Canada Opens Applications for AI Sovereign Compute Infrastructure Program in Bid to Build Domestic AI Capacity

Innovation, Science and Economic Development Canada has opened applications for the AI Sovereign Compute Infrastructure Program, a multi-billion-dollar federal initiative designed to build large-scale Canadian AI compute capacity that researchers, public-sector institutions, and Canadian companies can use without sending data to American or other foreign cloud providers. The launch, announced in mid-April, represents one of the most significant pieces yet of the broader Canadian Sovereign AI Compute Strategy, and arrives as the Carney government uses every available lever to position Canada as a credible AI player at a moment of intensifying global competition.

What the program funds

The AI Sovereign Compute Infrastructure Program will provide federal funding to support the construction and operation of large-scale AI training and inference facilities in Canada. Eligible projects include data centres specifically designed for high-density AI workloads, including the deployment of high-performance GPUs and specialised AI accelerators. Funding can also support the supporting infrastructure required for such facilities, including power supply, cooling, networking, and security systems.

The program is structured to support both new-build greenfield facilities and the expansion of existing data centre capacity. Applicants must demonstrate that the resulting compute capacity will be available to Canadian users, including academic researchers, public-sector institutions, and small and medium-sized Canadian companies, on terms designed to support their access. Pure commercial cloud capacity intended for global hyperscaler resale is not the focus of the program.

Funding amounts will be determined on a project-by-project basis, with the largest individual awards expected to support the construction of facilities in the multi-hundred-megawatt range. Applicants are required to provide significant matching investment, either from their own balance sheets or from private partners. The program is part of the broader Canadian Sovereign AI Compute Strategy announced in 2024.

Why sovereign compute matters

The Carney government has framed the program as a matter of economic sovereignty and national security. The argument, articulated by senior officials in recent months, is that Canada's growing dependence on foreign-controlled cloud infrastructure for AI workloads creates risks that go beyond ordinary commercial considerations. Sensitive Canadian data, including health, financial, and government information, is increasingly processed on infrastructure controlled by American hyperscale operators. AI models trained on Canadian data are increasingly developed using compute capacity located outside Canadian jurisdiction.

The argument for domestic AI compute mirrors longer-running arguments about energy security, telecommunications, and critical-minerals supply chains. Officials have explicitly drawn the parallel and have noted that the AI compute question is rising in strategic importance at a moment when the Canada-United States relationship is under acute strain.

The program is also intended to support broader Canadian competitiveness in AI research and commercial deployment. Canadian universities, including the Vector Institute in Toronto, Mila in Montreal, and the Alberta Machine Intelligence Institute, have been internationally recognised for their AI research output but have struggled with compute access in recent years. Canadian companies, including Cohere, have similarly used a mixture of domestic and foreign compute for their model training and inference workloads.

The Cohere question

The most prominent Canadian AI company, Cohere, has been the subject of significant attention in connection with the sovereign compute strategy. Cohere received approximately 240 million Canadian dollars in federal funding under earlier elements of the strategy and has been working on a 725 million Canadian dollar data centre project in Cambridge, Ontario, in partnership with American firm CoreWeave.

Cohere has also recently announced a planned merger with German AI company Aleph Alpha, with major investment from the Schwarz Group, a German retail conglomerate that committed approximately 500 million euros as the lead investor in an upcoming financing round. The merger, framed as the creation of a transatlantic AI champion, has raised questions in Ottawa about the long-term Canadian footprint of the combined entity.

Conservative MP Raquel Dancho raised the question publicly in late April, asking whether Cohere remained committed to a primarily Canadian operational footprint given that significant federal funding had supported infrastructure built and operated in part by an American firm and given that the merger introduced a substantial European stake. Cohere has not committed to specific operational benchmarks, although the company has emphasised its continuing Canadian headquarters and Canadian research presence.

Reaction from the AI community

The Canadian AI research community has broadly welcomed the new program. Leaders at the Vector Institute, Mila, and the Alberta Machine Intelligence Institute have indicated that domestic compute capacity is a critical input to the next phase of Canadian AI research. The Canadian Institute for Advanced Research, which has helped to coordinate Canadian AI research strategy for several decades, has issued a statement supporting the program and calling for sustained federal commitment beyond the initial round of funding.

Canadian AI startups have offered cautiously positive reactions. Leaders of several Toronto, Montreal, and Vancouver-based AI companies have welcomed the prospect of domestic compute capacity but have flagged concerns about the cost competitiveness of Canadian-hosted compute relative to American alternatives. Officials at Innovation, Science and Economic Development Canada have responded that the program is designed to address cost concerns through scale economies and infrastructure subsidies.

The country's largest data centre operators, including Telus, Bell, Rogers, and several independent operators, have indicated they are evaluating opportunities to participate in the program. Multiple consortia have begun forming around potential greenfield projects in Quebec, Ontario, Alberta, and British Columbia, where power costs and grid access make AI workloads particularly economically viable.

Power and environmental considerations

AI compute is among the most power-intensive workloads in modern data centre operations, and the program has prompted discussion in provincial capitals about the implications for electricity systems. Quebec's provincial utility Hydro-Quebec has been particularly active in courting AI workloads, given the province's surplus hydroelectric capacity, although the utility has indicated that new AI data centre proposals will be evaluated against the province's broader electrification needs, including for industry, transportation, and residential heating.

British Columbia's BC Hydro has indicated similar interest and similar caveats. Alberta, with its predominantly natural gas-based electricity grid, has been more cautious, with the provincial government emphasising that AI workloads will be evaluated against emissions and grid reliability considerations.

Ontario, with a mix of nuclear, hydroelectric, and natural gas generation, has positioned itself as a major potential location for AI infrastructure, particularly given the proximity to research talent in Toronto and Waterloo. The provincial government has signalled openness to AI data centre proposals subject to grid integration and rate design considerations.

The international context

The Canadian program arrives at a moment of intense global competition for AI infrastructure. The United States has launched the Stargate initiative, a major partnership between the federal government and private firms intended to support American AI compute capacity. The European Union has launched its AI Factories programme, supporting compute capacity for European AI development. The United Kingdom, Saudi Arabia, the United Arab Emirates, India, and Japan have all announced significant national AI infrastructure programmes in the past year.

Canadian officials have framed the sovereign compute program as a Canadian-scale response to the global trend rather than as a competitive bid against the largest national programmes. The view in Ottawa is that Canada's competitive advantages in AI research, in critical minerals, in clean energy, and in trusted regulatory institutions create a niche that does not require Canada to match the absolute scale of larger national programmes.

What's next

The application window for the program is now open, with proposals due over the coming months. Innovation, Science and Economic Development Canada has indicated that funding decisions will be made on a rolling basis and that the first major announcements should be expected by the end of the calendar year.

The program is one of several federal AI policy initiatives expected over the coming year. The Artificial Intelligence and Data Act, which would establish a comprehensive federal framework for AI regulation, has been in development for several years and is expected to be reintroduced in Parliament in the coming months. The federal government is also working on new procurement frameworks for AI tools used by the public sector and on guidance for the use of AI in federal operations.

For Canadian researchers and entrepreneurs, the message is one of cautious optimism. Domestic compute capacity has been the most consistent gap in Canadian AI infrastructure, and the program represents the federal government's most direct attempt yet to close that gap. Whether the resulting capacity is sufficient to support the next phase of Canadian AI development, and whether the cost structure makes it competitive with American alternatives, are questions that will be tested as the first projects come online.