Canada built the intellectual foundations of modern artificial intelligence. Geoffrey Hinton trained at the University of Toronto. Yoshua Bengio built Mila in Montreal into one of the most cited AI research institutions in the world. Vector Institute, Amii, and Mila collectively anchor a research ecosystem that the OECD has rated among the top five globally. The commercial return on that investment has been, to put it plainly, disappointing. Canada contributed C$223 billion in digital GDP in 2024. Only 12% of Canadian businesses currently use AI to deliver goods or services, the lowest rate in the G7. The country that built the tools is not using them.
This is not an accident. It is the result of a series of institutional and policy failures that have compounded over a decade. The Scientific Computing and Artificial Intelligence Program committed C$890 million to sovereign compute infrastructure, a genuine structural investment. But the broader commercialization architecture has not kept pace. Bill C-27, Canada's AI and privacy legislation, has been stalled since 2022 and remains unresolved in 2026. Without regulatory clarity, enterprise buyers defer AI procurement and Canadian AI companies face the same compliance uncertainty as their customers. The regulatory vacuum does not slow adoption uniformly: it slows domestic adoption while US and EU competitors operate under settled frameworks and capture the deals Canadian companies are positioned to win.
The talent story compounds the adoption story. Cohere, one of Canada's most significant AI companies and a direct product of the Toronto research ecosystem, merged with Germany's Aleph Alpha and structured the resulting entity around European operational headquarters. The story is not that Cohere left Canada. The story is that the structural incentives, compensation competitiveness, regulatory predictability, and institutional support, did not make staying the obvious choice. Meanwhile, Canada surrendered approximately C$7 billion in digital services tax revenue by deferring its DST to accommodate US trade negotiations, a concession that generated goodwill without generating equivalent returns. The DST capitulation did not produce a better trade framework for Canadian tech exports. It produced a gap in the fiscal architecture that was meant to fund domestic digital investment.
The trajectory is not inevitable. Canada's C$249 billion digital GDP projection for 2030 is achievable, but only if the gap between research capacity and commercial capture closes. That requires three things. First, SR&ED reform that actually reaches scale-up companies, not just early-stage research projects. The current programme architecture skews toward small companies doing basic R&D; the companies that generate export revenue are larger and faster-moving and the programme does not serve them well. Second, compute access parity: the SCIP investment needs to be operationalized in ways that give mid-market Canadian AI companies access to training infrastructure without requiring US hyperscaler dependency. Third, regulatory certainty. Bill C-27 needs to pass. iGaming and digital currency operators need a settled framework. Semiconductor import policy needs to explicitly protect Canada's Tier 1 designation. None of this is complex. All of it requires political will that has been intermittently available and inconsistently applied.
Canada's research advantage is real and defensible. The commercialization gap is also real and widening. The choice is not between ambition and caution. It is between acting on the structural investments already made and watching the returns flow elsewhere.