{"id":28215,"date":"2026-06-11T12:00:00","date_gmt":"2026-06-11T12:00:00","guid":{"rendered":"https:\/\/www.rbc.com\/en\/?post_type=rbc_tl&p=28215"},"modified":"2026-06-15T14:28:29","modified_gmt":"2026-06-15T14:28:29","slug":"sovereign-ai-shaping-canadas-next-digital-chapter","status":"publish","type":"rbc_tl","link":"https:\/\/www.rbc.com\/en\/thought-leadership\/the-growth-project\/sovereign-ai-shaping-canadas-next-digital-chapter\/","title":{"rendered":"Sovereign AI: Shaping Canada’s Next Digital Chapter"},"content":{"rendered":"\n
\n
\n

American excellence in artificial intelligence has had an unintended side-effect: a hyper concentration in computing software and hardware. Three U.S. tech firms account for around 85% of Canada\u2019s cloud infrastructure spending, while another three account for roughly 88% of enterprise foundation model usage. Meanwhile, NVIDIA makes up about 80-90% of the advanced AI chip market.<\/p>\n\n\n\n

Canadian firms and governments do not play a meaningful role in the global AI supply chain. But there is both a commercial and national demand to exert more control over our digital infrastructure to ensure deep AI capability as a country. The issue of \u201csovereign AI\u201d has emerged as a critical issue as the world is swept up by the new technology. At the same time, choosing to build sovereign infrastructure and sovereign AI systems, or creating sovereign requirements for existing cloud infrastructure could involve a cost premium and could reduce technological competitiveness. Canadian businesses pursuing sovereign AI initiatives should consider which workloads to keep on existing cloud infrastructure, and which may require new architecture.<\/p>\n\n\n\n

Canada is not alone in its pursuit of sovereign AI. More than 70% of global executives, investors and government consider sovereign AI as an \u201cexistential concern\u201d or strategic imperative\u201d to their goals, according to McKinsey & Co., which projects global sovereign AI to become a US$600- billion market by 2030.<\/p>\n\n\n\n

For several years, the Canadian government has pursued new legislation to strengthen privacy protections and modernize its digital regulation. At the same time, a more assertive posture from Washington heading into the forthcoming Canada-U.S.-Mexico Agreement (CUSMA) review has cast some of Canada’s digital regulatory efforts as potential trade irritants. The result is a live debate\u2014on both sides of the border\u2014over how Canadian businesses can continue to access the latest AI innovations while maintaining robust safeguards and protections.<\/p>\n\n\n\n

Canadian firms have found themselves making infrastructure, data, and vendor decisions in an environment that has materially changed in the last 18 months. The AI stack\u2014cloud, compute, foundation models, and the data those systems run on\u2014has simultaneously become a top subject of trade negotiations, regulatory design, procurement strategy, and operational risk management. The choices being made now at the negotiating table and in Canadian boardrooms will set the conditions for the next decade of the digital economy.<\/p>\n\n\n\n

For a start, the U.S. is now more aggressively pushing for its interests and tech sector dominance, with an immediate challenge being CUSMA\u2019s digital trade chapter, with a July 1, 2026, being the milestone for the six-year joint review.<\/p>\n\n\n\n

Unlike several other trade agreements, on data localization specifically, CUSMA goes further than comparable agreements: unlike the Comprehensive and Progressive Agreement for Trans-Pacific Partnership, its rule against requiring domestic computing facilities (Article 19.12) contains no \u201clegitimate public policy objective\u201d exception, narrowing the space Canada would otherwise have to mandate local data storage for public services or citizen data. <\/p>\n\n\n\n

Four provisions matter most for AI and data sovereignty: 1) limits on less-favourable treatment of foreign digital products, 2) restrictions on blocking cross-border data flows for business conduct, 3) prohibitions on requiring domestic computing facilities as a condition of operating, and 4) limits on source-code disclosure requirements. There is still room to move, but in narrower channels: the federal government\u2019s national security exception, the federal procurement exclusion, and prudential carve-outs in financial services to preserve system stability. The federal government is also working on over a dozen targeted exceptions for defence.<\/p>\n\n\n\n

Through a series of executive orders and policy shifts, the U.S. has hardened its position on digital trade. Washington is increasingly treating digital and AI regulation, including rules adopted outside its borders, as a trade irritant and barrier to market access. In March 2026 the United States Trade Representative (USTR) named Canada\u2019s sovereign computing initiative, digital rules, and proposed regulations as trade barriers in its National Trade Estimate report. U.S. Trade Representative Jamieson Greer has noted that all options are on the table regarding the future of CUSMA. A U.S. executive order in 2025 creating a federal AI Litigation Task Force, while a White House recommendation calling for a \u201cminimally burdensome national standard,\u201d point in the same direction: AI regulation is being framed as commercial friction. In a CUSMA submission, the American industry-led Computer & Communications Industry Association (CCIA) labelled Canada\u2019s Online News Act and Online Streaming Act as discriminatory, suggesting major headwinds ahead for the Canadian government and businesses to protect their interests.<\/p>\n\n\n\n

Ottawa moved on June 4, 2026, releasing AI for All, its national AI strategy, organized around trust, opportunity, and sovereignty. Two of its six pillars target exactly these dependencies\u2014a “sovereign AI foundation” in compute, data, and talent, and scaling Canadian champions\u2014under a “build-partner-buy” approach. The headline commitments are concrete: infrastructure “operated under Canadian control and Canadian law,” sovereign-compute partnerships of 850 megawatts by 2030 scaling toward 2.3 gigawatts, a public supercomputer by 2031, and $700 million in affordable compute for smaller firms. Depending on execution, this may help address some of the gaps.<\/p>\n\n\n\n

For Canadian leaders, sovereignty in the AI era should not mean isolation. It means freedom from coercion: the ability to choose which AI models to use, whose hardware runs AI inference, which jurisdiction governs data, and which providers can be substituted under pressure. Notably, Canada\u2019s domestic stack for infrastructure is developing faster than the prevailing narrative suggests, including Cohere\u2019s CoreWeave build out, Bell AI Fabric, TELUS AI data centres, the ThinkOn\u2013 Hypertec\u2013Aptum\u2013eStruxture consortium, and a new Canadian supercomputer are all building domestic AI and data infrastructure. Furthermore, Canada has many successful, global enterprise technology companies across AI models, financial services, healthcare technology, knowledge management and data storage, and beyond.<\/p>\n\n\n\n

\"Canadian<\/figure>\n\n\n\n

In Sovereign by Design <\/em>(Mullin & Khan, 2026) the above heat map rates “the severity of risk to Canadian AI sovereignty” at each layer\/dimension, where sovereignty means freedom from coercion, and the ability to structure dependencies so they can’t be used as leverage. The four tiers map to how concentrated the dependency is and whether Canada has a viable substitute:<\/p>\n\n\n\n

Reading the heat map.<\/strong> Each cell rates the severity of sovereignty risk where a layer of the AI stack (rows) meets a dimension of digital sovereignty (columns). Severity reflects the degree of foreign dependency, supplier concentration, and substitutability\u2014and how much leverage a disruption would hand to an outside actor.<\/p>\n\n\n\n