Sameer's Take is an opinion column. The views are Sameer Khan's own.
Last Updated: November 19, 2026
AI data centers just lost at the ballot box. On November 3, voters sided against data center development in 40 of 41 local ballot measures across 10 states. My view is simple: the tech companies building AI data centers should pay the full cost of the extra power they need, not the households and small businesses on the same grid. And here in Canada, I think the backlash is good news, because Canada can pick up the projects other places reject.
Let me break this down simply: why people are pushing back, what the data says about electricity bills, who should pay, and what the backlash means for Canada.
A bit about where I'm coming from: I have five years in research and advisory services and AI, and I run this newsletter. This is opinion. The facts below are sourced and dated so you can check them yourself.
Free download: The AI Stats Sheet 2026. 36 key AI numbers for business on four printable pages, with the source under every figure. Join 2,000+ business leaders who read AI Business Weekly and get it in your inbox right away.
What Is an AI Data Center?
An AI data center is a facility packed with specialized chips, mostly GPUs, that train and run AI models. It works like any data center, but it draws far more electricity, because thousands of those chips run flat out around the clock. The newest campuses are planned at a gigawatt or more, which is the scale of a large power plant.
That is why this debate is really about power. For the chips inside these buildings and who makes them, see our Nvidia and AI chips hub.
Why Are People Against AI Data Centers?
A big reason people are against AI data centers is cost. Opponents point to higher utility bills, and the midterms made that impossible to ignore.
Ballotpedia counted 41 local ballot measures on data center policy in 10 states on November 3, with Ohio alone accounting for 21, according to Ballotpedia's election-day preview. All but one went against data center development, and the restrictions passed by an average margin of 49 percentage points, political scientist Barbara Trish wrote in an analysis for The Conversation, republished by Fortune on November 13. The one exception was Harwood, North Dakota, where a project was already underway. The same analysis cites NBC exit polls showing 56% of voters called data centers a "major" or "crucial" factor in their congressional vote, and notes that opposition coalitions included voters dealing with high electricity costs.
Investors noticed right away. Data center and power stocks fell the next day, as we covered in our story on data center stocks sliding after voters rejected local projects.
Do Data Centers Raise Electricity Bills?
They can, and in the largest US grid region the cost is already showing up. The honest answer is that it depends on who pays for the new power and grid upgrades, which is exactly why I think the companies should cover it.
PJM runs the grid for 13 states and Washington, D.C. Its independent market monitor, Monitoring Analytics, said in its third-quarter State of the Market report that including existing and forecast data center demand in PJM's capacity market has caused "an irreversible increase of $29.4 billion in capacity market costs for all customers to date," S&P Global reported on November 12. The monitor estimated that data center load pushed PJM's third-quarter wholesale power cost up 9.8%, from $101.66 to $111.61 per megawatt-hour, and warned December's auction could add another $5 billion to $10 billion.
Household prices are climbing too. The US Energy Information Administration expected the average residential customer to pay 18.2 cents per kilowatt-hour in 2026, up nearly 5% from 2025, in its May outlook, Utility Dive reported. EIA pointed to fuel costs and grid spending, including spending to serve rising demand.
To be fair, the evidence isn't one-sided. A Lawrence Berkeley National Laboratory study found that from 2019 to 2024, the states with the fastest load growth actually saw average real prices fall, because fixed grid costs were spread over more sales, according to the American Public Power Association's summary. But the effect was weaker for residential customers, and the researchers said it is unclear whether it holds under significant nationwide load growth, which is what the forecasts below point to. Our AI energy statistics page has the wider numbers.
How Much Power Do AI Data Centers Need?
A lot more than today, and fast. Global data center electricity demand rose 17% in 2025, and the International Energy Agency expects data center consumption to double by 2030, with power use at AI-focused facilities set to triple, according to the IEA.
In the US, data centers use 4% to 5% of electricity today and could use 9% to 17% by 2030, depending on how many proposed projects get built, according to EPRI's Powering Intelligence 2026 study.
The money behind that demand is enormous. Amazon, Alphabet, Microsoft and Meta have combined 2026 capital spending guidance of $720 billion to $740 billion, based on the figures in our AI spending statistics. Whether all that spending pays off is a separate question, which I covered in Will the AI Bubble Burst?
Who Should Pay for AI Data Center Power?
The tech companies should pay the full cost of the extra power their AI data centers need. Not ratepayers. If a company needs a new power plant, new transmission lines or a substation upgrade to run its data center, that bill belongs to the company, not to the family down the road or the small business on the same grid.
The companies have already said they agree, at least on paper. In March, Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI signed a White House "Ratepayer Protection Pledge," committing to cover the cost of the new power generation their data centers require so it isn't passed on to households, according to the White House. It is a voluntary pledge, not a law.
Congress tried to make it stick. On September 16, the House passed the Ratepayer Protection Act 417-3. It would require state regulators to consider standards so loads over 100 megawatts cover the "full, incremental" cost of the grid upgrades they need, Utility Dive reported. The Senate version was blocked before the election, per the Fortune analysis above.
Some states aren't waiting. Virginia's regulator approved a new rate class for customers needing at least 25 megawatts, effective January 1, 2027, under which they pay for most of the infrastructure and capacity reserved for them. Oregon regulators approved an extra cent per kilowatt-hour for Portland General Electric customers using 100 megawatts or more, and Texas requires new loads of 75 megawatts or more to help pay for grid connections, Newsweek reported. For how these rules fit with wider AI policy, see our AI regulation guide.
Why the Backlash Is Good News for Data Centers in Canada
I think the local backlash in the US is good for Canada. The projects still need to go somewhere, and Canada can pick up the ones other places reject.
The money is already moving. Meta announced a C$13 billion, one-gigawatt AI data center in Sturgeon County, north of Edmonton, its first in Canada, powered by a natural gas power plant, The Canadian Press reported in July. In September, the federal government welcomed Bell's planned expansion of its AI hub in Saskatchewan, which could add up to 900 megawatts of new capacity with a total capital investment of up to $52.5 billion, according to Innovation, Science and Economic Development Canada. And in Nova Scotia, Anthropic signed a 20-year lease for a data center in a $10 billion EverWind plan, with EverWind saying it will cover the full cost of energy connections and network upgrades, as we reported in our story on Anthropic's Nova Scotia data centre lease.
Some provinces are writing the "pay your own way" idea into the rules. Alberta's Data Centre Regulation, in force since June 2026, moves projects that bring their own new generation or storage to "the front of the connection line," and the province says new supply is "paid for by the companies that need it," according to the Government of Alberta. Demand is not the problem: companies have requested about 19,565 megawatts of new data center load in Alberta, against a provincial record peak of about 12,800 megawatts, per the same page. Ottawa's national AI strategy, launched June 4, says high-performance computing should be "sustainable and aligned with Canada's clean energy expansion," according to the Prime Minister's Office.
Canada isn't immune to the same pushback. Oakville, Ontario passed a one-year interim control by-law on August 11 that temporarily restricts new data center development while it studies energy supply, water and other impacts, according to the Town of Oakville. That's why who pays for the power matters in Canada too.
Would I Support an AI Data Center in My Own Town?
Yes, if it builds its own power. That's my condition. If a company brings its own generation instead of leaning on the grid my neighbors pay for, I'd support an AI data center in my own town.
This isn't a fantasy. Cleanview identified about 60 planned data centers with a combined 97 gigawatts of their own "behind-the-meter" generation into the early 2030s, including a 7.5-gigawatt plant powering Amazon data centers in Texas, AFP reported in August, as published by Malay Mail. Google completed its $4.75 billion purchase of Intersect, a developer that builds new power alongside data center load, PV Tech reported in March. And at the White House event, xAI committed to develop 1.2 gigawatts of power as its supercomputer's primary power source.
Self-supply has its own critics. AFP reported that conservative estimates put annual emissions from the planned plants above 200 million tons of CO2, and that the NAACP has sued xAI over turbines powering its Memphis data center. My test is about the power bill: if the data center builds and pays for its own power, it shouldn't land on everyone else's.
Free download: The AI Stats Sheet 2026. 36 key AI numbers for business on four printable pages, with the source under every figure. Join 2,000+ business leaders who read AI Business Weekly and get it in your inbox right away.
Frequently Asked Questions (FAQ)
What is an AI data center?
An AI data center is a facility filled with specialized chips, mostly GPUs, that train and run AI models. It uses far more electricity than a typical data center, and the newest campuses are planned at a gigawatt or more.
Why are people against AI data centers?
A big reason is cost, including worries about higher utility bills. On November 3, 2026, voters sided against data center development in 40 of 41 local ballot measures in 10 states, and NBC exit polls found 56% of voters called data centers a major or crucial factor in their congressional vote.
Do data centers raise electricity bills?
They can. PJM's independent market monitor said data center demand has added $29.4 billion in capacity market costs for all customers to date. Whether bills rise depends on who pays for new power and grid upgrades.
Who should pay for AI data center power?
In Sameer's view, the tech companies should pay the full cost of the extra power their data centers need, not ratepayers. Seven companies signed a voluntary White House pledge in March 2026 to cover the cost of new generation.
Is Canada building AI data centers?
Yes. Meta announced a C$13 billion, one-gigawatt data center near Edmonton, Bell plans up to $52.5 billion for up to 900 megawatts in Saskatchewan, and Anthropic leased a data center in Nova Scotia.
The Bottom Line
Demand for AI data centers is still growing, but the bill for their power shouldn't land on ratepayers. The companies building them should pay the full cost of the extra power they need. And if one builds its own power, I'd support it in my own town.
For Canada, the US backlash is an opening. If AI data centers pay their own way, Canada can pick up the projects other places turn down.
Related Articles
AI Energy Statistics 2026: power use, carbon and the nuclear deals behind AI.
AI Spending Statistics: Big Tech capex and global AI spending.
AI Bubble: What the Spending, Revenue and Debt Data Shows: the full data behind the debate.
Will the AI Bubble Burst?: Sameer's Take on who gets hurt and who won't.
Nvidia and AI Chips Hub: the chips inside AI data centers.
AI Regulation Guide 2026: the laws shaping AI in the US, EU and beyond.
⭐ Get more AI Business Weekly in Google: add us as a preferred source on Google.
Written by Sameer Khan, with support from the AI Business Weekly Editorial Team. AI-assisted and fact-checked against the sources linked above.

