Ottawa just bought into the AI race. Now it has to pick a lane.

On October 29, in Montréal, Prime Minister Mark Carney announced that Ottawa will invest up to C$1.6 billion in Cohere, the Toronto-founded artificial intelligence company, as part of a Series E financing round targeting up to C$7 billion at a post-money valuation of C$35.7 billion, according to Reuters and The Globe and Mail's reporting on the announcement. The round is led by Germany's Schwarz Group, the retail group behind Lidl and Kaufland, and includes Nvidia, Bell Ventures, Bullhound Capital, Inovia Capital and Prosperity7 Ventures, according to Cohere and the reporting around the deal.

This is not a grant, and that distinction is the whole story. A grant is spent and gone. An equity stake, made through the Crown corporation Canada Development Investment Corporation on what Carney called a commercial basis, makes the taxpayer a part-owner, with a financial interest in whether Cohere wins. Carney's pitch, as carried by CBC News, was blunt about the motive: "We are building it right here, the right way."

The case for the deal is real, and it deserves to be stated at its strongest before anyone picks at it. Cohere is, by the reporting, the company that did not leave. Chief executive Aidan Gomez has said the company once turned down a nine-figure acquisition offer from a major American firm, summing up his view of selling as: "Acquisition is failure." In April 2026, Cohere merged with Germany's Aleph Alpha on terms that left Cohere shareholders with about 90 per cent of the combined company and kept the headquarters in Canada, according to reporting on the merger. Artificial Intelligence Minister Evan Solomon has argued the deal keeps Canada among a small group of countries — alongside the United States, China and France — with companies building advanced large language models. Without Cohere, that list gets shorter, and Canada's claim to a seat at the table where AI is governed gets thinner.

The conditions attached to the money are also more serious than a press-release promise. In exchange for the staged investment, Cohere has committed to creating at least 1,000 new high-paying Canadian technical jobs by 2030, committing at least C$1.7 billion to new computing infrastructure in Canada by 2030, maintaining its Canadian incorporation and headquarters, investing in research partnerships with Canadian academic institutions, and continuing to work with the Canadian AI Safety Institute, according to the Prime Minister's Office as reported by multiple outlets. The money is disbursed in stages tied to milestones, with regular reporting, which means Ottawa keeps leverage over time rather than writing one cheque and hoping. Note the scale of the second commitment: Cohere's own compute pledge is larger than the federal cheque. Ottawa is matching a private build-out, not funding it alone.

Now the sceptical case, which is equally real. The sharpest version, from Charles Lammam in The Hub, asks why a company that has already lined up most of a C$7 billion round from private investors obviously needs Ottawa as a co-investor at all, and warns that government money can breed dependency, with companies chasing Ottawa's cheques instead of customers. That objection cannot be waved away, because it points at the central question the announcement does not answer: what, exactly, does the public money buy that the private money would not?

Start with arithmetic. The announcement does not publish the size of the federal stake, the payment schedule, or any board arrangement, as one detailed analysis of the deal's terms noted. Simple division of C$1.6 billion by a C$35.7 billion valuation puts the stake near four to five per cent — but that is arithmetic, not a disclosed fact, and it should not be quoted as one. A stake in that range is a voice, not control. It does not, by itself, stop a future board from selling, relocating research, or shifting its most valuable work to wherever the compute and customers already are. Ownership of a financial stake, the location of a headquarters, and control over strategic decisions are three distinct things, and the deal's defenders and critics alike should be forced to keep them separate.

Then there is the question of scale. One analysis tallied the capital spending plans of Amazon, Alphabet, Microsoft and Meta at about US$725 billion in 2026, up 77 per cent from the previous year, mostly for AI data centres. Against that, Canada's entire Cohere stake, roughly US$1.1 billion, is about fourteen hours of the giants' spending. That comparison is not an argument for doing nothing — it is an argument for being precise about what this money is for. It cannot out-spend the American labs. It can only buy something they cannot sell Canada: a domestic company, anchored here, building models for regulated industries and governments that want control over their own data.

That is where the lane-picking comes in. Cohere's pitch has always been enterprise AI: systems for businesses and governments that need to keep sensitive information under their own control, rather than consumer chatbots. If Ottawa wants sovereignty rather than symbolism, the stake has to be the beginning of a procurement and infrastructure strategy, not the end of one. Three tests will tell us within a few years whether this was industrial policy or an expensive press release.

First, compute. The C$1.7 billion Canadian infrastructure commitment is the most important number in the deal, bigger than the cheque itself. But a data centre in Canada running on American chips, American cloud software and American networking gear is a Canadian building, not a Canadian stack. The test is whether the build-out creates domestic capacity that Canadian firms and governments can actually use on Canadian terms — and whether the pledge turns out to be genuinely new spending rather than investment Cohere would have made anyway. The public record does not yet settle that, and the staged payments should be used to find out.

Second, customers. A shareholder that never buys is a strange kind of champion. Ottawa, the provinces, hospitals, banks and Crown corporations are exactly the regulated customers Cohere is built for. If Canadian institutions keep buying their AI from American providers while Ottawa holds Cohere shares, the government will own a company its own system does not trust enough to use. Procurement is where sovereignty becomes real or remains a speech.

Third, talent. One thousand jobs by 2030 is a concrete commitment, but the deeper question is whether Canada keeps the researchers it trains. Cohere was founded in Toronto by people who could have built their company anywhere, and the history of Canadian technology is full of firms that stayed Canadian until the moment they didn't. The jobs pledge, the headquarters pledge and the academic partnerships all push in the right direction. They do not repeal gravity.

There is also a concentration risk that should be named plainly. Putting up to C$1.6 billion behind one company ties a large part of Canada's AI credibility to a single firm's fortunes in a capital-intensive industry where profits have so far been elusive, as The Globe and Mail's reporting noted. If Cohere stumbles against far better-funded rivals, taxpayers own part of that loss, and the political lesson drawn will be that government should never back a company again — which would be the worst possible conclusion, since the alternative to backing one champion is depending entirely on other countries' champions. Ottawa's June national AI strategy also includes C$500 million for a Canadian Tech Growth Fund and C$1.75 billion to seed venture-capital funds, according to reporting on the strategy. The Cohere stake should be judged as the flagship of that broader portfolio, not as the whole portfolio. A country that bets everything on one company has not made a strategy; it has made a wager.

So: is Canada now independent, or just a well-dressed customer of the same American cloud giants? Today, neither — and both readings are premature. What Ottawa has bought is time, leverage and a claim: a Canadian company, anchored here by contract, with milestones Ottawa can enforce by withholding the next tranche. That is genuinely more than a grant and genuinely less than control. The honest verdict is that the stake is a good opening move in a game Canada has been losing, and an opening move is only as good as the moves that follow it. Build the compute, buy the product, keep the talent, and publish the terms so taxpayers can judge whether their money bought influence or applause. Do that, and the C$1.6 billion will look cheap. Fail at it, and no share certificate will save the sovereignty speech.