On June 4, 2026, Prime Minister Mark Carney launched AI for All, Canada's first national AI strategy, pledging over $2.3 billion — north of $3.5 billion counting near-term commitments — to build a public AI supercomputer and "sovereign" compute and cloud infrastructure. The framing is explicitly geopolitical. "We will strengthen Canadian sovereignty at a time when it is being deeply challenged," the strategy reads, warning that Canada risks "leaving critical parts of our AI ecosystem under foreign control." Ottawa is targeting $200 billion in additional growth and 250,000 AI-related jobs by 2031, and lifting AI adoption from just over 12% to 60% by 2034.
The announcement reads as Canada's answer to the same anxiety driving Brussels — the European Commission unveiled its own Technological Sovereignty Package the day before. But Canada already runs a quieter, cheaper, and arguably more consequential experiment in loosening Big Tech's grip: the Competition Bureau's litigation against Google and Amazon. The two efforts share a premise. They point in opposite directions on cost and effectiveness.
The sovereignty case, fairly stated
The strongest argument for AI for All is real. Compute is now strategic infrastructure, and a small open economy that rents all of it from a handful of US hyperscalers is exposed — to price-setting, to export controls, and to the kind of cross-border political pressure Canada has felt acutely under a second Trump administration. A public supercomputer gives Canadian researchers and startups access to capacity the market may underprovide for non-commercial work, and a domestic cloud option is a hedge against a genuine single-point-of-failure risk. Sovereignty, on this view, is not protectionism; it is resilience.
That case deserves to be taken seriously rather than dismissed. But it should not crowd out the recognition that who controls the market matters more than who owns the servers — and on market control, Canada's competition enforcers are already doing the heavier lifting.
The enforcement track record
In November 2024, the Competition Bureau sued Google before the Competition Tribunal over its dominance of the online advertising stack, alleging the company unlawfully tied its ad-tech tools together to entrench its position. The Bureau estimates Google holds roughly 90% of publisher ad servers, 70% of advertiser networks, 60% of demand-side platforms, and 50% of ad exchanges in Canada. Its requested remedy is structural: divest two core tools, DoubleClick for Publishers and AdX, plus a penalty.
Google tried to kill the case on constitutional grounds, arguing that the administrative monetary penalty regime under section 79(3.1) of the Competition Act — which Google estimated could reach a startling $91 billion — amounted to a criminal sanction triggering Charter protections. In March 2026, the Tribunal dismissed that challenge (Commissioner of Competition v. Google Canada Corporation, 2026 Comp Trib 10), finding AMPs are regulatory deterrents, not penal punishment, and dismissing the $91 billion figure as "hypothetical." The case can now proceed to its merits.
The Bureau is running a parallel probe into Amazon, securing a court order in July 2025 to advance its investigation of the Marketplace Fair Pricing Policy — the rule that lets Amazon penalize third-party sellers who list lower prices elsewhere, a classic platform-most-favoured-nation concern.
Why antitrust is the better instrument
Here is the asymmetry that should shape Canadian policy. Subsidizing sovereign compute costs billions of public dollars and, even if it works, produces a Canadian alternative that must still win on price and performance against incumbents that enjoy enormous scale economies. Antitrust enforcement costs a fraction of that and attacks the actual mechanism of dependence: conduct that forecloses rivals and locks customers in. Breaking open Google's ad-tech tie or curbing Amazon's pricing leverage lowers costs for every Canadian publisher and seller immediately — no new data centre required.
The two tracks can also work at cross-purposes. "Sovereignty" is an elastic word, and the history of industrial policy is littered with national champions that delivered captive markets and higher prices rather than competitive frontier technology. If AI for All's procurement preferences and subsidies end up shielding a favoured domestic provider from competition, Canada will have recreated the very lock-in its competition law exists to dismantle — only this time with a maple-leaf logo. Genuine sovereignty is the ability to choose, which is what contestable markets produce; it is not a guarantee that the chosen vendor is Canadian.
A coherent strategy would sequence the two
None of this argues against public compute. Targeted investment in research infrastructure the market won't fund is defensible, and a resilience hedge against single-vendor risk is prudent. But the strategy's own logic — reducing dependence on dominant foreign providers — is better served by enforcing competition than by replacing one set of gatekeepers with another. Canada should let the Bureau's structural remedies run, resist building subsidies into permanent procurement carve-outs, and measure AI for All by whether it expands the menu of credible options rather than by how much Canadian-owned capacity it can point to.
The Google and Amazon cases show Ottawa already possesses a precise, low-cost tool for prying open concentrated tech markets. Spending $3.5 billion to build around the problem, while a $91 billion remedy sits one tribunal ruling away from going to the merits, would be a strange allocation of national effort. Build the supercomputer if the research case holds — but win the sovereignty argument in the courtroom first.