Canada’s infrastructure creates the opportunity, but governance is what sustains it, write Savar Suri and Perrin Beatty.Adrian Wyld/The Canadian Press
Savar Suri is an Émile Boutmy Scholar at Sciences Po Paris.
Perrin Beatty is a former cabinet minister and former CEO of the Canadian Chamber of Commerce.
The dominant narrative in the race for AI leadership has become familiar: For a seat at the table, you must produce the next OpenAI or Anthropic, train the researchers, and build a frontier model. By this measure, Canada, having no company on this scale, is already behind. The narrative, however, misunderstands the mechanisms of leverage in AI leadership.
The limiting factor in modern AI development is not talent or algorithms – it’s the physical infrastructure required to run the systems. Training and deploying advanced models demand conditions that make 30-year capital commitments viable. Chip design and manufacturing are immensely important too, but the leaders of that race have been all but decided. The more relevant question today is more specific: Who can provide the land, power and political stability required to run the systems being built? This is a different race, and one Canada is far better positioned to enter.
Developments across the United States make this a legitimate opportunity. In the first quarter of 2026 alone, more than US$40-billion worth of data centre projects were reportedly cancelled, with community opposition cited as a key issue. Electricity bills, water consumption, and noise have turned data centre construction into a flashpoint in local politics. At least 11 states have proposed legislation that would create moratoriums on new development, and federal legislation to pause construction has been introduced, although not yet passed. The computing infrastructure required by AI is running into resistance in the country most determined to build it.
Canada’s AI strategy to fund national health data project to improve care, investment, minister says
Canada offers a structurally different environment. Approximately 80 per cent of Canadian electricity comes from non-emitting sources, primarily hydroelectric. This is a differentiator that grows more significant as AI firms face mounting pressure over their carbon footprints. The cool Canadian climate substantially reduces the cooling costs, which can account for more than a third of a data centre’s operating expenses. As the second largest country in the world, Canada has an abundance of land. Crucially, although existing energy capacity is beginning to approach its limits, Canada’s untapped nuclear potential, alongside significant further hydroelectric renewable development opportunities, means the country has the potential to significantly increase its generation capacity over the coming decades. Perhaps most importantly for investors, Canada’s regulatory environment, despite delays, is coherent across jurisdictions and transparent in its processes. This combination matters to those committing capital over long time horizons.
Taken individually, these advantages are real but not decisive. What makes them so compelling is that no plausible competitor simultaneously holds all the same ones. The Gulf states have capital and cheap energy but operate in a geopolitically volatile region. Taiwan and South Korea host some of the world’s most critical technology facilities, but their geographic exposure is a material consideration for decade-long infrastructure bets. The Nordics have cheap energy and political stability, but limited land and higher costs at scale. Canada’s convergence of clean power, cold climate, abundant land, institutional stability and geographic separation from conflict is a combination no other candidate country holds.
Jurisdiction plays a role here too. Governments across Southeast Asia, Latin America and parts of Africa are actively seeking computing infrastructure that does not sit under American or Chinese legal control. Canada, as a stable middle power with no hegemonic ambitions and an established record of meeting international obligations, is the most credible alternative. But this opportunity is contingent on Canada developing a governance framework that delivers genuine data sovereignty to clients, not American infrastructure under a different flag. Critically, this framework must also reckon with Indigenous land rights at home. Many of the most viable sites for large-scale data infrastructure sit on, or are adjacent to, Indigenous territories. Failing to build meaningful partnership into the development is not only a legal and ethical risk, but it would directly undermine the regulatory coherence Canada should position as a selling point.
Canada is not without its own players in this space. The Toronto-based Cohere has quietly emerged as one of the best alternatives for institutions seeking sovereign AI deployments. This ambition is underscored by their plan to acquire Germany’s Aleph Alpha, a firm similarly focused on secure, aligned systems.
Google’s chief economist backs Ottawa’s ‘AI for all’ strategy
Canada’s infrastructure creates the opportunity, but governance is what sustains it. The next phase of AI competition will be decided not just by those who build the systems, but also by those who define how they are evaluated, trusted, and deployed across economies. Anthropic’s Mythos model has already demonstrated the importance of having policies in place that anticipate potential risks. Canada’s early investment in AI safety and ethics research is anchored by global leaders. The Canadian Institute for Advanced Research and MILA – Quebec AI Institute (headed by Turing Award recipient Yoshua Bengio), are both deeply embedded in global AI research and policy networks, which helps position Canada as a convincing standard-setter in a field that lacks one. Countries that establish these governance frameworks will determine which systems gain market access across the Western ecosystem, embedding their preferences into the global AI stack. This leverage compounds Canada’s infrastructure advantage.
The real obstacle, however, is domestic credibility. Canada has a history of announcing major infrastructure strategies and delivering delays, overruns, and cancellations. Ottawa’s recent call for proposals for sovereign, large-scale AI data centres, as well as the announcement of a National Electricity Strategy are valuable first signals, but the stages between identifying projects and executing their construction is the area where Canada has historically lost time.
The window is closing. In 2025, Microsoft, CoreWeave, and Brookfield committed more than US$15-billion to Nordic AI infrastructure. The Gulf states are building at an extraordinary scale. The structural advantages Canada holds are real, but they will not execute projects themselves.
Canada is not a conventional AI power, and nor is it likely to become one in the usual sense. Nonetheless, the competition for AI infrastructure is being decided in the field that Canada happens to occupy, and unlike the race for model supremacy, this is a race Canada can win. Whether these advantages are recognized and realized before the window closes is a policy question. That Canada has barely begun to act is a more immediate problem.