Mistral AI chief executive Arthur Mensch attends an AI summit at Imperial College London in July. The company plans to open an office in Montreal, and Mr. Mensch said it is interviewing candidates.LUDOVIC MARIN/Reuters
French company Mistral AI is scouting for customers in Canada and plans to hire artificial intelligence scientists in Montreal, chief executive officer Arthur Mensch said on Monday.
Mistral builds large language models and competes with OpenAI, Anthropic and Toronto-based Cohere Inc. The company is already working with European customers operating in Canada, Mr. Mensch said, adding that the country’s manufacturing and logistics sectors are opportunities as well.
The Paris-based CEO is in Montreal this week for All In, a two-day AI conference that starts Wednesday. “We’ve been spending some time in Canada for the last few months,” he said in an interview Monday. “There’s a lot of interest from the financial sector, and from the public sector.”
Cohere raises US$500-million, hires former Meta AI expert Joelle Pineau
Variational AI inks drug discovery deal with Merck worth up to $349-million
Mistral bears a few similarities to Cohere. Both are seen by their respective home countries as sovereign AI champions and alternatives to the U.S. tech giants. Like Cohere, Mistral has a heavy focus on business and public sector customers, as opposed to consumers, and works closely with clients to integrate AI.
“We go much more in-depth with enterprises,” he said.
Mistral does have a general purpose chatbot called Le Chat, however, which French President Emmanuel Macron urged people to download in a television interview earlier this year.
Mistral is smaller than its U.S. competitors and raised US$2-billion in September at a valuation of roughly US$14-billion. The financing round included funds from ASML, another European champion. The Dutch company is the only manufacturer of complex lithography machines that are crucial for producing advanced chips.
The sovereignty angle, along with the fact that customers can run AI models on their own infrastructure so that data does not flow back to Mistral, is helping the company win business in Europe, Asia and Africa. “What we see in Canada is that this is also a key aspect of our conversations here,” he said. “The dependency to the U.S. hyperscalers is a problem from an economic perspective.”
That may still be a tough sell with the Canadian public sector. The federal government, for one, has said it will use procurement policies to support domestic companies.
Mistral intends to open an office in Montreal, although there is no specific timeline, and Mr. Mensch said it is interviewing candidates. The company only recently hired its first employee in Canada. He’s familiar with Montreal, having completed a six-month internship at McGill University in 2014 working on machine learning for cardiac imaging.
He later worked at Google’s DeepMind office in Paris helping to build LLMs and left in 2023 to co-found Mistral. “We realized we could actually build a business in Europe, and we realized that the field was not taking the direction we wanted it to,” he said. AI development was becoming more centralized around large tech companies, for one thing. Unlike some of these players, Mistral has released open-weight models and source code, meaning they can be downloaded and fine-tuned for specific applications.
A number of reports lately have found that the investment in generative AI has yet to pay off. A study from the Massachusetts Institute of Technology in July said that 95 per cent of organizations are seeing zero return on investment.
Mr. Mensch said such findings show the importance of working with experts when implementing AI, and directing it at the right problems. “Instead of just equipping your employees with AI tools, which is useful but definitely not sufficient, enterprises need to look at what is driving their costs, what is driving the revenue, and how AI can be integrated,” he said. “That takes more sweat than just creating prototypes.”
He gave the example of CMA CGM Group, a France-based shipping company that Mistral has worked with to automate some aspects of its operations. The process of accounting for and directing containers as they come off cargo ships is complex and involves multiple pieces of software. Mistral was able to develop AI agents to assist and reduce costs by 80 per cent, Mr. Mensch said.