Mark Carney's government is taking a direct equity stake in Cohere, the Toronto firm now valued near C$36 billion. It is the clearest sign yet that national governments, not just Silicon Valley, intend to own the sovereign AI they depend on. The move hands Canadians a share of the upside and ties the company to the country for the next decade.
By Blumefield | October 30, 2026
A government becomes a shareholder
Canada did something on Thursday that Western governments have spent the better part of two years talking about and almost never doing. It bought in. Standing at the Montréal research institute Mila on the opening day of the Canada-European Union Summit, Prime Minister Mark Carney announced that Ottawa will take a direct equity stake of up to roughly C$1.6 billion in Cohere, the Toronto company that has become the country's best-known name in sovereign AI. This is not a grant, not a loan guarantee, and not a research subsidy dressed up in a press release. The federal government is buying shares, in stages, on what it calls a commercial basis, and it expects a return.
The structure matters as much as the headline number. Canada is investing alongside private backers, including Nvidia, the telecom group Bell, and Germany's Schwarz Group, the retail and cloud conglomerate. By taking equity rather than handing out cash, Ottawa positions itself to share in Cohere's upside if the company keeps growing at its current pace. "By taking an equity stake, Canada will share in the company's success," the government said, framing the deal in the language of an investor rather than a patron. The money will be released in tranches, with regular reporting against targets the company has agreed to hit.
For a sector used to American and Chinese capital setting the terms, the signal is loud. A G7 treasury is now a named shareholder in a frontier model developer. The question Carney is betting the answer to is whether a mid-sized economy can own a piece of the sovereign AI it has decided it cannot afford to rent.
The conditions behind the cheque
State money rarely arrives without strings, and this deal is no exception. In exchange for the investment, Cohere has committed to a specific set of obligations that read like an industrial policy wish list. The company will create at least 1,000 new high-paying technical jobs in Canada by 2030. It will commit at least C$1.7 billion to additional computing infrastructure inside the country over the same period, a figure that exceeds Ottawa's own cheque and anchors the physical build-out of models and data on Canadian soil. And it will maintain both its Canadian incorporation and its Canadian headquarters, a direct answer to the long-running fear that the country's best technology companies get built at home and then sold or relocated south.
Two further commitments round out the package. Cohere will expand research partnerships and create openings for students and recent graduates, feeding the sovereign AI talent pipeline that Carney repeatedly returned to. And it will keep working with the Canadian AI Safety Institute, tying the commercial bet to the government's parallel interest in governing the technology responsibly. The disbursement schedule is deliberately conditional. Money moves as milestones are met, with progress on jobs and domestic investment reported along the way.
Evan Solomon, the minister of artificial intelligence and digital innovation, put the political logic plainly. "It is not enough for great companies to start in Canada. We want them to grow here, build here, and create good jobs here," he said. "Investing in Cohere sends a clear message: you should not have to leave Canada to build a company that succeeds around the world."
Why sovereignty, why now
The timing was not accidental. Carney made the announcement hours after releasing Canada's first national security strategy in twenty years, and on the same day he opened a summit designed to deepen ties with a European Union that is itself seeking to reduce its dependence on the United States. Sovereign AI, in this framing, is not a technology story bolted onto a policy agenda. It is the policy agenda.
The rationale the government offered is built around control of data. Cohere sells secure AI systems to businesses and governments, the kind of customers who care intensely about where sensitive information is stored, how it is processed, and who can reach it. Carney leaned into that distinction. "Today, true sovereignty depends on our ability to build, adopt, and govern AI on our own terms," he said. "The bigger and faster Cohere can scale up their models and operations, the more businesses and governments can control where sensitive data is stored, how it is used, and who can access it." He closed with the line that captures the whole gambit: "It is in all our interests for Cohere to succeed, and now all Canadians will have a share in that success."
There is a hard-nosed strategic read underneath the rhetoric. A country that runs its hospitals, courts, banks, and defense systems on models hosted by a handful of American hyperscalers has quietly outsourced a layer of its statecraft. Carney's wager is that spending public money to keep a domestic champion independent is cheaper than the alternative of permanent dependence. For readers tracking how this fits the wider global contest, Blumefield has followed the sovereign AI buildout as it has spread from Paris to New Delhi to the Gulf.
Cohere's bet, and the risks
Cohere is an unusual national champion. Founded in 2019 by Aidan Gomez, a co-author of the landmark research that introduced the transformer architecture now underpinning modern AI, the company made an early decision to skip the consumer chatbot wars and sell to enterprises and governments instead. That focus looked unfashionable when every rival was chasing viral growth. It looks shrewd now that the market cares about security, deployment, and data control more than demos. The company has grown quickly through this year, pushing deeper into secure enterprise deployments and expanding abroad, including a recently completed merger with Germany's Aleph Alpha and the purchase of the Czech cybersecurity startup Aisle.
Gomez framed the investment as validation of a thesis he has argued for years. "We founded Cohere in Canada with the belief that this country could build AI companies capable of competing at the highest level globally," he said. "As AI becomes increasingly critical to economic growth and competitiveness, organisations need solutions that give them greater control over their data, infrastructure, and how AI is deployed." He described the company as helping to "build the future of sovereign AI" from Canada for customers worldwide.
The money lands at a rich valuation. According to reports around the announcement, the government's stake is part of a financing round of roughly C$7 billion that values Cohere near C$35.7 billion, a steep climb from the figures circulating only a month earlier. That number is the source of the risk. Governments do not have a distinguished record of picking technology winners, and a public shareholder buying near the top of an AI valuation cycle is exposed to the same correction that private investors fear. If enterprise AI margins compress as open-weight models from China and Europe commoditize the market, Ottawa's equity could look expensive. The counterargument is that the strings, the jobs, the domestic compute, the headquarters lock-in, deliver value to Canada even if the share price disappoints.
A template for the middle powers
What Carney has really done is import an idea that has drifted through European capitals for two years and give it a concrete form. France has championed Mistral as a continental answer to American labs. Brussels has poured rhetoric and regulation into digital sovereignty. India has stood up its own national model effort. But a direct government equity stake in a frontier developer, structured as a commercial investment with enforceable conditions, is a sharper instrument than subsidy or industrial strategy papers. It is state capitalism applied to sovereign AI, the most strategically sensitive technology of the decade.
The approach carries obvious tensions. A government that is both regulator and shareholder of an AI company invites questions about conflicts of interest, about whether the Canadian AI Safety Institute can scrutinize a firm the treasury is invested in, and about what happens if Cohere needs a bailout rather than a brag. There is also the awkward presence of Nvidia and a German retailer as co-investors in a deal sold on the language of national independence, a reminder that even sovereign AI runs on globally sourced chips and capital.
Yet the logic is spreading because the problem is real. Every government that has digitized its services now depends on models it does not build and cannot fully audit. Canada has decided the answer is ownership, and it has put more than a billion dollars of public money behind the conviction that sovereign AI is infrastructure, not a product to be bought off a foreign shelf. If the bet pays off, expect other middle powers to copy the playbook within the year. If it does not, Ottawa will have learned an expensive lesson about the difference between funding a champion and controlling one. Either way, as Blumefield has argued throughout this cycle, the era of governments treating AI as someone else's business is over.
Primary sources: Office of the Prime Minister of Canada, Cohere, and Mila.