Canada contributed C$223 billion in digital GDP in 2024, leads the G7 in AI research, and holds the lowest publicly available AI compute infrastructure of any G7 country. Only 12% of Canadian businesses use AI to deliver services. The gap between what Canada has built in research and what it has captured commercially is the central digital economy policy challenge of the decade.
Canada's digital economy contributed C$223.45 billion to GDP in 2024, with 12% of the Canadian workforce employed in digital sectors and the ICTC projecting C$249.65 billion by 2030.1 Canada was the first country in the world to introduce a national AI strategy, in 2017, and has produced world-class AI research through the Vector Institute, Mila, and the Alberta Machine Intelligence Institute. Its universities train AI researchers who are recruited globally. Its cities host R&D operations for Google, Microsoft, Amazon, and virtually every major technology company. And yet, as of late 2025, only approximately 12% of Canadian businesses report using AI to produce or deliver services, and Canada holds the lowest amount of publicly available computing infrastructure and performance of any G7 country.2 The gap between research excellence and commercial application is one of the most consequential structural weaknesses in the Canadian economy.
The sovereignty dimension of this gap is not rhetorical. Policy Magazine's January 2026 analysis argues that Canada faces a binary choice between building the data infrastructure and compute capacity that would allow it to participate meaningfully in the AI era, and becoming what the analysis calls a digital colony, subservient to AI infrastructure and large language models controlled by a small number of companies in Silicon Valley and Beijing.3 Canada cannot compete with the United States or China on foundational model development. The resources required, estimated at over US$100 billion per model generation, are simply unavailable in the Canadian public or private economy. But Canada does not need to compete on foundational models to have meaningful digital sovereignty. It needs to build compute infrastructure for inference and deployment, establish data governance frameworks that ensure Canadian data generates value in Canada rather than exporting it to US platforms, and develop the commercialization pathways that translate Canadian AI research into Canadian companies.
The Digital Services Tax episode illustrates the sovereignty challenge concisely. Canada enacted a 3% DST on the Canadian revenue of large digital platforms in 2024, after years of negotiation and OECD commitments that had not materialized into a multilateral framework. The United States government then launched a CUSMA dispute against the DST, framing what was effectively corporate lobbying by Meta, Google, Amazon, and Apple as a US national interest trade grievance. Under US pressure, the Carney government announced repeal of the DST in 2025, sacrificing C$7 billion in projected five-year revenue and, more significantly, demonstrating that Canada's ability to independently regulate its digital economy is constrained by the political leverage that US-based platform companies can bring to bear through the trade relationship.4
This is not primarily a story about corporate tax. It is a story about whether Canada can govern its own digital economy. The Online News Act, the Online Streaming Act, the Bill C-27 data protection legislation that remained stalled in committee through multiple parliamentary sessions, and the DST repeal form a pattern: Canada attempts to regulate the digital economy in the Canadian public interest, and US-based platforms mobilize trade leverage to neutralize those regulations. The legal and policy framework for Canadian digital governance needs to be built and defended with the same strategic clarity that Canada applies to resource sector sovereignty. The two are increasingly similar problems.
The compute infrastructure deficit is both an economic and a sovereignty problem. Torys' April 2025 analysis documents that Canada trails every other G7 country in publicly available AI computing infrastructure and performance, based on The Dais's assessment for Toronto Metropolitan University.2 The federal AI Compute Challenge and the C$240 million Cohere investment are partial responses to this gap, but as the researchers Tusikov and Haggart observe at Policy Options, simply buying compute capacity in Canada does not make Canada richer or better at generating, retaining, and commercializing AI if the compute infrastructure is operated by a US firm and the IP produced on it flows to US corporate owners. The distinction between data sovereignty, compute sovereignty, and IP sovereignty is crucial for Canadian policy, and it is a distinction that current federal policy has not fully made.
Indigenous data sovereignty is a distinct and important dimension of this broader conversation. The FNIGC's First Nations Data Governance Strategy and the OCAP principles (Ownership, Control, Access, and Possession) established by First Nations researchers represent the most developed framework for Indigenous data governance in Canada. The Government of Canada's digital sovereignty framework acknowledges that Indigenous data sovereignty is "led by Indigenous partners through separate processes," but the integration of OCAP principles into federal data governance, AI procurement, and digital infrastructure investment has been slow and partial. When government AI systems make decisions affecting Indigenous people, including in immigration, social services, and resource management, the data on which those systems are trained and the values they embed are sovereignty questions that deserve the same governance attention as territorial sovereignty.
The gender dimension of digital economy access and participation is structural. Women are underrepresented in computing, AI research, and technology leadership in Canada, as in most comparable countries. This is not primarily a pipeline problem. Research consistently shows that the gap is created and sustained by workplace cultures, compensation practices, and promotion structures in the technology sector that disadvantage women, particularly women of colour, Indigenous women, and women with disabilities. A digital economy that excludes half the population from its highest-value roles is both a social inequality problem and an economic efficiency problem: Canada's ability to compete in AI and technology at the scale the opportunity requires depends on drawing on the full range of Canadian talent.
Digital sovereignty is the ability of Canada to exercise autonomous control over its digital infrastructure, data, and intellectual property. The Government of Canada's framework on digital sovereignty, published in 2025, defines this for internal government operations. The broader question, whether Canada can govern the digital economy as a whole in Canadian interests, involves different actors, different legal frameworks, and considerably more political complexity.
The concentration of the digital economy in US-based platforms creates a structural sovereignty problem that is qualitatively different from other trade dependencies. When Canada buys steel from the United States, it buys a product and retains ownership of the steel. When Canadians use US digital platforms, the data generated by that use flows to US corporate owners, is used to train AI systems owned by those corporations, and generates advertising and service revenue that is taxed in whatever jurisdiction those corporations prefer. Canada's attempt to capture a portion of that value through the DST, the Online News Act's news media bargaining provisions, and the Online Streaming Act's Canadian content requirements all represent variations of the same policy challenge: how to ensure that economic activity occurring in Canada generates value for Canadians, not just for US platform shareholders.
The data centre investment debate illustrates the complexity. When the federal government invested C$240 million in Cohere, a Toronto-based AI firm, and CoreWeave built a data centre in Cambridge, Ontario to support it, the optics were of Canadian AI investment. But as intellectual property expert Jim Hinton observed in Policy Options, the US cloud provider building the facility retains ownership of the computing infrastructure, and the AI models trained on that infrastructure may have IP ownership structures that do not benefit Canada. Simply buying compute capacity does not make Canada richer if the value created on that capacity flows elsewhere. The policy distinction that matters is between investing in Canadian-owned compute infrastructure, which builds Canadian assets and retains IP in Canada, and subsidizing US-owned compute infrastructure in Canada, which creates jobs but not lasting Canadian economic assets.
Bill C-27, the federal data protection legislation that would have updated Canada's privacy framework and introduced AI governance provisions, remained stalled through multiple parliamentary sessions and has not yet been enacted. This represents a significant governance gap. Canada's current data protection framework, the Personal Information Protection and Electronic Documents Act enacted in 2000, predates the smartphone, social media, cloud computing, and generative AI. Updating it is not a regulatory preference. It is a basic precondition for Canadians to have meaningful control over their personal data in the current digital environment. Quebec's updated privacy law, enacted in 2021 and in force since 2022, provides a provincial model that federal legislation has not yet matched nationally.
Canada's AI research infrastructure is genuinely world-class. The Vector Institute in Toronto, Mila in Montreal, and the Alberta Machine Intelligence Institute collectively represent one of the highest concentrations of AI research talent outside of the major US and Chinese technology centres. Geoffrey Hinton's foundational work on deep learning was done at the University of Toronto. Yoshua Bengio's recurrent neural network research at Universite de Montreal has been central to modern AI development. The academic lineage of Canadian AI research is exceptional.
The commercialization gap is equally real. Canadian AI researchers have consistently left for better-resourced positions at US technology companies, taking their intellectual capital and research directions with them. The firms founded by Canadian AI researchers have typically scaled in the United States, not Canada, because US capital markets, US corporate customers, and US compute infrastructure are more accessible. The result is that Canada generates a disproportionate share of AI research talent and captures a disproportionately small share of AI commercial value. This is the same dynamic as the critical minerals value-added gap: Canada produces the raw material and exports it for value to be created elsewhere.
The Pan-Canadian AI Strategy's second phase, focused on commercialization rather than research, represents the federal government's attempt to address this gap. The AI Compute Challenge, the Cohere investment, and the broader compute infrastructure agenda are the practical expressions of this shift. Whether they produce a measurable change in Canadian AI adoption rates and the retention of AI IP in Canada is a question that will be answered by data over the next five years. The 12% business adoption rate cited in the Policy Magazine analysis is the baseline. A doubling of that rate to 24% would be a significant signal of policy effectiveness. At current trajectories, that is not the most likely outcome in the near term.
Digital trade in services, defined as cross-border delivery of software, data processing, professional services, financial services, and other digitally enabled services, is the fastest-growing dimension of international trade and the dimension where Canada has the most underexploited competitive advantage. CETA, CPTPP, and CUSMA all contain digital trade provisions that create preferential access for Canadian digital services exports to the EU, Indo-Pacific, and North American markets. Canadian software companies, AI firms, cybersecurity providers, fintech operators, and digital media companies have rights under these agreements that they are systematically underusing.
The EU's AI Act, which entered into force in 2024 and creates graduated compliance requirements for AI systems by risk level, is simultaneously a barrier and an opportunity for Canadian AI exporters. Companies that develop compliant AI systems for the EU market build a product that can be sold in any well-governed jurisdiction. Canadian AI companies that are closely aligned with EU regulatory standards have a head start on EU market access over US competitors whose regulatory posture is less harmonized with European requirements. CETA's digital services provisions should be explicitly leveraged to open EU procurement opportunities for compliant Canadian AI and cybersecurity companies.