Canada is buying into its own AI champion. On October 29, Prime Minister Mark Carney announced that the federal government will invest up to C$1.6 billion (about US$1.14 billion) in Toronto-based Cohere, taking an equity stake as part of a Series E round that Cohere says targets up to C$7 billion at a C$35.7 billion valuation. It is one of the largest direct government bets on a single AI company anywhere outside the US and China.
Key takeaways
- Ottawa will put up to C$1.6 billion into Cohere in stages, as an equity investment rather than a grant or loan.
- The money is part of a Series E targeting up to C$7 billion (roughly US$4.5 billion) at a C$35.7 billion valuation; only a first close has been completed so far.
- Germany’s Schwarz Group leads the round, with Nvidia, Inovia Capital, Prosperity7, Bell Ventures and Bullhound Capital also in.
- In return, Cohere commits to staying headquartered in Canada and, according to reports, investing at least C$1.7 billion in Canadian compute infrastructure by 2030.
- For developers and enterprises, the deal strengthens the “sovereign AI” option: models you can deploy on infrastructure you or your government controls.
What happened
Cohere announced the initial close of its Series E on October 29, describing it as the largest private financing ever undertaken by a Canadian technology company. The same day, Carney confirmed in Montreal that the Government of Canada would join the round as an investor. The stake was negotiated through the Canada Development Investment Corporation, a Crown corporation that manages government assets, and the money will flow in stages, according to BetaKit.
The round is led by companies of the Schwarz Group, the German retail giant behind Lidl and Kaufland, which pledged €500 million earlier this year alongside Cohere’s merger with Germany’s Aleph Alpha. Returning investors include Nvidia, Inovia Capital and Prosperity7 Ventures; Bell Ventures and Bullhound Capital are new. Cohere has not said how much has been committed at this first close or when the round will finish, so the C$7 billion figure is a target, not money in the bank.
The government’s terms are notable. Cohere must keep its incorporation and headquarters in Canada and report regularly on the jobs and investments it makes there. Several outlets, citing the government’s announcement, also report a commitment to spend at least C$1.7 billion on new computing infrastructure in Canada by 2030. Cohere’s own release confirms it is “making a number of commitments” without listing them.
Carney framed the deal as a matter of national control over a strategic technology. “The most powerful AI systems are concentrated in a very small number of companies, in a very small number of countries,” he told reporters, according to AFP. AI minister Evan Solomon said the stake gives Ottawa “a seat at the table.”
Who Cohere is, and why it matters
Founded in Toronto in 2019 and led by CEO Aidan Gomez, a co-author of the original Transformer paper, Cohere has never chased the consumer chatbot market. It sells models and tools to businesses and governments: the Command family of language models, Embed and Rerank for search and retrieval, and North, an agent platform aimed at enterprise workflows. Its pitch is that customers can run these models privately, including on their own infrastructure, which matters in banking, healthcare, defence and the public sector.
The company says that pitch is working. In its announcement, Cohere reported annual recurring revenue up 165% year over year at the end of Q3 2026 and a 20x increase in customers over the past year, to 300. Named customers include RBC, TD Bank, Bell Canada, Fujitsu and PwC. These figures are company-reported and have not been independently audited, but they suggest real enterprise traction rather than hype alone.
Gomez called the round “the largest private capital raise in Canadian history” and told AFP that Cohere is now the largest AI company outside the US and China. That second claim is his framing; it is hard to verify given how differently companies report revenue and valuation, but at C$35.7 billion Cohere is clearly in the top tier of non-US, non-Chinese AI labs, alongside France’s Mistral.
Why it matters: the state as shareholder
Governments have funded AI for years through research grants, compute subsidies and procurement. Taking equity in a frontier-adjacent lab is different. It ties public money to the company’s success and gives the state leverage over where the company is based and where it builds infrastructure. It follows the US government’s stake in Intel and Canada’s own plan to buy into miner Teck Resources, and it signals that AI is now being treated like critical minerals or semiconductors: a strategic asset worth owning.
It also reflects a growing transatlantic axis. With Schwarz leading the round and Cohere absorbing Aleph Alpha, Cohere effectively becomes a Canada–Germany champion. Schwarz Digits says it plans to run Cohere’s models on STACKIT, its European cloud. For governments that are wary of depending entirely on US hyperscalers or Chinese open models, that combination of a non-US lab plus non-US cloud is exactly the point.
The scale is still modest next to the US giants. AI startups raised hundreds of billions of dollars this year, much of it going to a handful of American labs, as we covered in our breakdown of Q1 2026 AI investment. A US$4.5 billion round would be a rounding error for those companies’ compute budgets. Cohere’s strategy depends on not competing head-on for the biggest general-purpose model, and instead winning regulated customers who value control over raw benchmark scores.
What it means for developers and businesses
For most developers, nothing changes overnight: Cohere’s APIs, pricing and models are the same today as last week. But the funding has practical implications worth tracking.
- More credible private deployment. If your organisation needs models that run inside its own cloud or data centre, Cohere is now better capitalised to support that over the long term. That lowers vendor risk for multi-year contracts.
- Canadian and European data residency. New Canadian compute capacity and the STACKIT partnership could make it easier to keep data and inference inside Canada or the EU, a frequent requirement in finance, health and government work.
- Retrieval tooling remains a strong point. Cohere’s Embed and Rerank models are widely used in retrieval-augmented generation (RAG) pipelines, even by teams that use other vendors for generation. Continued investment there is good news for anyone building search over private documents.
- A reason to design for model portability. Sovereign AI is fragmenting the market by geography. Building your stack so you can swap model providers, or self-host, is increasingly a sensible default. Our guide to running LLMs locally covers the on-premise side of that trade-off.
For Canadian startups, the more interesting signal may be what comes next. Solomon said the Cohere deal will not be the last time Ottawa takes a stake in a tech company, pointing to a forthcoming C$500 million Canadian Tech Growth Fund that will sometimes take equity in small and medium-sized AI firms.
Open questions and risks
Several details are still unknown. The government has not published the size of the stake as a percentage of the company, the schedule for its staged payments, or what happens if Cohere misses its commitments. Cohere has not disclosed how much of the C$7 billion target is already committed. And public ownership raises fair questions about how Ottawa will balance its role as investor with its role as regulator and customer.
There is also execution risk. Training frontier-level enterprise models, which CFO Francois Chadwick named as a goal for the funds, is expensive, and Cohere must integrate Aleph Alpha and several other 2026 acquisitions while scaling across North America, Europe and Asia.
What to watch next
- The final close of the Series E and the total amount raised.
- Details of the government’s stake and the conditions attached to it.
- Where and when the C$1.7 billion in Canadian compute is built, and which providers supply it.
- New model releases under the Command and North brands, and whether they close the gap with US frontier models.
- Whether other governments, especially in Europe, follow with direct equity stakes in their own AI labs.
The bottom line: Canada has decided that owning a piece of an AI lab is cheaper than depending on someone else’s. For builders, that means one more well-funded, sovereignty-focused option on the menu, and one more reason to keep your AI stack flexible.