Everything About AI
NextFin News - Prime Minister Mark Carney's drive to make Canada an AI data-center superpower is colliding with a wall of local opposition, and the federal government's answer is a five-point rulebook that companies can sign but nobody has to obey. On September 3, 2026, Artificial Intelligence Minister Evan Solomon unveiled a national framework for responsible data-centre development in Markham, Ontario, backed by more than 20 signatories including OpenAI, Anthropic, Amazon Web Services, Microsoft, Google and Meta. The problem: the framework is not legally binding, and the backlash it is meant to calm is rooted in something harder than public relations — the physical limits of Canada's electricity grid.
The Framework and the Backlash
The tension is easy to state and hard to resolve. Ottawa wants data centers because AI infrastructure is the fastest-growing source of industrial electricity demand in a generation. A government pitch deck prepared for Solomon's office, obtained through access-to-information and published in July 2026, put the scale in blunt terms: Canada has about 337 megawatts of AI data-center capacity today, with more than 20 gigawatts — 20,000 megawatts — in projects "under planning or development." That is a roughly 60-fold gap between what exists and what is on drawing boards.
The natural-resources ministry's own "Powering Canada Strong" strategy, released alongside the government's National Electricity Strategy, projects that electricity demand from AI data centers could reach three to five gigawatts by 2030 and as much as 10 gigawatts by 2050. For comparison, U.S. data centers are projected to account for 8.6% of all American electricity demand by 2035, more than double the 3.5% share today. Canada's total electricity demand is expected to double by 2050. The math is unforgiving: a single hyperscale facility can consume as much power as a mid-sized city, and the country is being asked to build dozens of them while households, factories and electric vehicles compete for the same electrons.
Solomon's framework tries to defuse the politics without touching the physics. Developers must minimize water use and other environmental effects, be transparent about local impacts, not shift electricity costs onto Canadians, create lasting local benefits, and bring "strategic value" to the country. The principles were developed with the Federation of Canadian Municipalities and input from industry, Indigenous groups and environmental organizations, and are meant to "complement existing provincial, territorial, municipal and Indigenous regulatory processes by setting a national recognized baseline," according to the government's news release.
That last phrase is doing a lot of work. The framework does not override municipal zoning, provincial grid-connection rules, or local environmental assessments. It sets a baseline that municipalities and provinces can hold companies to — but only if those municipalities and provinces already have the power to do so.
"I'm not trying to fight with councillors and cities who have real questions," Solomon said in an interview. "Maybe some city councillors haven't had a good assurance that there's an answer, and so then they're like, 'Let's have a moratorium.'"
The moratoriums are real, and they are spreading. In Manitoba, Premier Wab Kinew rejected a proposed 500-megawatt gas-fired data center southeast of Winnipeg in June 2026, calling it "a big threat to the environment and not much benefit to the economy." The project, from Las Vegas-based Jet.AI and Vancouver's Consensus Core, would have sat on 142 hectares of farmland near Ile des Chênes and burned natural gas around the clock. In Hamilton, Ontario, city councillors advanced steps toward a 12-month moratorium on AI data centers after hundreds protested a proposed National AI Compute Facility on the city's harbourfront. Oakville, Ontario, introduced a one-year moratorium in August 2026, and Mississauga's council moved to consider a similar pause. Quebec, Ontario and British Columbia have all taken steps to restrain runaway data-center development, particularly around access to electrical capacity.
Public opinion is not on the government's side either. A Leger poll of 1,505 Canadians published in July 2026 found 42% oppose data-center development in their own province, compared with 44% who support it — a near-even split in which Albertans were the most likely to strongly oppose development, at 31%. That Alberta finding matters, because Alberta is simultaneously the epicenter of the federal government's data-center ambitions and the province where opposition is hardest.
Why Alberta Became the Battleground
The heart of the contradiction sits in Alberta. The province is aggressively courting AI infrastructure with an Artificial Intelligence Data Center Strategy that aims to attract $100 billion in investment. An August 2025 government document, labeled "commercially confidential," reported that projects to date amounted to more than 18 gigawatts — nearly the entire national pipeline — and that this capacity would be supported by natural-gas power plants rather than the provincial grid. The same document calculated that if 18 gigawatts were added on gas, it could generate 20 million tonnes of CO2-equivalent emissions, or almost 3% of Canada's total emissions in 2023.
The flagship project makes the trade-off concrete. On July 8, 2026, Meta announced its first data center in Canada — a C$13-billion campus in Sturgeon County, about 35 kilometers north of Edmonton, and the company's largest facility outside the United States. The campus starts at one gigawatt of power and can scale to 1.8 gigawatts, making it one of the largest data-center developments under construction anywhere outside the U.S. It will be powered by the Greenlight Electricity Centre, a 932-megawatt combined-cycle natural-gas plant being built by a consortium of Pembina Pipeline, Morgan Stanley Infrastructure Partners and Kineticor Asset Management, with a long-term tolling agreement signed by Meta. The plant is scheduled to come into service in late 2030; until then, Calgary-based Capital Power will supply 250 megawatts from its existing gas-fired fleet for the next decade. The facility will require roughly 150 million cubic feet of natural gas per day, Pembina said.
The Alberta government's sales pitch is that this model costs taxpayers nothing and actually lowers bills. The province says the Meta project alone will deliver about C$250 million a year in royalties, taxes, levies and transmission fees, create more than 3,000 construction jobs at peak and about 300 permanent positions, and — because large customers pay the same transmission fees as everyone else — could reduce the transmission portion of household power bills by up to 6%. "Built with private money, powered by new private generation, and paying its own way," the province's project site says.
Meta's own framing leans on community benefit.
"We believe that the success of a data centre is only possible when the community itself succeeds along with it," Gary Demasi, Meta's vice-president of data centre strategy and development, told a news conference in Calgary on July 8. "More than that, we want Sturgeon County and Alberta to thrive."
But not everyone in Sturgeon County is convinced. In Olds, Alberta, residents have rallied against a separate proposed 10-building campus with 1.4 gigawatts of gas-fired generation from Synapse Real Estate. And the environmental critique is sharper still. Keith Stewart, a senior energy strategist with Greenpeace Canada, argued that Prime Minister Carney could have insisted that data centers be powered by renewable energy. Instead, he said,
"This makes it clear that one of the main reasons that the Carney government wants to expand AI data centres is that so Big Oil can sell more gas."
The criticism lands because it exposes the central contradiction in Ottawa's position. The same government that launched a National Electricity Strategy promising to double power capacity by 2050 is also promoting an AI strategy — "AI for All," unveiled June 4, 2026 — that the Privy Council documented as requiring roughly 5.5 gigawatts of AI compute by 2030, with about 15.5 gigawatts of new gas capacity listed in planning documents for data centers. Carney's strategy language promises to "link new data centre development with clean energy expansion, robust environmental standards, and tangible benefits for local communities." The pipeline says otherwise: gas is what is actually available, at the speed the market wants.
The Second-Order Problem: Jurisdiction, Not Just Power
The first-order story is electricity. The second-order story is who gets to decide. Canada's Constitution gives provinces authority over natural resources and electricity generation; Ottawa controls trade, taxation and federal environmental assessment. A voluntary federal framework cannot force a municipality to rezone farmland, cannot compel a province to release grid capacity, and cannot stop a premier from saying no.
This is why the backlash is structurally durable rather than a temporary public-relations problem. Opposition to data centers is not a single-issue protest that will fade after a town hall. It is the collision of three jurisdictional layers — municipal land use, provincial power allocation, and federal industrial policy — each with its own electorate and its own incentives. A mayor facing a moratorium petition answers to residents worried about noise, water and property values. A premier controlling grid connections answers to voters worried about household power bills. A federal minister answering for AI strategy answers to a national productivity agenda. None of them can overrule the others, so the path of least resistance is delay.
The delay has a cost, and it is measured in investment share. Data-center developers buy two things: land and power. Alberta offers both quickly, with gas that does not need to wait for transmission lines or clean-generation buildout. Ontario and Quebec offer cleaner grids but tighter capacity and higher scrutiny. British Columbia has hydro but limited surplus. The framework does not change this arithmetic; it merely asks developers to be better neighbors while provinces decide who gets the electrons.
There is also a credibility gap in the "local benefits" pillar. The Alberta government's claim that household transmission bills could fall up to 6% rests on spreading fixed grid costs across more large customers — a claim that holds only if those customers actually connect and pay. The Manitoba premier's rejection rested on the opposite calculation: that the energy consumed and community impact "outweigh the limited gains." Both sides are doing the same cost-benefit math and arriving at different answers because they weight the same inputs differently. A 500-megawatt facility consumes roughly as much electricity as 400,000 households by the Electric Power Research Institute's rule of thumb — 100 megawatts equals about 80,000 homes. For a province like Manitoba, whose identity is built on clean hydro and whose rural communities are small, that trade is existential. For Alberta, whose economy is built on selling energy, it is a new customer.
The Counter-Thesis: Alberta Is Right, and the Backlash Is Noise
The strongest case against the pessimistic read is straightforward: the market has already voted, and it voted for gas. More than 20 companies signed Ottawa's framework on day one, including the world's largest cloud providers. Meta committed C$13 billion in a single announcement. Alberta's model — private, behind-the-meter generation that does not touch the public grid or the public balance sheet — solves the two problems that actually matter to developers: speed to power and cost certainty. If the alternative is waiting a decade for clean transmission in Ontario or Quebec, gas in Alberta wins every time, backlash or not.
There is force in this argument. The voluntary framework's defenders note that public commitments create accountability even without legal force: municipalities and provinces can hold signatories to the principles they signed. And the economic logic is hard to dismiss — a C$13-billion investment, C$250 million a year in fiscal benefits, 3,000 construction jobs. In a productivity-starved economy, that is real money.
But this thesis fails on one test: timing. The 18 gigawatts of Alberta projects are not built; they are announced. The Greenlight plant does not deliver power until late 2030. Between now and then, every project must survive municipal zoning, provincial environmental review, and — in a country where elections happen regularly — the possibility that today's welcoming premier is replaced by tomorrow's skeptical one. Manitoba's reversal did not come from a change in the economics; it came from a change in political judgment. The same switch can flip in Alberta.
The falsifying signal is specific: if Alberta secures final approvals and construction starts on more than half of the roughly 20-gigawatt national pipeline by 2028, while Ontario, Quebec and British Columbia continue to restrain connections, then the backlash is indeed noise and the gas-powered model wins by default. If instead approvals stall and the 20 GW pipeline shrinks materially, the constraint is real and the framework's voluntarism is the problem.
What Comes Next
The short-term outlook is more friction. Municipal moratorium votes will continue through 2026 and 2027; each one is a discrete binary risk for a specific project. The framework gives Ottawa something to point to in those fights, but not a veto. Developers will keep signing the principles because the cost of signing is zero and the cost of not signing is losing access to federal goodwill.
The medium-term outlook turns on the grid. The National Electricity Strategy's promise to double capacity by 2050 does not help a data center that needs power in 2029. The binding constraint for the next five years is not money — hyperscalers have plenty — but megawatts and permitting timelines. Projects paired with dedicated generation, like Meta's, will advance; projects dependent on public-grid connections will queue.
The long-term outlook is structural, and it favors the provinces that can reconcile clean power with speed. If Canada cannot build clean capacity faster than data-center demand grows, gas will fill the gap regardless of what the framework says — and the government's climate commitments will absorb the emissions. If clean buildout accelerates, the backlash loses its strongest argument and the framework becomes a genuine baseline rather than a wish list.
Three signals will tell the story first: the final disposition of Hamilton's moratorium and similar municipal votes; the pace of grid-connection approvals in Ontario and Quebec relative to Alberta's gas projects; and whether the 20-gigawatt pipeline survives contact with actual environmental assessments. The first measures political will, the second measures infrastructure reality, and the third measures whether the framework's principles mean anything when they stop being voluntary.
Carney's data-center push is not failing because Canadians dislike AI. It is failing to land because the federal government is asking municipalities and provinces to host the infrastructure of a national industrial strategy without giving them either the power to stop it or the incentive to welcome it. A rulebook that everyone signs and nobody enforces is not a policy; it is a hope dressed as a framework. The electrons, unlike the signatures, will have to be real.
Explore more exclusive insights at nextfin.ai.
