Speaking alongside French President Emmanuel Macron last month, Prime Minister Mark Carney named four countries at the “forefront” of artificial intelligence: France, Canada, China and the United States.
It was a flattering grouping for Canada, and a generous one. Experts say only two countries truly sit at the top.
“China and the U.S. are in a league of their own,” said Sean Mullin, a former special adviser on economic affairs in the Prime Minister’s Office.
“If you started to look at that next tier there’s a good argument for Canada and France to be in that next tier.”
Even so, experts say there are good reasons for Ottawa to keep building Canada’s AI industry and capabilities.
“We are a tier-one country in certain areas of AI,” said Jaxson Khan, a former AI advisor to the federal government, noting Canada has been “the crucible” of various industry advances.
Canada’s case
Canada can fairly claim to be one of the birthplaces of AI.
At the University of Toronto, Geoffrey Hinton helped launch the machine learning revolution, earning him the 2018 Turing Award and 2024 Nobel Prize in Physics.
He shared the Turing Award with Université de Montréal professor Yoshua Bengio and French researcher Yann LeCun, and the 2024 Nobel prize with American researcher John Hopfield.
Hinton’s former student Ilya Sutskever co-founded OpenAI, which brought AI chatbots into the mainstream.
In Montreal, Bengio founded Mila, now one of the world’s largest academic AI institutes. His brother Samy, also a Université de Montréal graduate, went on to do AI research at Google and now Apple.
In Edmonton, University of Alberta professor Richard Sutton shared the 2024 Turing Award with American researcher Andrew Barto for pioneering reinforcement learning, which teaches AI systems through trial and error.
But American companies have often been the first to turn Canadian research into commercial products.
“[AI is] certainly an area where Canada could and should punch more above the weight than it actually does,” Mullin said.
Mullin notes that it is hard for Canada — or any country — to compete with Silicon Valley.
“Almost every other Western country loses talent to Silicon Valley because it is really the epicentre,” said Mullin.
“And if you want to build world-leading tech companies, sometimes being at the epicentre matters more than anything else.”
In 2025, private investment in AI reached a staggering US$286 billion in the United States, according to Stanford University’s 2026 AI Index Report. By contrast, the other global leader, China, saw investment of just US$12.4 billion, while Canada received US$4.3 billion.
Despite having a fraction of the private sector investment, Chinese models have become competitive with American ones.
“The U.S.-China AI model performance gap has effectively closed,” says the report, which notes that “looking at just private investment figures likely understates China’s total AI spending, given its government guidance funds.”
“U.S. and Chinese models have traded places at the top of performance rankings multiple times since early 2025,” it added, noting Chinese company DeepSeek briefly matched OpenAI’s O1 model in early 2025.
The index lists 12 companies that released so-called “frontier” AI models in 2025: eight were American, two were Chinese, one was Israeli and one was French.
The Cohere bet
Founded in 2019, Toronto-headquartered Cohere is Canada’s leading homegrown developer of AI models, selling its products to businesses and governments around the world.
Khan describes Cohere as “a world-class model company.”
Its language models and AI agent platform, North, let organizations use their own data while controlling where their systems run.
“In general, [Cohere is] really focused on sovereign AI deployments, enterprise AI deployments, sensitive AI … which I think is certainly always useful as a country to have,” said Khan.
Both Innovation, Science and Economic Development Canada and Shared Services Canada have awarded Cohere software licensing contracts. The government has also committed to exploring its use in other departments.
However, Cohere is not reliant on government contracts, Khan notes. About 85 per cent of Cohere’s revenue stems from corporate contracts, the company told Reuters in May 2025.
And most of this business is from international firms, Khan says.
“It’s not like the Canadian government is propping this company up … there’s very strong international market validation for the firm.”
On Oct. 29, Cohere announced the initial close of a Series E funding round targeting up to $7 billion, which would value the firm at about $35.7 billion and rank as the largest private financing by a Canadian company.
As part of that round, the federal government is taking an equity stake of up to $1.6 billion. Prime Minister Mark Carney said Cohere will remain incorporated and headquartered in Canada while committing at least $1.7 billion to Canadian computing infrastructure by 2030.
Cohere did not respond to a request for comment for this story.
Role of a champion
Khan and Mullin, who have both advised Ottawa on its AI strategy, say the federal government has a role to play in advancing an AI champion and Canada’s AI capabilities.
In December 2024, the Trudeau government announced a $2-billion Canadian Sovereign AI Compute Strategy. And this June, the Carney government announced an AI for All strategy that committed $2.3 billion over five years.
Khan argues that backing Cohere preserves Canadian expertise in building advanced models and gives businesses and government departments a domestic supplier for sensitive applications.
“This is just essential capacity for us to have as a country,” he said, noting Canada may not want sensitive health or financial data to go to foreign-owned firms.
Mullin agrees. “I think the idea here is that we need to get behind our best companies to support them, so that we don’t just get locked into U.S. or Chinese tech,” he said.
Khan and Mullin also say it is important for Ottawa to help firms meet their compute needs.
Ottawa has committed up to $240 million to help Cohere expand its domestic compute capacity by partnering with the American company CoreWeave to build a data centre in Cambridge, Ont.
Ottawa has also pledged up to $300 million to help smaller firms pay for computing power.
However, government subsidies will not enable Canada to match the scale of American or Chinese giants.
“There’s no way you’re going to be able to train a globally leading model through a $2-billion government program alone,” said Khan.
Khan also notes that industrial support can only go so far. Another piece of the puzzle, he says, is fostering AI literacy among the general population.
“We can talk about infrastructure, we can talk about industrial policy, all those things are important,” he said.
“But … if you want to have broad-based adoption of this technology, if you want to use it, [you have] to train your people how to adopt it.
“Otherwise, we’re not going to see the kind of broad-based productivity gains and benefits from AI. So I would really focus on that. I’d try to get a K-to-12 curriculum. I’d probably 10x our investments in training, and that’d be a good start.”
