Canada won the AI talent race, but we’re losing the infrastructure one

Sami Kibria is co-founder of flexgrid.cloud and Abdullah Snobar is the executive director of DMZ and chief executive of DMZ Ventures.

Meta’s META-Q recent $13-billion data centre investment in Alberta is a genuine testament to Canada as a global player on the artificial-intelligence stage, with the country leading in research, education institutions and talent. The investments have also pushed AI infrastructure into public discourse, which is long overdue.

While large investments into data centres to create more AI capacity should be celebrated, Canada needs to look at how our entrepreneurs, researchers, businesses and institutions will actually be able to use said infrastructure to build the next generation of AI companies right here at home.

Canada had a fully funded national AI strategy in 2017, before any other country in the world did. These investments built Vector Institute, Amii and MILA into research institutions that are renowned globally. We’re seeing payoffs in the numbers, such as Canada officially outpacing the United States when it comes to the rate of growth of AI talent.

Canada’s plan to counter the U.S. and China in AI is taking shape. Will it work?

With Prime Minister Mark Carney’s pledge to create 250,000 AI jobs by 2031, it’s clear our government is serious about having AI at the centre of Canada’s economic agenda.

Employment targets, strong talent pools and research isn’t enough without access to compute – processing power – and the federal government’s sovereign AI strategy recognized a massive problem back in 2024 that we can’t keep ignoring. Without enough access to AI compute, Canadian innovation will be built here but scaled elsewhere because the infrastructure to do it at home doesn’t exist.

Our default response to growing AI demand has been simply to build bigger data centres, and we all know the adage: Bigger doesn’t necessarily mean better. Canada needs more capacity, but capacity is not the only factor that decides who leads the AI economy – access is.

Take Meta’s investment in Alberta, it’s a major vote of confidence in Canada, but there’s an important distinction to appreciate. The data centre will serve Meta’s AI work, but it won’t expand the compute power for Canadian businesses, researchers and entrepreneurs to develop their own AI projects.

Anthropic is reportedly eyeing Alberta for a data centre too, based on a handful of recent roles they’re hiring for in Canada. Whether it’s Meta or Anthropic setting up shop in Alberta, it’s not a given that any Calgary-based startup close by will get to access its capabilities.

That difference matters because the easier it is for Canadians to get their hands on compute, the more our research turns into companies, jobs (both direct and indirect) and economic growth. The federal government’s new Responsible Data Centre Development Principles announced by AI Minister Evan Solomon recognizes this distinction by calling for data centre projects in the country to provide tangible benefits to Canadians, including access to compute and jobs, but we still have more to do.

Foreign investment undeniably has a place in Canada’s AI strategy and encouraging these players to share more of the upside is undoubtedly progress, but it can’t be our standalone solution.

Every economic era had a defining infrastructure from railroads, electricity and then the cloud. None of them were revolutionary until ordinary businesses and people could access them.

AI asks far more of our computers today than it did even a year ago. We’re feeding it more complex tasks, expecting faster responses and more users are hitting the system at once.

Compute is the thing you need to compete, and access to it is the bottleneck. The countries that make that access easy are the ones that turn AI research into real outputs such as companies and jobs.

Distributed AI infrastructure, meaning a mesh of compute across multiple locations instead of packed into one mega facility, is how you can solve for access.

It brings compute closer to the people and businesses actually using it, which makes the system harder to knock over and allows businesses more visibility over their workloads.

The countries that will lead the AI economy will be the ones to make it easiest for their own entrepreneurs to build and access sovereign compute. Canada has spent well over the last decade building one of the strongest AI ecosystems and now we must give them the infrastructure to lead it.