A Toronto-based AI company that has spent three years positioning itself as the non-American alternative to OpenAI is now in talks to raise money on a scale no Canadian private startup has seen. Cohere is discussing a round of $2 billion to $3 billion at a valuation of about $20 billion, according to four people familiar with the matter cited by the Globe and Mail on September 11.
The valuation is nearly three times the $7 billion the company was worth a year ago. That jump, if the round closes, would make Cohere the most valuable private company in Canadian history and the financing the largest ever raised by a private startup in the country.
The proposed round has an unusual shape. The Canadian government is expected to contribute, according to people familiar with the talks, with existing shareholders following on. Germany and the Schwarz Group, the retail conglomerate behind Lidl, are also in discussions, and Schwarz has committed $600 million, the report said.
Cohere’s pitch is sovereignty. The company sells large language models to enterprises that want an AI partner based outside the United States, one that will store their data under rules they can trust and that does not answer to a Silicon Valley parent. Europe and Canada have become its natural markets, and a German industrial conglomerate writing a large check fits that story exactly.
The numbers that underpin the valuation are small by the standards of the AI leaders. Cohere’s revenue run rate was roughly $240 million last year, a fraction of what OpenAI and Anthropic generate. The $20 billion figure implies a multiple that only makes sense if buyers believe the sovereign-AI niche will grow into a much larger market, and that Cohere will hold a leading share of it.
The company’s backers have so far been willing to make that bet. Nvidia, Oracle and Salesforce have all invested, drawn by a team led by Aidan Gomez, one of the authors of the 2017 paper that introduced the transformer architecture at the core of modern AI. That pedigree gives Cohere a claim to technical seriousness that few of its competitors can match.
What Cohere has not done is chase consumer attention. There is no mass-market chatbot, no viral product, no attempt to outspend the giants on advertising. The company sells to banks, insurers and governments, and its model lineup is tuned for the kind of quiet, high-stakes work those buyers do. Analysts said that focus is exactly what makes it attractive to a government and a conglomerate rather than to a venture firm chasing a consumer hit.
The talks carry risk. A round this size, with a sovereign government and a foreign conglomerate at the table, involves more than price. It involves questions of national security, data residency and control that venture rounds rarely raise. Those questions can slow a deal even when the money is committed, and the people familiar with the matter cautioned that the terms could still shift.
If it closes, the round would give Cohere the capital to build out its own computing capacity rather than renting it, a step the company has said is necessary to serve European clients who refuse to let data cross the Atlantic. That is an expensive ambition, and the $2 billion to $3 billion is priced to pay for it.
Cohere’s founders bring a claim to technical authority that money alone cannot buy. Aidan Gomez was one of the eight authors of the 2017 Google paper “Attention Is All You Need,” which introduced the transformer architecture that underlies nearly every major AI system today. The company was founded in 2019 and has raised from Nvidia, Oracle and Salesforce, three names that signal where it intends to sell rather than merely who wrote the checks.
The sovereign-AI market it is chasing has real money behind it. European governments and companies have watched American and Chinese firms pull ahead and have responded by funding homegrown alternatives, from France’s Mistral to Cohere itself. A German conglomerate committing $600 million is the clearest expression yet of that demand, and it suggests the next phase of the AI race will be fought as much in capitals as in data centers.
For the broader AI industry, a Cohere round at $20 billion would be a data point in favor of the idea that the market can support more than one or two winners. The sovereign-AI story, once dismissed as a niche, is now attracting the kind of capital that suggests it has moved to the center of the debate over who controls the technology.
The round’s size also reflects how much it costs to compete in AI at the frontier. Training and serving large models requires computing clusters that run into the hundreds of millions of dollars, and a company that wants to serve European clients without sending data to American clouds must own or control that capacity itself. The capital Cohere is raising is, in part, the price of that independence.
