Canada Needs Homegrown AI Infrastructure

The American giant OpenAI has reportedly been in discussions with the federal government about building, or partnering on, AI data centres in Canada. Data centres like these are AI’s physical footprint, housing the high-performance hardware required to train and operate its models. The offer seemed, on the surface, like a win-win. Canada’s abundant clean-energy supply would feed OpenAI’s voracious demand for computing power, and OpenAI would help Canada achieve one of Mark Carney’s stated nation-building goals: expanding domestic AI capacity, which recently got a billion-dollar investment from the federal budget. Building AI data centres in Canada will create skilled jobs and anchor our national capabilities in a sector that’s become fundamental to economic competitiveness. 

But that OpenAI offer also suggested, without providing much detail, that it can help with another of the Liberals’ AI ambitions: to build a “sovereign cloud.” Because Canada’s data infrastructure has long been owned and operated by foreign-controlled firms, the government intends to build AI infrastructure that’s operated and governed under Canadian law. 

Related: Boycotting U.S.Lettuce? Here’s How.

This is where OpenAI’s offer contradicts itself: the more Canada relies on foreign AI investment, the less control we’ll have over our own data. If Canada partners with foreign AI giants to accelerate our infrastructure build-out, we’ll gain their speed, scale and access to the frontier of computing technology. Yet we risk giving up something essential: the power to decide where data lives, how it is used and who captures its value. More and more, the ability to shape an economy depends on who controls the systems that turn data into knowledge. Canada’s economic strength will be determined by whether it builds that capacity at home or yields it to others.

The U.S. already has the advantage here. American laws like the CLOUD Act allow American authorities to access data held by U.S. companies, even when it’s stored abroad. That ensures that data managed by U.S. companies can be access by U.S. authorities, even when it’s stored on servers in Canada. That includes Canadian data. And because American firms retain commercial control over data stored on their platforms, it will fuel foreign research, innovation and profit. The result for Canada is capacity without sovereignty.

The dynamic is familiar. For more than a century, Canada has exported raw materials—lumber, minerals, oil—only to import finished goods at a premium. Data is following the same path. Former Research In Motion co-CEO Jim Balsillie has argued that data is “the raw material of the 21st-century economy.” It includes health records, environmental readings, industrial metrics, online behaviour, public research and more—the core inputs that train and refine AI systems. For example, Canadian health and genomic data is among the richest in the world; it’s routinely accessed by foreign tech firms and research networks to train medical AI systems. The resulting models and patents are commercialized abroad. And Canadian location and mobility data from platforms such as Google Maps and Uber is extracted and used to train and optimize U.S. systems, generating patents, algorithms and commercial products that are owned and monetized abroad.

This is the digital equivalent of sending our crude oil away to be refined elsewhere. Without sovereign control, Canada becomes a supplier of raw material, rather than a builder of the products and intellectual property it fuels. When our datasets are processed abroad, the resulting models, insights and patents are created and owned elsewhere. The economic returns—jobs, research advantage, and strategic leverage—accrue elsewhere as well. The resource may be new, but the pattern of exporting value instead of creating it here at home is old.

This dependence on foreign systems is not hypothetical. Canadian organizations already spend more than $6 billion a year on U.S.-controlled cloud services, effectively paying foreign rents on domestic information flows. Every terabyte of Canadian data sent abroad feeds intellectual-property portfolios worth hundreds of billions globally. Once those models are commercialized, Canadian firms pay again in licensing and API fees to access insights built on their own citizens’ data.

Canada now faces a familiar dilemma. It can permanently cede the idea of true sovereignty to secure a share of investment and infrastructure offered by American tech giants. Or it can take the slower, harder path toward building a foundation of its own. That path means more than hosting data on Canadian soil—it means building the infrastructure to govern and process it under Canadian law. With that control, Canada can begin to build and train domestic AI models on homegrown datasets, turning local information into new research, products and intellectual property. Realizing that vision will require a long-term commitment to align public goals with private capability, and to treat data infrastructure as a national asset, central to Canada’s sovereignty.

The potential payoff is huge. The Canadian Institute for Health Information estimates that better domestic access to health data alone could unlock more than $10 billion a year in value. The same logic applies across the economy: in clean energy, education, finance, AI research and manufacturing. The wealth created from Canadian data should reinforce Canada’s economy rather than fuel others.

It is true that pursuing that path is a gamble. Canada does not yet offer cloud or compute services on the scale of the U.S. companies, and catching up outright is unlikely. But that doesn’t mean Canada should resign itself to dependency. Other nations have shown that strategic coordination and clear standards can shift the balance of power. From a security perspective, initiatives such as the EU-backed Gaia-X project and France’s SecNumCloud require that cloud providers handling sensitive national data be domestically governed. Europe’s goal is to ensure that information critical to the public interest remains under national law, even when delivered through private platforms. Canada’s challenge goes further: to combine that same legal protection with an economic strategy that keeps the value of domestic data, and the innovation it enables, at home.

The path to doing so requires precision, not protectionism. Canada can still welcome international providers, while defining clear conditions for the storage, processing and use of domestic data. That means directing federal innovation funding toward Canadian-owned cloud and compute providers, like Toronto-based ThinkOn, and emerging public-sector data-centre partnerships. It also means embedding sovereignty clauses into government cloud contracts. Data residency and legal control should be treated as pillars of economic strategy, rather than hollow declarations of principle.

The federal government has already signalled its intent. Ottawa has floated mobilizing pension-fund investment and public incentives—roughly $10 to $15 billion—to grow green data centres. But money alone is not enough: if those facilities are operated by firms subject to foreign law, Canada will have invested in capacity it does not truly control.

OpenAI’s offer should be a wake-up call, not an easy win. It highlights the opportunity to design a trusted, Canadian-governed AI foundation that invites collaboration but keeps jurisdiction where it belongs. Canada has always advanced by building the infrastructure that connects and sustains it, from railways to radio waves to power grids. The next system is digital, and its course will be shaped by how, and by whom, it is governed. Canada still has time to ensure the next industrial revolution runs on rails we actually own.

Joshua van Es works in corporate law, and is the founder of Upper Harbour, a Canadian initiative focused on data sovereignty and AI infrastructure.