What Canada's AI Strategy Gets Right, and What It's Missing

March 2026 | midagent | John Pope

Canadian owned AI applications enable national sovereignty and economic security.

Let's begin with a statement that doesn't get made often enough in Canadian policy circles, because the temptation to critique is always stronger than the impulse to give credit: Canada's AI strategy is genuinely impressive.

Not impressive in the way that government press releases describe things as "impressive". Impressive in a measurable, internationally recognised, historically significant way.

Canada holds the distinction of being the first country to create a fully-funded national AI strategy. Storyboard18 That was 2017. While other G7 governments were still debating whether AI constituted a strategic priority, Canada was already funding researchers, establishing national institutes, and building the talent infrastructure that would become the envy of economies twice its size.

The results are not hypothetical. Canada ranks first among G7 nations in AI research publications per capita. Six percent of the world's most elite AI researchers — the top 0.5 per cent globally — call Canada home. Tradingkey Three of the people most responsible for the deep learning revolution that made ChatGPT, Gemini, and every other modern AI system possible — Yoshua Bengio, Geoffrey Hinton, and Richard Sutton — did their foundational work at Canadian universities, supported by Canadian public funding, at a time when the rest of the world thought neural networks were a dead end.

Hinton, who received the Nobel Prize in Physics in 2024, put it plainly: "Fundamental basic research often doesn't have an effect for many, many years." Canada bet on it anyway. We were right. Bloomberg

That bet deserves to be named, acknowledged, and built upon. Not because congratulations are the point, but because understanding what worked is the precondition for understanding what still needs to be fixed — and something important still needs to be fixed.

The Architecture of What's Been Built

https://www.youtube.com/watch?v=yiB6Wl8yCU4

The Pan-Canadian AI Strategy organises its activity across six interrelated priorities: talent and research; national AI institutes; commercialisation and adoption; standards and conformity; compute and infrastructure; and responsible AI and ethics. Fox Business This is not a superficial framework. Each pillar has received real money and produced real outcomes.

Budget 2024 committed $2.4 billion in new AI investments Storyboard18 — layered on top of the $125 million first phase and the $443 million second phase that preceded it. The money has gone toward a $2 billion AI Compute Access Fund and Canadian AI Sovereign Compute Strategy, $200 million to Canada's Regional Development Agencies for AI start-up support, and $100 million for the NRC's AI Assist Program for small and medium-sized businesses.

In November 2024, the Canadian Artificial Intelligence Safety Institute launched — a research and coordination centre mandated to advance understanding of risks from advanced AI systems, working with CIFAR, the National Research Council, and Canada's three national institutes: Mila, Vector, and Amii. Bloomberg

More than 125 Canada CIFAR AI Chairs are currently active, advancing cutting-edge research across AI safety, drug discovery, machine learning for health, and natural language processing. Fortune

The government has also moved, in the past eighteen months, with unusual speed on the infrastructure question. The Canadian Sovereign AI Compute Strategy commits $1 billion toward a national supercomputing facility. The $15 billion pension fund co-investment program — described in our previous post — creates a structural mechanism for private capital to build sovereign AI data centre capacity at scale.

In a policy landscape defined by chronic under-investment, regulatory paralysis, and institutional timidity, this is a genuine and meaningful body of work. It should be said plainly.

The Gap That the Strategy Has Not Closed

And yet Canada's AI talent is largely working for foreign corporations. Global Finance Magazine

Recent research shows the share of Canadian-founded companies being built in Canada has collapsed from 70 per cent to 32 per cent. Sevenpeakssoftware The researchers Canada trained, the ideas incubated at Mila, Amii and Vector, the graduate students funded by $250 million in annual federal scholarships — a staggering proportion of the commercial value they create flows to companies headquartered in Seattle, San Francisco, and New York.

Nearly two in three graduate researchers in Canada are considering leaving the country upon completion of their education. Kavout

It takes an AI company an average of eighteen months to begin building a product in Canada. In the United States, the same process takes four months. Wikipedia That is not a talent gap. That is a commercialisation gap — a structural failure to translate world-class research into global-scale Canadian tech companies — and it is the central unresolved challenge in Canada's AI strategy.

The AI Strategy Task Force convened by Minister Evan Solomon in the fall of 2025 was admirably direct about this in its published summary. Task Force members called for national AI adoption playbooks, dedicated digital transformation programs, and coordinated leadership to address systemic barriers including outdated IT systems, siloed data, and risk-averse purchasing behaviours. Yahoo! They emphasised, repeatedly, the need to protect and retain Canadian intellectual property, and to maintain domestic control over technological breakthroughs through ownership requirements and specialised public-private partnerships.

The diagnosis is now clearly established and broadly shared. The prognosis depends on what gets built next.

The Missing Layer

Here is the structural gap that the strategy has identified but not yet filled, stated as precisely as possible: Canada has invested heavily in the supply side of AI — research, talent, compute capacity, safety frameworks — but has not yet built the sovereign commercial infrastructure through which that AI capability can be deployed at scale, on Canadian terms, serving Canadian economic interests.

Consider what this means in practice.

A Canadian researcher at Mila develops a breakthrough in natural language processing. That research is published. A US technology company reads the paper, hires the researcher, and builds a commercial product. The product runs on AWS infrastructure. Canadian businesses pay to use it. The data those businesses generate trains the next version of the model. The value compounds — in California.

Canada invested in the research. Canada trained the researcher. Canada is now paying, through its businesses, to use the product. And the data intelligence generated by Canadian commercial activity is feeding a foreign AI system that Canada does not own, cannot audit, and has no legal authority over.

As intellectual property expert Jim Hinton put it, when the government announced its Cohere investment: "Simply buying compute capacity doesn't make Canada richer or better at generating, retaining and commercialising AI." Trade with Estonia

He is right. Compute is the factory. What matters is who owns the factory, who runs it, what it produces, and who captures the value of what it produces.

The C.D. Howe Institute's recent analysis identified the missing pillar directly: Canada has spent considerable effort on compute and infrastructure, but less attention has been paid to the data supply chains and commercial platforms through which AI value is actually created and distributed. Center for Data Innovation Without Canadian-owned, Canadian-governed commercial infrastructure through which AI-native commerce and government intelligence can operate, the compute strategy risks building a very expensive highway that Canadian cars cannot drive on.

What a Complete Strategy Looks Like

The good news is that the architecture of a complete strategy is now visible, and the policy levers to build it are all in place. They simply need to be pointed at the right target.

The Sovereign Compute Strategy provides the hardware layer. Mila, Vector, and Amii provide the research layer. Cohere and an emerging cohort of Canadian AI companies provide the model layer. CAISI provides the governance and safety layer. The pension fund co-investment program provides the capital structure.

What is missing is the application layer — the sovereign, open-standard, AI-native commercial and intelligence platform through which all of that upstream investment can be deployed in the real economy, generating returns that stay in Canada, training models on Canadian data that Canada owns, and providing government and business intelligence that operates within Canadian legal jurisdiction.

As the Hub's analysis of the strategy consultations concluded, open-source AI offers a guiding philosophy that prioritises experimentation, skills development, and organisational control — auditable models, transparent documentation, and the ability to adapt systems to Canadian legal, democratic, and cultural contexts. These benefits complement, rather than replace, commercial platforms. The result is not a trade-off, but a pipeline: open approaches expand participation at the front end, while commercial platforms help successful use cases scale across the economy. European Commission

That pipeline is the missing piece. And it is precisely what a sovereign, open-standard AI commerce and intelligence platform is designed to provide.

The Moment Canada Has Prepared For

Minister Solomon was right when he said countries that master AI will dominate the future. The question is not whether Canada has the ingredients to master it. We have been demonstrating for thirty years that we do. The question is whether we will build the commercial infrastructure that allows Canadian mastery to generate Canadian prosperity — or whether we will continue the pattern of funding the research, training the talent, and then watching the value migrate south.

Canada did not build the Trans-Canada Highway so that American trucks could profit from it. We did not invest in hydroelectric infrastructure so that American manufacturers could buy the power cheaply and sell the products back to us at a premium. The logic of sovereignty in physical infrastructure is well understood. It is time to apply the same logic to digital infrastructure — completely, coherently, and with the same national ambition that made Canada a world leader in AI research in the first place.

The Pan-Canadian AI Strategy got the foundation right. It gave Canada a decade's head start on every comparable nation. The researchers are here. The institutes are here. The capital mechanisms are in place. The policy intent is declared.

What remains is the will to build the final layer — the one that turns Canadian AI research excellence into Canadian AI economic sovereignty.

That layer is not a research grant. It is not a compute cluster. It is not a safety institute, as important as all of those are.

It is a platform. Canadian-owned, open-standard, sovereign by design, and built to ensure that the next thirty years of Canadian AI development generates prosperity for Canadians — not just for the foreign platforms fortunate enough to have been built first.

We have done the hard part. We invented the technology. We trained the world's researchers. We built the institutes. We wrote the first national AI strategy on the planet.

All that remains is to own what we built.

No more excuses, Canada. Elbows up.