Essay · October 2025
“This would build compute capacity and data centres that we need to underpin Canada’s competitiveness, to protect our security, and to boost our independence and sovereignty. This will give Canada independent control over advanced computing power while reinforcing our leadership in AI and quantum.”
— Prime Minister Mark Carney, on the Major Projects Office and Canada’s sovereign cloud initiative (Edmonton, September 2025; as reported by The Quantum Insider)
There is a contradiction at the heart of Canada’s digital economy that rarely gets named directly. Canada has built some of the world’s leading AI research institutions — Vector, Mila, the Alberta Machine Intelligence Institute — and invested heavily in the talent pipeline feeding them. And yet the infrastructure those researchers depend on — the cloud platforms, the compute capacity, the data centres — is overwhelmingly owned and controlled by American companies operating under American law.
A Digital Stack Canada Does Not Control
This is not a minor administrative detail. It is a structural vulnerability. In June 2025, Anton Carniaux, Microsoft France’s director of public and legal affairs, testified under oath to a French Senate inquiry that he could not guarantee customer data would never be handed to U.S. authorities without the French government’s approval. That admission made explicit what many had preferred to leave implicit: data held by a U.S. provider remains subject to U.S. law, including the CLOUD Act, wherever it is stored. Canada runs on the same providers, and Canada does not control its own digital stack. Ninety-three percent of the office-software market is held by two U.S. firms, Microsoft and Google, and sixty percent of the cloud market by five U.S. companies (Beaumier, Policy Options, 2025). American companies own 84 of Canada’s 286 data centres (Investigative Journalism Foundation, 2025). Even traffic that never needs to leave the country often does: researchers at the University of Toronto’s IXmaps project have documented how Canadian internet traffic routinely “boomerangs” through the United States before returning home.
Canada has world-class AI research and foreign-controlled AI infrastructure. That gap is the sovereignty problem.
What Digital Sovereignty Actually Means
Digital sovereignty is often framed as a nationalist or protectionist concept. That framing misses the point. At its core, digital sovereignty is about a nation’s capacity to govern its own digital ecosystem — to have meaningful control over where data is stored and computed, who designs and operates critical technology, and which jurisdictions have legal authority over the systems that run public services, financial infrastructure, and private communications.
It operates across three dimensions: data sovereignty (where is data stored and who can access it?), technology sovereignty (who owns and operates the infrastructure?), and operational sovereignty (who has legal authority to enforce the rules?). Canada currently has significant gaps in all three — not because of bad policy, but because an open, innovation-friendly digital economy was built on the assumption that foreign infrastructure was neutral. It isn’t.
Three Models, Three Lessons
The global landscape offers three distinct approaches to digital governance, each with instructive failures.
The U.S. model — entrepreneurial capitalism — produced extraordinary innovation velocity and the dominant global tech platforms. It also produced winner-take-all market concentration. The companies that built the open internet now control the infrastructure that everyone else depends on, including Canada.
The Chinese model — state-guided capitalism — achieved rapid catch-up growth and strategic coordination. It also produced resource misallocation, overcapacity, and a digital ecosystem that is functionally closed to the outside world. It is not a model Canada should replicate.
The European model — big-firm capitalism with strong regulatory frameworks — has produced the world’s most ambitious digital governance legislation (GDPR, the AI Act, the Digital Markets Act) and very little sovereign digital infrastructure. Europe regulated the American platforms. It did not build alternatives. The Draghi report, published in 2024, identified this as a core competitiveness failure.
Canada sits in a fourth position: hybrid dependency. A trust-based ecosystem, strong research institutions, and a government that funds strategic industries — but critical infrastructure (cloud, AI compute, semiconductors) remains foreign-controlled. The research is Canadian. The stack is not.
The Case for Balanced Sovereignty
The answer is not to close Canada’s digital economy. Isolation would sacrifice the innovation benefits of openness and the economic relationships that depend on interoperability with global systems. The answer is balanced sovereignty: targeted strategic autonomy in critical domains, combined with continued openness in others.
Three policy directions follow from this framing.
A National AI Sovereign Compute Program. Canada lags on national compute capacity — the foundational infrastructure that determines whether domestic AI development is even possible at scale. The OECD has flagged this as a core competitiveness gap. Carney’s sovereign cloud initiative is a start, but the scale of investment required is larger than what has been announced.
Closing the SME diffusion gap through procurement. Large organizations can negotiate cloud contracts with sovereignty provisions. Small and medium enterprises cannot. Procurement-led adoption — using government purchasing power to pull SMEs toward sovereign-compliant infrastructure — is the mechanism that bridges this gap. Tax support for domestic cloud adoption would accelerate it.
A Digital Sovereignty Legal Pack. Bill C-27, which would have created the Artificial Intelligence and Data Act, died on the Order Paper when Parliament was prorogued in January 2025. Canada still has no federal AI-specific law. Government cloud guidelines exist but are not binding. A coherent legal framework — connecting privacy law, cloud procurement rules, and research infrastructure commitments — would close the gap between what Canada says it values and what its contracts actually require.
The Risk Worth Naming
Industrial policy always carries the risk that politically chosen winners become economically inefficient losers. The case for sovereign compute is not a case for protectionism — it is a case for building the foundation that makes meaningful competition possible. Sovereignty is a means to strengthen competitiveness. It is not an end in itself.
The question Canada has to answer is not whether to have a digital sovereignty strategy. Microsoft’s admission that it cannot guarantee customer data against U.S. legal demands already answered that question. The question is whether Canada will build the infrastructure to give its sovereignty strategy actual teeth, or continue to regulate foreign platforms while depending on them for everything that matters.
Key takeaways
- Canada’s AI research is world-class, but the cloud, compute and data-centre layers it runs on are largely U.S.-controlled and subject to U.S. law.
- Neither the U.S., Chinese nor European model fits; Canada needs balanced sovereignty: strategic autonomy in critical domains, openness elsewhere.
- Three levers: a national sovereign compute program, procurement-led SME adoption, and a binding legal framework now that AIDA has lapsed.
References
- Beaumier, G. (2025, May 6). How Canada could achieve digital sovereignty. Policy Options.
- DLA Piper. (2025, January). Canadian privacy and AI horizon shifts again.
- Draghi, M. (2024). The future of European competitiveness. European Commission.
- Kasonga, M. (2025, September 17). American firms own nearly a third of Canada’s data centres. Investigative Journalism Foundation.
- The Quantum Insider. (2025, September 12). Carney pushes for sovereign cloud as Canada navigates quantum future.
- The Register. (2025, July 25). Microsoft admits it ‘cannot guarantee’ data sovereignty.
The full research deck
Leave a Reply