Could power-hungry AI data centres actually make electricity cheaper?

Amid a growing backlash against AI data centres, one potentially transformative idea is getting lost in the noise. If regulated with care, the argument goes, their enormous appetite for electricity could help usher in an era of energy abundance—and even make power cheaper for everyone. Currently, the dominant narrative is running in the exact opposite direction. Pennsylvania Governor Josh Shapiro is the latest to slam the brakes on data centres, pulling a slew of projects out of the regulatory fast lane, citing legitimate concerns about what data centres could mean for “our communities, our environment, and our utility bills.” His counterparts in New York and even Texas have echoed similar worries, showing that resistance to data centres runs the political gamut. Even Alberta—arguably Canada’s most enthusiastic adopter of the new industry—is discovering that not every project touted by Premier Danielle Smith will get the green light. Last week, the Alberta Utilities Commission (AUC) rejected a proposed 1,400-megawatt natural gas power plant intended to serve a massive Synapse data centre near Olds, a town of about 9,200 people an hour north of Calgary. Notably, electricity supply or prices were not the problem. The facility was designed to be self-reliant, providing enough power equivalent to serving all of Edmonton exclusively to the data centre without connecting to the province’s electricity grid. The proposed development would have placed the gas plant, along with 600 emergency diesel generators, directly beside an established community. The area of Olds, Alta., where Synapse Real Estate Corp. planned a data centre complex with 1.4 gigawatts of natural gas-fired power generation is shown on Thursday, June 4, 2026. Lauren Krugel/The Canadian Press. The AUC concluded that Synapse had failed to justify why a project of that scale needed to be built within 800 metres of some 700 homes, given the local concerns about noise and air quality. The proponent, for its part, said while the project will not proceed as proposed, its commitment to Alberta remains “steadfast.” The decision was quickly absorbed into the broader backlash against data centres. Online reaction celebrated it as another setback for an extractive industry hell-bent on using up all of the land and power. Days later, the anger spilled into two packed town halls in Redwater and Lacombe, where residents greeted Technology Minister Nate Glubish with jeers and boos. Rival town halls Alberta’s NDP has seized on many of the familiar anxieties, including around energy affordability. In fact, the opposition party preempted the government by launching its own town hall series two weeks earlier. “Even if you save 6 percent on the transmission of electricity, but your energy prices go up way more, then ultimately your overall bill will go up,” said Nathan Ip, MLA for Edmonton-South West and the party’s technology and innovation critic. The 6-percent figure refers to an entirely separate data centre development—the $13-billion Meta project in Sturgeon County, north of Edmonton. Premier Smith has held it up as a model for how Alberta can accommodate enormous new electricity demand without passing the costs on to everyone else. Meta is expected to pay nearly $200 million a year in transmission fees, which Smith says will reduce the transmission portion of household electricity bills by up to 6 percent. Alberta Premier Danielle Smith announces Meta’s plans to build a data centre in Sturgeon County during a Calgary Stampede event on July 8, 2026. Credit: Meta. Ip says those claims deserve scrutiny. “I would love to see a third-party independent analysis that actually looks at the impact on electricity prices, natural gas prices, and ultimately what Albertans are going to pay,” he said. The Hub solicited opinions from five electricity experts—and found that the government’s central premise largely holds up, albeit with important caveats. None argued that adding giant new loads will automatically lower electricity bills. Indeed, they identified circumstances in which the opposite could happen. But there was broad agreement on a more counterintuitive point. If structured properly, data centres could make Canada’s electricity systems more efficient, spur investment in new generation, and spread the enormous fixed costs of the grid across more users, thereby lowering utility bills. Alberta, in particular, is well suited to reap those potential benefits. More supply than demand Unlike other provinces, Alberta already operates a competitive electricity market in which private investors often decide when to build new generation. Under its new approach to data centres, developers proposing projects above 75 megawatts are further being pushed toward BYOG—bring your own generation. “You need more supply than you need demand,” said Tim Boston, a former executive with ENMAX and Capital Power who now runs his own consulting firm. “What you’re seeing with those hyperscalers is that they’re going to bring usually one-and-a-half times the generation they need.” That overbuilding is driven by the need for reliability. The biggest data centres tend to want their computing capacity available around the clock, with what’s known as “five nines” reliability—99.999-percent uptime. “I go back to my Economics 101 a thousand years ago at the University of Alberta,” Boston mused. “The more supply you have, the lower prices become.” Power lines are pictured near Pincher Creek, Alta., Thursday, June 6, 2024. Jeff McIntosh/The Canadian Press. The important part is what happens to all that extra generation. Boston says it can be sold into Alberta’s electricity market. “They’re going to bid in at a very low cost,” he said. “They will drive prices down because they’re bringing in extra power.” Alberta’s oilsands already provide a real-world example of the concept. Many facilities operate cogeneration plants that produce both electricity and the thermal steam needed for industrial processes. When those facilities generate more power than they need themselves, the excess can be sold into the provincial grid. CNOOC’s Long Lake oilsands facility, for example, has 170 megawatts of cogeneration capacity. In the winter of 2022, it generated electricity at an average rate of more than 160 megawatts while consuming only around 40 on-site, leaving well over 100 megawatts on average to flow into the grid. A much larger example came online recently. Suncor began replacing its aging coke-fired steam boilers at its Base Plant with cogeneration two years ago, adding more than 800 megawatts of capacity. By late 2025, the units were fully commissioned, with the additional supply already putting downward pressure on pool prices and contributing to lower prices in early 2026, according to Alberta’s Market Surveillance Administrator. Boston argues data centres could produce a similar effect. “Everything over their marginal cost, which will be a very low number, everything above that is a win for them.” A behind-the-fence generator can primarily serve its own data centre while still maintaining the ability to export surplus electricity to the grid. Under current requirements, it would have to register as a pool asset within the Alberta market. Emergency backup generation, by contrast, operates when the data centre is disconnected from the transmission network and does not synchronize with the grid. Regardless, adding more generation is only one way these computing facilities could put downward pressure on our electricity bills. The other is much closer to the argument Smith herself has been making. More customers to spread the cost “The cost of Alberta’s transmission system—the towers and lines that move power around the province is shared by everyone who uses the grid,” Smith said in a recent video. “When a customer that size joins our electricity system, your share of the transmission bill gets smaller.” Tom Chapman, an electricity consultant and former director of transmission and distribution policy with the Ontario energy ministry, says the math makes sense. “By attracting new loads and adding to the rate base, some of those fixed system costs will now be spread over—be paid for by—more consumers,” Chapman said. “Everybody’s unit costs should come down a little bit.” Ben Hilborn, the Victoria-based entrepreneur behind Elastic Energy, a startup focused on shifting electricity demand to make better use of the grid, likens it to an arbourist who borrows $200,000 to buy a boom truck. If the arbourist expects to complete only 10 jobs a month, he argues, the cost of financing that truck has to be recovered across those 10 customers. But if the same truck can suddenly be used for 20 jobs a month, the cost of the asset can be spread across twice as much work. “You get greater utilization of this asset that has essentially a fixed finance rate,” Hilborn said in a recent episode of Alberta Edge. “If you can leverage that asset to provide more service to more people, that is a downward pressure on how much you have to charge for that service.” Credit: The Hub/OpenAI. Transmission lines, substations, and transformers are very expensive assets built to handle periods of high demand, but just like the boom truck, they are not operating at full capacity every hour of the day. Blake Shaffer, an electricity economist at the University of Calgary, describes the addition of data centres as a “numerator and denominator” balancing act. A new data centre adds costs to the electricity system (the numerator), but also adds an enormous amount of demand over which those costs can be spread (the denominator). “In the extreme, if a new load is only during periods when the wires have spare capacity, they don’t require capacity expansions and simply make use of the existing grid, increasing its utilization rate,” Shaffer wrote in an email. “This boosts the denominator with no increase to the numerator,” Shaffer said. “Average costs go down—dream scenario.” The catch is that data centres don’t automatically behave this way. For the economics to work in markets with grid constraint, Shaffer says, much depends on when they use electricity. And here, flexibility is key. Where latency is less of a concern, data centres can shift computing to off-peak hours, making better use of the grid when local electricity demand is low. Geography can also work to our advantage. If data centres anchored in Alberta provide computing services primarily to customers halfway across the globe, that “digital pipeline” can take advantage of time zone differences, matching periods of high computing demand elsewhere with periods of lower electricity demand here. A flare stack burns off excess gas at a processing facility near Crossfield, Alta., Tuesday, June 13, 2023. Jeff McIntosh/The Canadian Press. However, first-tier and hyperscale data centres tend to want to run, or at least have the capacity to run, 24/7 through the normal peaks and valleys of power use. “That sucker is going to run hard,” Boston said. The bigger prize, in his view, comes back to the over-supply of new generation that these new data centres promise to bring with them. What about the price of natural gas? Shaffer and his colleague, Kent Fellows, both pointed to another potential complication—natural gas. Alberta is attracting data centres partly because its natural gas is extraordinarily cheap, at times falling into negative prices. But if the industry succeeds on the scale envisioned by the province, that may not remain true forever. While the premier has focused on the upside of increased natural gas consumption—particularly the extra royalty bucks it could generate for government coffers—observers point to the other side of the equation. Remember: much of the new generation proposed to power Alberta’s data centres is gas-fired. Developers may foot the bill for power plants and grid connections themselves, but they can’t insulate the broader market from what happens to the price of the fuel. For all the clarity Smith has provided on why a play on natural gas could help maintain the Alberta tax advantage, she’s said considerably less about the price of the commodity itself. Building the scaffolding for more power, including renewables The concept of energy abundance naturally requires thinking beyond a fixed supply of energy. In the long run, this could mean not just more natural gas, but more of everything else. Boston points to renewables as an example. Alberta is already home to well over 6,000 megawatts of wind and solar, which together supply roughly one-fifth of the province’s electricity. The transmission needed to carry that electricity must be capable of handling peak output even though intermittent generators don’t produce at full capacity all the time. Concerns about the costs and reliability challenges accompanying rapid renewable buildout were among the issues raised by the government when it imposed its controversial seven-month pause on new renewable approvals in 2023. Data centres, Boston argues, could change that cost-benefit calculation by making much more intensive use of the expensive wires. “When you build transmission out, you’ve got to build for the big number,” Boston said. This essentially flips the order of development. Workers with Skyfire Energy install racking to mount solar panels on a home in Calgary, Alta., Tuesday, July 30, 2024. Jeff McIntosh/The Canadian Press. Instead of building transmission to reach individual new sources of renewable generation, the enormous infrastructure investment needed to serve data centres can create the backbone first. Once those wires are in place, additional wind, solar, and other generators can potentially slot into a massively expanded system partly paid for by those new industrial customers. Chapman describes this as the “enabling infrastructure” effect of transmission. “By contributing to that, I can see this data centre facilitating additional renewable generation,” he said. All about system design Alberta is unusually well positioned to test the energy abundance thesis because of its deregulated electricity market, which gives private investors greater freedom—and incentive—to respond to new demand. Things look considerably different in provinces that are less resource-endowed or have vertically integrated systems. But even in Ontario, Chapman argues a giant new electricity customer can still help spread the fixed costs of its system across more users, especially when it comes to making better use of the province’s growing nuclear fleet, which is expensive to start and stop. “Ontario has big plans for new nuclear facilities,” he said. “It will be building out those facilities in the 2030s, and large loads like data centres will be a great customer for that nuclear power.” The difference is who bears the risks. If a public utility builds billions of dollars of infrastructure in anticipation of data centre demand that doesn’t materialize—or disappears years later—existing ratepayers could ultimately be left holding the bag. That makes the BYOG principle in Alberta all the more compelling. “The logic passes the smell test,” Chapman concluded. “Meta is paying for its own generation. It’s contributing to public infrastructure.” He says the province appears to have anticipated several of the most obvious ways the data centre boom could go wrong. Beyond rules and regulations giving households and essential services priority, the government is also developing siting standards aimed at preventing the kinds of conflicts that derailed the Olds project. It’s neither possible nor necessarily desirable to prescribe every condition of a rapidly evolving industry in advance. Some flexibility allows the province to adapt as new problems emerge. But those remaining unknowns have given the NDP an opening. “I don’t think they’ve examined the entire life cycle of data centres,” said Ip, adding that his party is now in the process of crafting a shadow policy as a point of contrast. Both the UCP government and its opposition are now running competing town halls. Albertans clearly still have plenty of questions—and, apparently, plenty to yell about once someone hands them a microphone. The province should not pretend it has every answer. But nor should legitimate questions about electricity prices, natural gas, environmental impacts, and land use become an argument for shutting the door on the industry altogether. The more useful question is whether Alberta can design—and keep adapting—the rules so that developers will actually add to the overall energy pie, rather than just take from it. Get that right, and the defining feature of AI data centres—the extraordinary amount of power they consume—could turn out to be one of their biggest economic advantages. The article explores the potential for AI data centres to contribute to energy abundance and lower electricity costs in Alberta, despite growing political resistance. While concerns about environmental impacts and utility bills are prevalent, experts suggest that if managed properly, data centres could enhance the efficiency of electricity systems, spur investment in new generation, and distribute fixed costs across more users. The Alberta government is pushing for data centres to bring their own generation, which could lead to lower prices and better utilization of the grid, although challenges remain. Ask about this article — or anything in Canadian politics, economics, and public policy — powered by The Hub’s 5,000-article archive and deep area expertise. Comments (1) Great column as usual. I would just add that Virginia, a leader in data centres did not see their rates rise like other states. Lawrence Berkeley National Laboratory did a study on increased electricity demand and saw states like Virginia and North Dakota saw their rates fall by a few cents per kilowatt hour while California where demand decreased saw a 6c raise. There are plenty of studies that say the same.