AI on the Horizon: September 28, 2026

Is Canada ALL IN?

Canada’s latest AI news on the economy, society, and policy. In this issue of AI on the Horizon, we look at Canada’s recent ALL IN AI summer and Cohere’s summer investment strategy. One question lingers: measured against our peers, does any of it move us up the investment table?

Canada’s AI play

Canada showed up, and so did everyone else. 2026’s ALL IN in Montréal drew its largest crowd yet, with Germany as the country of honour, having more than 40 German companies in attendance.

The headline was safety. Canada and Germany are committing up to $300 million to LawZero, the Montréal non-profit founded by Canadian AI pioneer Yoshua Bengio. LawZero’s approach (Scientist AI) focuses on designing trustworthy AI from inception. Bengio’s strategy prioritizes transparency, evidence-based reasoning, and reliability over autonomy. This funding will help develop sovereign compute infrastructure in Canada and establish an office in Berlin, Germany.

On rules and regulations, the minister confirmed a Canadian Digital Safety Commission, a regulator able to issue binding orders and monetary penalties. The proposed penalties run to $10 million or three per cent of global revenue. The new regulator would also oversee private-sector privacy and help implement the new Safe Social Media Act.

The second theme was open-source models. Mila now holds the open-source file in Canada’s national AI strategy, though the federal AI minister declined to name a figure. “It’s in the tens of millions,” he said. “It’s serious backing, and we’re backing Mila.” Mila also partnered with Mozilla on a ready-to-use open-source AI package for business, with $5 million from Mozilla and $1 million from Montréal’s Hypertec.

But American frontier labs arrived with interests of their own. OpenAI’s managing director, George Osborne, was notably complimentary about Canadian data centre prospects. OpenAI has also begun offering free credits to startups at Toronto’s Creative Destruction Lab, which works with roughly 700 firms a year. Anthropic made a similar move in July, with $10 million in credits for researchers at Mila, the Vector Institute and Amii as well as several universities and health organizations across Canada.

Canada’s Cohere competes

Cohere dominated the week. The company signed its business merger with Berlin’s Aleph Alpha, five months after the two announced they would join forces. The combined firm keeps the Cohere name, operates dual headquarters in Toronto and Berlin, and retains a research centre in Heidelberg, and employs more than 1,000 people. Aidan Gomez (Cohere’s CEO), described the result as “a Canadian-German champion at the frontier on the world stage.”

And the capital is following. Cohere is currently in advanced talks to raise a Series E fund between US$2 billion to US$3 billion, and the reporting suggests Canadian and German government money could form part of it. For a company that has raised US$1.6 billion over its entire life, that is a different order of magnitude, and the scale comparison is unavoidable. Anthropic was last valued at US$965 billion – Cohere is seeking US$20 billion.

Finally, scaling compute is another hurdle for Canada’s AI darling. Cohere is buying capacity from a Bell data centre in Merritt, British Columbia in a transaction Buzz HPC values at US$220 million. And they are also partnering with OpenText to put their North platform in front of roughly 120,000 enterprise clients early next year. But an important question remains.

On the Horizon: How do we actually stack up?

Canada’s AI for All strategy is worth about $2.3 billion over five years. The European Union is mobilizing €200 billion through InvestAI, including €20 billion for AI gigafactories. France is opening up 5GW of AI data centre capacity with up to €75 billion of it from SoftBank alone. The United Kingdom is spending £2 billion simply to grow public compute twentyfold by 2030. And South Korea more than tripled its national AI budget, proposing to spend US$7.27 billion in 2026.

Then there is the private side. The four largest American hyperscalers (Alphabet, Microsoft, Meta, and Amazon) have guided to roughly US$700 billion in capital spending this year. That’s nearly US$2 billion a day. Canada’s entire five-year national AI strategy is worth about one of those days.

So, since Canada is not going to win on sheer investment, the more useful question is whether our bet on specialization is the right one. We’re banking on a collaborative approach based on open source, safety, and adoption with countries in a similar position. If this is our only chance, then let’s not be afraid to go all in.

Want to learn more about AI’s productivity potential in Canada? Check out our latest report, The Right AI Tool for the Job.