Ottawa’s new national framework for artificial intelligence data centres is voluntary, and it gives the federal government no authority over where any of those facilities actually get built, according to the government’s own announcement.
Innovation, Science and Economic Development Canada launched Canada’s Responsible Data Centre Development Principles on Sept. 3, 2026. Evan Solomon, minister of artificial intelligence and digital innovation, announced the framework alongside Tim Tierney, president of the Federation of Canadian Municipalities. Twenty-three companies signed on, including OpenAI, Anthropic, Google, Meta, Microsoft, Amazon Web Services, Bell Canada, Telus, Equinix, Cohere and Quebec-based QScale.
The framework sets five expectations: that data centres create lasting local benefits, that they not shift electricity costs onto Canadians, that they minimize water use and environmental impacts, that they be transparent about local impacts, and that they bring strategic value to Canada.
None of that is enforceable. The department’s own release states the principles are “intended to complement existing provincial, territorial, municipal and Indigenous regulatory processes by setting a nationally recognized baseline that communities can use when considering major projects.” It also notes that development decisions “are fundamentally local, shaped by regional priorities, regulatory approvals and unique community needs.” In practice, a signatory that ignores every principle faces no federal penalty, because Ottawa is not the body issuing the permit.
That gap matters because the opposition the framework is meant to address is already organized. CTV News reported on Sept. 6 that the principles were failing to win over critics. Angus Reid Institute polling found more than two-thirds of Canadians oppose having a data centre built near their home, and a Leger poll of 1,505 Canadians published in July found 42 per cent opposed development in their own province against 44 per cent in support. Residents demonstrated outside Hamilton city hall against AI data centres on July 15, 2026.
There is a second point the announcement did not address. The federal government does hold one direct lever over large AI data centres, its call for proposals under the sovereign AI compute strategy, which sought projects with planned capacity above 100 megawatts and offered memoranda of understanding to successful proponents. That intake closed on Feb. 15, 2026, roughly six and a half months before the principles were published. Projects submitted into that process were therefore assessed against the department’s stated criteria on energy, Indigenous participation and environmental impact, not against the five principles announced this month. The government has not said whether the new principles will be applied to proposals already in that pipeline.
The scale involved is not modest. The Canadian Press reported on July 8, 2026 that a document prepared for Solomon to use in pitching international investors identified proposed projects that could take Canadian data centre capacity to 20 gigawatts. One development near Olds, Alta., planned by Synapse Real Estate Corp., includes 1.4 gigawatts of natural gas-fired generation on site.
For Canadian businesses and ratepayers, the second principle is the one with a testable outcome. Whether data centre load shifts costs onto other electricity customers will be decided in rate applications before provincial energy regulators in Alberta, Ontario, Quebec and Saskatchewan, not in a federal document that no signatory is obliged to follow.
