Cohere co-founders Nick Frosst, Aidan Gomez and Ivan Zhang. Photo: Cohere
The Government of Canada is taking an equity stake of up to CAD $1.6 billion in Cohere. It is the largest direct investment Ottawa has ever made in a technology company, it values the Toronto AI firm at CAD $35.7 billion, and the terms of the stake have not yet been made public.
This is an opinion piece, so for transparency: I’ve built Fello for ten years without public money, so I come at this as an operator, not a policy person. I’m undecided on whether this is a good move. I love taking risks, and I can also see why people criticize the deal. We’ll know in a few years. What I’ll say up front is that I’m always cheering for a founder who takes a swing. All figures are in Canadian dollars unless marked USD.
I’m not here to argue the deal should be undone. It’s signed. I’m here to lay out what was announced, how the people who build in this country reacted, and what has to happen next for it to pay off.
What was announced
Prime Minister Mark Carney announced the stake on October 29 at Mila in Montreal, the opening day of the Canada-EU summit.
The money is part of Cohere’s Series E, a round targeting up to CAD $7 billion. Germany’s Schwarz Group is leading it, with Bell Ventures, Bullhound Capital, Nvidia, Inovia and Prosperity7 alongside. It values Cohere at CAD $35.7 billion, up from roughly USD $7 billion a little over a year ago.
In exchange, Cohere has committed to stay incorporated and headquartered in Canada, invest at least $1.7 billion in Canadian computing infrastructure by 2030, and create at least 1,000 technical jobs. The money is paid in stages tied to those commitments. AI Minister Evan Solomon told BetaKit the tranches are tied to milestones where jobs and revenue targets are key.
Cohere has gone further on its own. In its release the next day, the company committed to more than CAD $4 billion in R&D and AI infrastructure in Canada through 2031, more than 1,500 jobs, 250 paid student roles, and research partnerships with the University of Toronto and the University of Waterloo.
This isn’t Ottawa’s first cheque to Cohere. In December 2024 the government committed up to $240 million toward a $725 million compute project, money that went to capacity at a CoreWeave data centre in Cambridge, Ontario. CoreWeave is American. In March of this year, Cohere also won two federal contracts worth a combined $2.3 million for its North platform. Shared Services Canada said it chose North because it was the only fully Canadian-developed option that met its sovereignty requirements. One of those contracts was sole-sourced.
What hasn’t been disclosed: the size of the stake, the share class, the price per share, whether Ottawa gets a board seat, and what happens if Cohere misses its milestones. Solomon told BetaKit the money would be staged on milestones but declined to share the commercial terms. The stake is held through the Canada Development Investment Corporation, the Crown corporation that manages Ottawa’s equity holdings. Cohere says annual recurring revenue grew 165% year over year and it has 300 customers. It has not disclosed the revenue figure.
How the tech world reacted
The deal split Canadian tech within hours. Not along party lines. Along the question of whether this is how you build an ecosystem.
Greg Isenberg, who grew up in Montreal, put the ecosystem case most clearly. His worry wasn’t Cohere. It was every other founder: when the government owns part of a company, every Canadian AI startup is now competing with a company its own government owns. He also asked why Cohere shared 165% ARR growth but not the ARR itself, and whether “sovereign” means models built in Canada or just hosted here.
Bram Sugarman looked at it as a deal structure. If he’d negotiated it for the government, he wrote, there would be no cash investment: warrants or options, a commercial deal for government to use Cohere’s products, a distribution deal through government suppliers, and abundant compute supply. His read is that Ottawa is really stepping up as a landmark customer. The published terms don’t confirm that yet.
Humphrey Su got the most mileage out of a single line: he was now an investor in Cohere “as a LP of Canada.” The joke lands because a real limited partner gets the fund terms, quarterly reporting and information rights. So far, Canadians have the press release.
Nick Dooley runs Augure, a Montreal-hosted sovereign AI platform for lawyers, therapists and defence contractors, which makes Cohere his biggest competitor. His video is more generous than his caption. His read: the market is just getting started, and this is a way to fund a global player from Canada. He gives the deal credit for keeping Cohere and its jobs here, for the progress reports built into the staging, and for the lift it gives everyone building in the country. He’s candid that it’s not an easy thing to watch as a competitor. He’s also still building. His line: if he were scared of competition, he’d be in the wrong business.
The best defence came from Nolan Druid, a Carleton computer science student. His point was that tech people calling critics idiots are missing the room. A construction worker whose day starts at 6am doesn’t check benchmarks. He checks gas prices and ER wait times. Nolan still backs the deal: Cohere could have left, the money is staged, and the alternative is becoming fully dependent on companies outside Canada. He also admits that if Cohere loses, taxpayers own the loss.
The sharpest criticism went after the product. Alex Grenier called the deal a scandal, arguing taxpayers are backing models that can’t bench anywhere near the frontier, and said it was disgraceful. Demetree asked which frontier-level Cohere model he would actually use, given Anthropic, OpenAI, Google, Meta and the Chinese labs.
That’s the spread. The louder argument is about what the state is doing on the cap table.
The open questions
Most of the debate online is about benchmarks. The questions that will actually decide whether this works are quieter, and they’re the ones worth asking Ottawa and Cohere directly.
The terms. Did Ottawa get the same shares at the same price as Schwarz? Preferred or common? A board seat? Clawbacks if the jobs don’t show up? Ottawa’s own $1 billion Venture and Growth Capital Catalyst Initiative requires $3 of private money for every $1 of public money. Nobody has yet explained why this deal was structured differently.
The compute. The $1.7 billion Cohere promised to spend on Canadian infrastructure is the headline that makes the deal look balanced. But the last sovereign compute money went to an American-owned data centre on Canadian soil. There’s a better sign this time: Cohere has a CAD $309 million GPU cloud deal with Bell AI Fabric, and Bell is Canadian. Whether the rest of the $1.7 billion follows that pattern, and whose chips end up in the racks, is still the question.
Procurement. Shared Services has already sole-sourced Cohere once. Solomon has said Ottawa will use Canadian-made AI in its own operations and that startups will be “validated” with government contracts. Government as anchor customer is how the US defence industry got built. It’s also how protected industries get built. Which one this becomes depends on whether Cohere keeps having to win those contracts on merit.
Who advises whom. Jordan Jacobs, co-founder of Radical Ventures, an early Cohere investor, sits on the Prime Minister’s new AI council. Cohere’s Chief AI Officer Joelle Pineau sat on the task force that produced the strategy pledging to anchor Cohere in Canada. None of that is improper. Advisory councils are supposed to include people who know the industry. It’s just worth saying out loud that the strategy, the investment and the company share a small circle of people.
Who’s next. Alongside the Cohere stake, Ottawa is standing up a $500 million Canadian Tech Growth Fund built to take equity positions in AI companies. Cohere is the pilot for a policy, not a one-off.
What this comes down to
Both sides have a real point, and it’s worth being honest about that instead of picking a tribe.
The critics are right that the private money was there. A German retail giant is leading the round and Nvidia is back in, so Ottawa’s cheque wasn’t the difference between Cohere raising and not raising. They’re right that a government on the cap table changes the game for every other Canadian founder. And they’re right that the terms haven’t been made public. Those are fair things to want answered.
The defenders are right too. Cohere could have left, and plenty of Canadian companies have. The money is staged against jobs and revenue, which is better than a grant. And after this summer, when US export controls took Anthropic’s top models offline for nearly three weeks, the idea that Canada should have at least one serious AI company it can count on stopped being abstract. A country that wants to build has to accept risk. Investing is a bet: you put money in, the company gets better, you earn the return. That’s the whole idea. Only time will tell if this one pays.
That’s not how I’ve built, for what it’s worth. I understand you need to raise money to make money, but my outlook has always been control of the company and its revenue. Fello has gotten leaner every year and made more, consistently. I respect the game, and if you have a good model, I get that it’s all about scaling and moving fast. It just means I look at a $7 billion round with a different set of eyes.
What’s not in dispute is what the signature changes for Cohere. This isn’t a grant you spend and move on from. It’s a stake every Canadian now holds, whether they wanted it or not. Cohere should treat that as a massive responsibility. No amount of government support gets a company to the top of this market. No procurement rule makes a model better. Cohere has to climb on its own, against competitors that are bigger and better funded.
Taking the money was the easy part. Now Cohere has to go build.
Five things that will tell us if it worked
The terms get published. Share class, price, board rights, clawbacks. Real investors never accept a deal on a press release. Taxpayers shouldn’t either.
Cohere competes for government work. If future contracts are open and Cohere wins them on product, fine. If sole-sourcing on “Canadian-developed” becomes the norm, it’s a protected incumbent.
Most of the revenue comes from outside Canada. This isn’t a Cohere rule, it’s the rule for every Canadian company. Exports are how a country of 40 million wins, and a company that sells to the world lifts everyone around it. The number to watch is the share of revenue earned outside Canada. Cohere building for the world is the good version. Cohere leaning on Canadian procurement is the other.
The models are still worth running in 2028. The sovereignty case only holds if a Canadian ministry can use Cohere without falling meaningfully behind people using frontier models. If the gap widens, the thesis doesn’t hold.
Who outside Canada is using it. This is my personal measure, and it’s about names, not percentages. We live in a globalized economy, and a bank in Frankfurt or a telecom in Singapore has every model in the world to choose from. If those logos keep showing up on Cohere’s customer list, the company is winning on product. If the list stays mostly Canadian, it’s sovereignty carrying the case, not the product.
If those break the right way, this was a smart bet and Canada has a global AI company it part-owns. If they break the wrong way, it’s an expensive lesson. Either way, the outcome is now in Cohere’s hands, not Ottawa’s. That’s the part that matters.
For what it’s worth, I’m rooting for them. Anyone who takes a risk this size is good in my books, and I wish Cohere all the best. The reason this story is worth covering is that I’m not a fan of government being involved in business's like these.
Sources
Prime Minister’s Office: Canada announces landmark investment in Cohere
The Logic: Cohere valued at $35.7B as Canada’s government takes a stake
The Globe and Mail: Ottawa buys $1.6-billion stake in Cohere
BetaKit: Feds to buy stake in Cohere as part of planned $7-billion Series E
ISED: Government of Canada finalizes $240M investment in Cohere
Canadian Shield Institute: Cohere investment sovereignty score
Open Government: Shared Services Canada sole-sourced Cohere contract
The Logic: Carney picks researchers and investors for new AI council
The Logic: Ottawa will spend big to back Canadian AI, Solomon says
Published in Fello Foundry by Fello
