Oregon’s Power Problem Isn’t AI. It’s Supply.
Oregon’s power problem isn’t AI—it’s supply. The Northwest once powered aluminum smelters, mills, and heavy industry at enormous scale. Today, data centers are the convenient villain. But if Oregon wants electric cars, electric heat, and a digital economy, we need more reliable generation—not fewer places to plug in.
You’d think electricity demand was a brand-new problem.
At least that is the impression you might get from the current debate over data centers, artificial intelligence and high-performance computing. Listen long enough and you would think the Pacific Northwest spent the last century quietly powering refrigerators and porch lights until somebody plugged in a server farm and broke the grid.
History tells a very different story.
Long before anyone asked ChatGPT a question, the Northwest was feeding enormous amounts of electricity into aluminum smelters, steel and ferroalloy operations, pulp and paper mills, chemical plants and other heavy industry. At full production, the region’s ten aluminum smelters alone could consume about 3,150 average megawatts of electricity — roughly 27.6 terawatt-hours every year. The Northwest Power and Conservation Council says aluminum production grew alongside the Columbia River hydroelectric system and ultimately employed around 11,000 people while consuming power on a scale comparable to several large cities.
Bonneville Power Administration’s broader group of direct-service industries reached its peak in 1995, when 35 industrial customers consumed about 2,500 average megawatts. BPA says that was enough electricity to power more than two million homes.
Compare that history with today’s admittedly massive data-center industry.
A 2026 study by ECOnorthwest and the University of Virginia estimates Oregon’s 111 operating data centers consumed approximately 14 terawatt-hours in 2025, about 22.8 percent of the state’s retail electricity sales. That represents an average annual load of roughly 1,600 megawatts. Another 32 facilities are reportedly planned or under construction.
That is a lot of electricity, and pretending otherwise would be silly. But it is also important context:
Oregon’s entire current data-center load has not yet clearly eclipsed the electricity once consumed by the Northwest’s aluminum industry alone.
The comparison isn’t perfect because one number is Oregon-specific and the other covers the broader Northwest. But that is precisely the point. The Pacific Northwest power grid is regional, and enormous industrial loads are hardly a new problem for it.
Commercial Electricity Apparently Counts Differently
There is another wrinkle hiding in the statistics.
Semiconductor manufacturing is normally classified as industrial electricity consumption. Standalone data centers have traditionally been classified as commercial consumption. Oregon’s own Energy Strategy now separates data centers and chip manufacturing into a special category called “tech loads” because those traditional classifications increasingly make little sense when discussing the grid.
The transformer doesn’t care.
Whether electricity is consumed by an aluminum potline, an Intel fabrication plant or racks of Nvidia processors, somebody has to generate it, transmit it and deliver it.
Moving a giant electrical customer from the “industrial” column to the “commercial” column doesn’t magically make its electrons heavier.

Here’s Something We Actually Got Right
While industrial electricity demand was changing, the Pacific Northwest accomplished something genuinely impressive: we became dramatically more energy efficient.
Oregon residential electricity consumption was approximately 18.2 terawatt-hours in 2000. In 2024 it was about 19.8 terawatt-hours. In nearly a quarter-century, residential electricity consumption increased only about 9 percent.
That didn’t happen because Oregon stopped building houses.
Beginning in the 1980s, the Northwest adopted increasingly aggressive Model Conservation Standards for residential construction. Better insulation, higher-performance windows, improved heating systems, appliances, lighting and eventually increasingly stringent energy codes meant each generation of homes could accomplish more with less electricity.
Manufactured housing provides a particularly interesting example.
In 1992, Bonneville’s Manufactured Housing Acquisition Program, or MAP, began paying manufacturers to build homes meeting Super Good Cents and Model Conservation Standards. BPA records describe factory certification followed in many cases by installation inspections.
“Eventually, all manufactured homes shipped into the four Pacific Northwest states will be built to Super Good Cents standards.”
February 19, 1992 | Oregonian
The program was remarkably successful. About 50,000 Super Good Cents manufactured homes were constructed between 1992 and 1995, with the Northwest Power and Conservation Council estimating that the electricity savings cost roughly two cents per kilowatt-hour.
Similar efficiency improvements occurred throughout site-built residential and commercial construction. Conservation essentially became another power resource, allowing population and economic activity to grow without electricity consumption rising at anything resembling the same rate.
That was smart policy and smart engineering.
But efficiency does not repeal arithmetic.
Now We Want to Electrify Everything Else
Oregon’s own Energy Strategy projects electricity demand rising substantially in the coming decades. In the near term, the largest increase comes from technology loads — principally data centers and semiconductor manufacturing. After 2030, however, the state’s modeling shows electric vehicles becoming a major driver of electricity-demand growth, followed by further electrification of commercial, industrial and agricultural activity.
This is the part of the conversation that deserves considerably more attention.
We are simultaneously discussing electrifying transportation, replacing direct fossil-fuel heating with electric heat pumps, increasing industrial electrification, expanding semiconductor manufacturing and accommodating a rapidly growing digital economy.
Every one of those things ultimately ends at an electrical outlet.
Oregon has not yet experienced anything approaching universal EV adoption. Nor have we converted every gas furnace, water heater, commercial boiler and industrial process to electricity. If those policy goals continue moving forward, today’s data-center demand isn’t the end of the load-growth story. It may be the beginning.
Meanwhile, Look at the Supply Side
Oregon has also retired some very large sources of dependable generation.
The 1,130-megawatt Trojan Nuclear Plant stopped operating in 1993. Oregon’s approximately 550-megawatt Boardman coal plant ceased operation in 2020.
That does not mean the Northwest stopped building electricity generation. It didn’t. The region added substantial amounts of wind, natural gas, solar and, increasingly, battery storage. Anyone claiming we simply quit building power plants would be ignoring reality.
But the more useful question is whether the region added enough dependable generation, storage and transmission capacity to match the electrification we are now asking the grid to support.
The Northwest Power and Conservation Council’s answer is revealing. Its draft Ninth Power Plan, released in August 2026, concludes that the region may need more than 11,000 megawatts of additional generation, more than 5,000 megawatts of storage and another 1,060 average megawatts of energy efficiency by 2032.
That’s not exactly the profile of a region drowning in surplus electricity.
BPA’s own Columbia River history notes something else worth remembering: fish-passage requirements that increased spill at federal dams historically reduced power-system generation by an estimated 1,000 average megawatts.
That doesn’t mean fish protection doesn’t matter. Of course it does. The dams themselves profoundly altered salmon habitat, and responsible management requires addressing those impacts.
But it should also mean we demand evidence that actions taken in the name of fish recovery actually benefit fish.
A USGS study of the experimental deep drawdown at Lookout Point Reservoir found that the 2023 operation failed to create conditions that broadly improved juvenile Chinook passage and survival, with warm-water conditions associated with high mortality during part of the drawdown. A subsequent 2026 study found a more complicated picture: passage improved when the reservoir reached its lowest elevations, while earlier warm-water conditions negatively affected survival, leaving important uncertainty about the overall population benefit.
That is exactly why energy and environmental policy ought to be driven by measurable outcomes instead of assumptions.
If reducing hydropower production produces a demonstrable ecological benefit, policymakers can weigh that benefit against its energy cost. If it doesn’t, then perhaps continuing to sacrifice dependable electricity deserves another look.
Maybe the Customer Isn’t the Problem
None of this means data centers should get a free ride. A 200-megawatt customer should pay the real costs associated with connecting and serving a 200-megawatt load. Utilities should plan for those customers, transmission capacity should be accounted for, and residential customers shouldn’t unknowingly subsidize massive industrial-scale infrastructure.
But blaming data centers for Oregon’s rising electric bills confuses demand growth with the entire cost of electricity.
The Oregon Department of Energy identifies rising wholesale power costs, transmission and distribution investment, inflation, wildfire mitigation, extreme-weather damage and infrastructure replacement among the major forces pushing utility costs higher. Growing electricity demand matters, but it isn’t the only item on the bill.
Forty years ago the Northwest sent enormous quantities of electricity into aluminum pots and pulp mills. Then Oregon’s industrial economy increasingly shifted toward semiconductors and advanced manufacturing. Today another transformation is underway, with electricity being converted into computation, artificial intelligence and data services that are sold around the world.
Meanwhile, residential efficiency bought us decades of breathing room. New homes became more efficient, appliances improved, conservation programs worked and electricity consumption per household declined relative to what planners once expected.
That success may have also helped hide an uncomfortable reality: conservation can reduce how quickly demand grows, but it cannot supply electricity that doesn’t exist.
If Oregon intends to electrify transportation, heating and more of its economy while simultaneously attracting semiconductor manufacturing, artificial intelligence and other electricity-intensive industries, then the central question shouldn’t simply be, “Who is using all this electricity?”
A better question is: What kind of power system are we building to support the economy and lifestyle we say we want?
Data centers are simply the newest large customer to arrive at the meter. History suggests the Northwest has handled customers this large before. The difference is that previous generations built an electrical system around expanding supply while today’s debate too often begins by arguing over which customer ought to consume less.
Before blaming the server rack, perhaps we should spend a little more time looking at the other side of the meter.
That’s my viewpoint, but you can find some of my sources below the line.
Sources & Further Reading
Northwest Power and Conservation Council — Columbia River History: Aluminum
Historical overview of the Northwest aluminum industry and its approximately 3,150-average-megawatt peak electricity requirement.
Northwest aluminum industry history
Bonneville Power Administration — Direct Service Industries
BPA’s history of its large industrial customers, including the 1995 peak of approximately 2,500 average megawatts.
BPA direct-service industry history
ECOnorthwest / University of Virginia data-center study — reported by OPB
Oregon’s 111 operating data centers, estimated 2025 electricity consumption and planned growth.
OPB: Oregon data centers by the numbers
U.S. Energy Information Administration — Oregon Residential Energy Consumption
Historical Oregon residential electricity data from 1960 through 2024.
EIA Oregon residential energy data
Northwest Power and Conservation Council — Manufactured Housing Acquisition Program
History of MAP, Super Good Cents and energy-efficient manufactured housing in the Northwest.
Council housing-stock and MAP history
1992 Oregonian Article – Super Good Cents
Northwest Power and Conservation Council — Energy Efficiency History
Includes the approximately 50,000 Super Good Cents manufactured homes built through the early MAP program.
Northwest energy-efficiency history
Oregon Energy Strategy — Modeling and Technical Analysis
State projections showing near-term technology-load growth followed by increasing EV and broader electrification demand.
Oregon Energy Strategy modeling
Northwest Power and Conservation Council — Draft Ninth Power Plan
2026 regional assessment of new generation, storage, efficiency and transmission needs.
Draft Ninth Northwest Power Plan
Oregon Department of Energy — Boardman Coal Plant
Official history of Oregon’s approximately 550-MW Boardman generating station and its 2020 closure.
ODOE Boardman Coal Plant history
Oregon Department of Energy — Trojan Nuclear Plant
Official history of Oregon’s former 1,130-MW commercial nuclear generating station.
ODOE Trojan Nuclear Plant history
U.S. Geological Survey — Lookout Point Reservoir Drawdown
Research examining juvenile Chinook passage and survival during experimental deep reservoir drawdowns.
USGS Lookout Point Chinook study
Oregon Department of Energy — Electricity Rate Increase Drivers
State analysis of wholesale power, infrastructure, wildfire, weather and other factors contributing to Oregon electricity-rate increases.
ODOE electricity-rate drivers
Discover more from
Subscribe to get the latest posts sent to your email.
