Canada’s AI for All Plan Bets on Sovereign Compute

Prime Minister Mark Carney launched AI for All on June 4, 2026, pairing 90,000 youth placements with a public supercomputer. The five-year plan also targets 250,000 AI-related jobs and a jump in business adoption from just over 12 percent to 60 percent by 2034.

Three months later, Ottawa has started a $13 million literacy program in Edmonton, while the costly work remains a national machine, gigawatts of data-centre power, and a fight over who owns the stack.

We just launched Canada’s new AI Strategy: AI For All. 
 
We’re taking control of our future — with AI that’s governed by Canadian values, AI that’s accountable to Canadians, and AI that serves all Canadians.

— Mark Carney (@MarkJCarney) June 4, 2026

Two Job Numbers, One Five-Year Plan

The Prime Minister’s Office put both figures in the same release, and they are not the same promise. One is an economic-growth target of 250,000 new AI-related jobs over five years, tied to an extra $200 billion of output. The other is up to 90,000 AI-related jobs and work placements for young Canadians.

Carney framed the choice as distribution, not invention. Canada already employs about 150,000 people in AI work across more than 3,500 firms, and the digital sector as a whole accounts for about 800,000 workers and more than $140 billion of GDP, according to the strategy published by Innovation, Science and Economic Development Canada.

AI is here. The question is whether it will improve the lives of all Canadians or benefit only a few. AI can shorten our emergency room wait times and make a small business more competitive, if it is governed by Canadian values with a clear goal of improving the lives of all Canadians. That’s why we need an ambitious new strategy: AI for All.

Mark Carney, Prime Minister of Canada, June 4, 2026 launch

The same document says Canadian firms have raised more than CAD$37 billion in venture funding for AI products. The political problem is not a missing lab. It is that those labs have not pulled the rest of the economy with them, and that firms that do adopt often froze hiring around AI tools instead of rewriting the work.

Ottawa’s answer is a National AI Literacy Initiative, trusted AI agents for every post-secondary student, and a pledge to use federal buying power as an anchor customer for Canadian vendors. Minister of Artificial Intelligence and Digital Innovation Evan Solomon called that mix trust, opportunity, and sovereignty. The spend sheet underneath those words is mostly machines.

A Public Supercomputer by 2031

The strategy’s sovereignty chapter is the part with a delivery date and a procurement vehicle. Canada will build a public AI supercomputer by 2031 so researchers and smaller firms can run work on Canadian terms, and it will back data centres that can scale to at least 100 megawatts. ISED’s own analysis says commercial players will need 5.5 gigawatts of AI compute by 2030.

Partnerships now being finalized, the department wrote, have proposed 850 megawatts by 2030, with room to scale toward 2.3 gigawatts. That is the industrial plan sitting behind the jobs headline. Chip fabrication still sits almost entirely offshore, which the strategy admits, so “sovereign” here means ownership, jurisdiction, and power, not a Canadian GPU plant.

The AI Sovereign Compute Infrastructure Program is the nearest cash instrument for the public machine. ISED set an $890 million funding envelope from Budget 2024 and Budget 2025, wants a single lead recipient or consortium, and expects significant service within 18 months of a contribution agreement. That 18-month clock is not the 2031 supercomputer date. One is a build-and-run contract. The other is a national target.

NAMED FUNDING ENVELOPES

Those lines are named envelopes, some of them restated from earlier budgets, and they should not be added into a single new pile. The Prime Minister also pointed to a $500 million Tech Growth Fund that can take equity, a $700 million top-up to the Compute Access Fund, and a $500 million Regional AI Initiative, in the launch speech that accompanied the release.

Foreign hyperscalers are still welcome where Ottawa sees a Canadian benefit. The strategy says as much. The political product is a Canadian-controlled slice beside that import market, and the early deals show who can book that slice. A Canada-Saudi arrangement has already handed Cohere 50 megawatts of compute, which is how sovereignty looks when it is a power contract rather than a slogan.

The 2017 Strategy Already Bought the Talent

Canada was first to a national AI strategy in 2017, when the Pan-Canadian Artificial Intelligence Strategy put $125 million through the Canadian Institute for Advanced Research from 2017 to 2022. Phase two, launched in 2021, added $443.8 million through 2031, including $208 million for CIFAR and $40 million for the Digital Research Alliance of Canada to add research compute.

An ISED evaluation dated January 2025 found that money did what talent money does. Canada CIFAR AI Chairs reached 129 in 2023-24, against a target of 135 by March 31, 2026, and 84 percent of chairs surveyed called the second phase extremely important. The same review listed the holes that AI for All now tries to fill: weak commercialization, IP that leaves the country, foreign-heavy venture capital, and compute funding that had “yet to hit the ground.”

FROM TALENT GRANTS TO A MACHINE

1. 2017: Ottawa launches PCAIS 1.0 with $125 million to recruit and keep researchers at Amii, Mila, and the Vector Institute.
2. 2021: PCAIS 2.0 adds $443.8 million through 2031 for commercialization, standards, talent, and research compute.
3. June 4, 2026: Carney launches AI for All, adding a 2031 public supercomputer and the 90,000 youth-placement target.
4. Early September 2026: The government stands up Digital Transformation Canada to buy and use technology, including AI, as an anchor customer.
5. Sept. 9, 2026: Solomon launches the literacy program in Edmonton with Amii.
6. Sept. 21, 2026: Student and educator courses are due to open, with a public course later in 2026.

The research brand is not in dispute. The strategy itself names Nobel laureate Geoffrey Hinton and Turing Award winners Yoshua Bengio and Richard Sutton. What the 2017 play did not buy is broad use inside ordinary firms, or a Canadian-owned training cluster large enough to stop labs from renting time in Virginia.

Why Most Canadian Firms Still Skip AI

The government’s working baseline is blunt. Statistics Canada found that 12.2 percent of firms used AI to produce goods or deliver services in 2025, double the 6.1 percent recorded a year earlier, and that is the “just over 12 percent” the June plan uses. Among small and medium firms, ISED puts adoption at about 8 percent, against 29 to 42 percent in the Nordics, 26 percent in Germany, and 18 percent in France.

A later Statistics Canada reading, covering the 12 months before the second quarter of 2026, found 19.2 percent of businesses used AI in production or services, triple the 6.1 percent of the second quarter of 2024. That survey window mostly predates AI for All, so it is not a report card on Carney’s plan. It does show the floor moving before the literacy course existed.

ADOPTION SNAPSHOT

• 12.2 percent: Share of firms that used AI to make goods or deliver services in 2025, per Statistics Canada, the baseline baked into AI for All.
• 19.2 percent: Share in the 12 months before the second quarter of 2026, still far from the 60 percent target for 2034.
• About 8 percent: SME adoption cited in the federal strategy, the group the Compute Access Fund is meant to reach.
• 40.0 percent: Share of businesses in the later survey that said AI is not relevant to them.

That last figure is the objection the launch event did not dwell on. Free chatbots are already on every laptop. A firm that says the tools are irrelevant is not waiting for a password. It has not found a job the tools should do, or it does not trust the output enough to put them near a customer or a payroll file.

The global market the Prime Minister’s Office cited is projected at U.S.$4.8 trillion by 2033, which is the prize used to justify speed. Canada’s slower uptake, the June release warned, risks sending talent and startups abroad and leaving “critical parts of our AI ecosystem under foreign control.” Consultations in 2025 drew more than 11,000 submissions and a 28-member task force. The public asked for safety and jobs. The budget that followed still reads like a power-and-procurement plan.

Amii’s Free Courses Open on Sept. 21

The classroom piece is the first part of AI for All that a student can actually touch. On Sept. 9, 2026, in Edmonton, Solomon launched the National AI Literacy Initiative as a $13 million partnership with the Alberta Machine Intelligence Institute. Cam Linke, Amii’s chief executive, stood with him at Amii headquarters, alongside Edmonton-Riverbend MP Matt Jeneroux.

The June plan had already promised entry-level training for all Canadians and a reach of one million post-secondary students. The Edmonton event put a vendor, a dollar figure, and a start date on that pledge. Student and educator material is due on Sept. 21, 2026. A course for the wider public is due later in 2026, before year-end. Modules run about three hours, without a live instructor, on a simple ladder: understand AI, use AI, and, for some, build with it.

THE THREE FREE STREAMS

• Essentials for Students: Aimed at up to one million post-secondary students, with the first course due on Sept. 21, 2026.
• Essentials for Educators: Written for more than 50,000 kindergarten-to-Grade-12 teachers, opening with a first chapter on Sept. 21 and more chapters through the fall.
• Essentials for Canadians: A public course, delivered later in 2026 through community partners, covering bias, misinformation, and privacy loss.

Solomon said education is how a country gets through a technology shift. Linke treated the contract as a scale-up of work Amii was already doing, not a new invention.

This national mandate is a profound validation of Amii’s proven track record in AI literacy, and we are deeply proud to answer this call.

Cam Linke, CEO, Alberta Machine Intelligence Institute, Sept. 9, 2026

$13 million is a rounding error next to the $890 million SCIP call. That is the point of the second-order read. The course is what the public can see this month. The machine is what the strategy is actually buying, and it still has no named operator.

Local Power Grids Are the Binding Constraint

ISED’s own value-chain sketch is candid about electricity. Canada’s grid is clean and predictable, the strategy says, and currently constrained. A companion electricity plan, Powering Canada Strong, talks about doubling electricity kit by 2050 so data centres can sit on sustainable baseload. Until that kit exists, every new 100-megawatt hall is a fight with a utility, a town, and a ratepayer.

Since March 2025 the government has signed AI and technology statements with Australia, the European Union, Finland, Germany, India, Norway, Qatar, Saudi Arabia, Spain, Sweden, the United Arab Emirates, and the United Kingdom, and it has stood up a Sovereign Technology Alliance. Those deals are how Ottawa wants to move “from reliance to resilience.” They also confirm that Canadian compute will be built with foreign capital, foreign chips, and, in some cases, foreign demand on Canadian rivers and gas plants.

In early September 2026, Carney launched Digital Transformation Canada, a new federal body to buy and run digital systems, including AI, and to use procurement as a test bed for Canadian firms. That is the demand side of the same bet: if the public supercomputer and the 850-megawatt partnership slate are the supply, the government itself is supposed to be a customer that keeps the work onshore.

Safety law is the other half of the trust pitch. The June plan promises updates on personal information, deepfakes, algorithmic manipulation, surveillance pricing, and an online-safety regime for chatbots and social platforms, plus a larger role for the Canadian AI Safety Institute in model evaluations. The Protecting Victims Act, tabled in December 2025, already tries to ban non-consensual sexual deepfakes. There is still no single AI statute on the books.

WHERE EXPERTS DISAGREE

• Ottawa’s frame: Carney and Solomon treat sovereignty over compute, data, and models as the condition for trust and for keeping value at home, with a public supercomputer as the flagship.
• The integration critique: Sean Speer, a policy writer at The Hub and the Manhattan Institute, argued after the launch that Canada cannot recreate a frontier stack against American and Chinese capital budgets, and that the useful strategy is to become essential inside one of those stacks.
• The adoption critique: A three-hour public course can teach vocabulary. It cannot give a construction firm or a diner a reason to rebuild its workflow, which is why 40.0 percent of businesses still call AI irrelevant.

Speer wrote that “Canada’s AI future is more likely to depend on integration than independence,” and that a homegrown frontier stack is “an unserious starting point.” That argument does not cancel the 5.5 gigawatt planning figure. It asks who those watts will actually serve once they are plugged in.

The next date that is real for a student is Sept. 21, 2026, when Amii’s first courses are due to go live. The next date that is real for the strategy as written is a contribution agreement for one national machine, then 18 months of installation, then a public supercomputer that the plan still places in 2031.