections. By tying the definition of a large energy use facility to both demand level and industry code, the measure narrows its focus to data and cloud infrastructure rather than sweeping in other industrial users. That targeted approach reflects lawmakers’ concern that AI and cloud growth could otherwise overwhelm existing regulatory tools.
PGE’s rate docket and the commission investigation driving the numbers
Oregon’s Public Utility Commission is already examining how to handle large-load customers through Docket UM 2377, an investigation into marginal cost study treatment of costs for large customers and further modifications to PGE Rule C and Rule I. The docket, referenced on the commission’s about page, signals that regulators recognized the problem before the legislature acted: existing rate design was not built for the kind of concentrated demand that AI-driven computing brings to a utility’s system.
The UM 2377 proceeding and the POWER Act work in parallel. The docket provides the technical framework for calculating what large loads actually cost the grid at the margin, while the legislation gives the commission statutory authority to enforce a separate rate class. Together, they create a two-track system where residential and small commercial customers are insulated from infrastructure spending tied to facilities they never use.
In practice, that means PGE must update its marginal cost studies to reflect the unique profiles of data centers: high, relatively constant demand; sensitivity to outages; and potential clustering near specific substations or transmission lines. Those characteristics can trigger substantial upgrades, from new transformers to expanded transmission capacity. Under the emerging framework, those costs would be assigned directly to the data-center class rather than rolled into systemwide averages.
The commission, which outlines its broader regulatory role on the agency homepage, will ultimately decide how those marginal costs translate into tariffs. That includes choices about demand charges, time-of-use pricing, and any special contract provisions for very large facilities. Each of those design elements could shift the final percentage changes away from the headline 29 percent increase and 1.3 percent decrease now circulating in policy discussions.
Unresolved questions about Oregon’s data-center rate experiment
Several gaps remain in the public record. No approved rate schedule or commission order has yet been published confirming the exact 1.3 percent residential reduction or the 29 percent data-center increase. Those figures have circulated in reporting but do not appear in the engrossed statutory language or in publicly available UM 2377 filings. Until PGE files updated tariff sheets and the commission issues a final order, the precise bill impacts for both customer classes remain subject to change.
The POWER Act’s definition of a large energy use facility also leaves room for interpretation. Facilities just below the 20-megawatt threshold could still impose significant local grid costs, especially if multiple medium-sized data centers cluster on the same feeder. Regulators will have to decide whether to address that risk through additional rules, case-by-case conditions on interconnection, or future legislative tweaks.
Another open question is how the policy will interact with economic development goals. Counties and cities courting data centers often offer tax incentives or streamlined permitting. If state-level utility policy makes Oregon comparatively more expensive for very large loads, local officials may press for exceptions or transitional arrangements. The commission will be under pressure to balance those concerns against its mandate to ensure just and reasonable rates for captive residential customers.
There is also the matter of precedent. If Oregon succeeds in holding household bills flat or lower while accommodating rapid AI-related growth, other states may borrow elements of the model. Conversely, if data-center developers delay or cancel projects in response to higher tariffs, the experiment could become a cautionary tale in how not to manage the energy transition.
For now, the promise to lower residential bills and shift costs toward data centers remains a directional signal rather than a finalized rate. The contours are clear: separate the biggest new loads, measure their true grid impact, and prevent those expenses from spilling onto everyone else’s bill. The details that will determine who ultimately pays what-exact rate levels, contract terms, and implementation timelines-will be hammered out in the commission’s dockets over the coming months. Until those orders are issued, Oregon’s bid to rewrite the relationship between AI infrastructure and household energy costs remains a closely watched work in progress.
