Canada is not going to outbuild the United States or outspend China in artificial intelligence. The more practical question is where a middle power can still exert control, reduce risk, and build influence.
That question surfaced repeatedly at AI Governance: New tradeoffs for sovereignty, trust and sustainability, hosted by the Lawrence National Centre for Policy and Management at the Ivey Business School. Guided by Professor Erik Bohlin, Ivey Chair in Telecommunications Economics, Policy and Regulation, and Assistant Professor Romel Mostafa, Director of the Lawrence National Centre, the discussion turned to a distinctly Canadian challenge: how should the country think about AI sovereignty when it cannot control every layer of the technology?
For Mark Daley, Chief AI Officer at Western University and NSERC Scholar in Residence in AI, the answer begins with realism.
"Where sovereignty is non-negotiable, we have to build for ourselves," Daley said. "Partnering is where most of our value is going to come from."
Rather than treating sovereignty as total self-sufficiency, Daley argued that Canada should focus on preserving room to act where it matters most.
"Sovereignty is about preserving optionality across as many layers of that stack as possible," he said.
Sovereignty is not just about where data sits
That argument was echoed from an industry perspective by Mark Graham, Senior Vice-President of legal and regulatory affairs at Bell, and Alexandre Guilbault, Vice-President of AI enablement at Telus.
Both warned against treating sovereignty as a narrow question of whether data is stored on Canadian soil. Graham said enterprises and governments need much more than a domestic server location if they want AI systems they can trust.
"You need a full stack of services," he said, including data centres, compute, storage, networks, and the governance systems that control access and encryption. "All of that needs to be controlled from Canada."
Guilbault made a similar point, but framed it around dependence. AI is becoming embedded in software development, operations, and core business processes, he said. That makes compute power a strategic asset, not just a technical input.
"We cannot predict the future," Guilbault said, "but we know for sure that the compute power is the most critical link."
He argued that if AI becomes as central to business as electricity or telecommunications, Canada cannot afford to leave itself fully exposed to outside control.
"We wouldn't want to have our electricity and telecommunication being fully dependent on sovereign actors," Guilbault said. "That's why we need to do the same thing for AI."
The point was not isolationism. It was resilience.
Guilbault pointed to the U.S. CLOUD Act as an example of why geography alone does not settle the issue. "Even if the data is hosted in Canada, if it's hosted by an American company, they can go and query this data," he said. "Our IP is not secure."
Graham pushed the same logic further. As more payroll systems, customer service operations, transport systems, and other essential functions become AI-enabled, the central risk is straightforward: "Who can turn it off?"
Canada does have advantages
Guilbault argued that one of the country’s clearest advantages is energy. Canada has abundant renewable electricity, a cooler climate, and political stability, all of which matter as AI systems demand more power and more data-centre capacity.
At Telus, he said, that has already shaped investment decisions. The company’s sovereign AI factory in Rimouski, Que., is already sold out, and Telus has announced three more AI facilities, drawing largely on renewable power.
"The middle powers can benefit from green energy in a stable democratic country," Guilbault said.
Bell is making a similar bet. Graham said the company expects to have more than 300 megawatts of Canadian AI data-centre capacity operating by the end of next year, funded privately.
Daley argued that Canada’s most durable advantage may be less physical than human.
"Canada is acknowledged as a first mover," he said. "Canada has prestige in the AI space."
That matters because Canada is unlikely to become a leading producer of frontier chips in the near term.
"We will not have optionality on chips in Canada in the next five years," he said.
But talent, trusted institutions, clean power, and a stable democratic environment are all areas where Canada can still build leverage.
A middle-power strategy means choosing where to build
Daley’s broader argument was that Canada should stop thinking like a frustrated superpower and start thinking like a middle power.
That means being precise about where domestic control is essential, where partnerships can extend Canadian influence, and where the country can still shape the rules of the game.
On infrastructure, that could mean building Canadian capacity in areas such as power, trusted procurement, secure data infrastructure, and verification. On the diplomatic side, it means forming coalitions with other countries that share similar constraints and interests.
Guilbault made that point directly.
"All the middle powers need to work together in order to make that possible and compete against China and the United States," he said.
Daley described this as a form of issue-by-issue partnership building. Canada may not have leverage at every layer of the AI system, but it does not need to. What matters is preserving enough control and enough alliances to avoid strategic dependence.
Standards as a strategic lever for middle powers
Daley argued that international standards are one of the few areas where middle powers can still punch above their weight.
The reason is simple. Large AI firms still need access to large markets. If a group of like-minded jurisdictions sets credible standards, companies will have to respond.
"Collectively that group has a pretty strong lever," Daley said, referring to the potential influence of countries such as Canada, Japan, Korea, and members of the European Union.
His case was not that standards are flashy. It was that they are practical. Canada already has credibility in AI research and policy. The next step, he said, is to turn that credibility into influence by getting more respected Canadian experts onto international standards committees.
That would not solve every problem. It would not give Canada control over frontier models or chip manufacturing. But it could give the country a meaningful role in shaping the governance frameworks that companies must follow.
For a middle power, that may be the point. The goal is not total independence. It is to secure enough capability, enough influence, and enough flexibility to protect national interests in a technology landscape dominated by larger players.
That makes sovereignty less an all-or-nothing ambition than a strategic exercise in choosing where Canada must build, where it should partner, and where it can still lead.
Related Articles
