Mistral becomes the talk of Davos as business leaders seek AI gains

Stay informed with free updates

Simply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.

As artificial intelligence dominated conversations between political and business leaders at the World Economic Forum this week, a relative newcomer emerged as the toast of Davos: French AI start-up Mistral.

Chief executives of three large tech companies told the Financial Times the group’s latest AI model was one of the best available products, according to technical benchmarks used by companies to evaluate their performance.

One US Big Tech executive said that nine-month old Mistral was doing “a great job” competing against sophisticated models made by large US companies like OpenAI and Google.

The interest in Mistral’s technology challenges the prevailing view that the race to dominate generative AI — systems that can spew out humanlike text, media and code within seconds — has become a two-way battle between Google and the multibillion-dollar alliance of Microsoft and OpenAI.

In Europe and elsewhere, rising interest in Mistral has rekindled the possibility that later entrants can secure a meaningful share of a fast-growing market, as costs of computation in AI development come down steadily. 

Mistral, which builds large language models, the underlying technology that powers generative AI products such as chatbots, secured a €2bn valuation last month in a funding round worth roughly €400mn. But it faces even better funded rivals, such as OpenAI, the maker of ChatGPT, which is estimated to be worth $86bn.

The French start-up is backed by investors such Silicon Valley heavyweights including General Catalyst and Andreessen Horowitz.

This week, the company brought on Nvidia, considered the world-leading maker of AI chips, as an investor and strategic partner. Florian Bressand, Mistral’s chief business officer told the FT that the move provides the start-up with access to the chip company’s latest innovations.

Microsoft’s chief executive Satya Nadella recently called out the Paris-based company — founded by Arthur Mensch, Timothée Lacroix and Guillaume Lample, a trio of former Meta and Google researchers — as one of the innovators building AI on its Azure platform.

The comment comes despite Microsoft’s $13bn investment in OpenAI to date, making it San Francisco-based start-up’s largest investor. Last November, OpenAI’s chief executive Sam Altman was ousted briefly by his board, a move that shocked the business world and forced companies to consider diversifying their AI providers.

“Companies can’t just rely on a single supplier,” Mistral’s Bressand said.

He added: “We are working with large corporations: we have 10 interesting proofs of concepts with companies all over the world, and not only in France. Half of the usage from our platform comes from the US, which is not a surprise because it is a more mature market.”

The company exemplifies the current divide in the AI community between open and closed-source models — or between systems whose technical details are transparent to third parties, compared with those that remain proprietary to a single company.

Mistral, whose founders helped build Meta’s open-source LLAMA model, say this will give it an edge with companies that need to build customised features in their software.

Open-source models were particularly attractive to state-owned or highly regulated entities, such as defence companies or banks, who wanted to experiment with generative AI but could not do it with proprietary software because of compliance reasons, Bressand said.

BNP Paribas and Salesforce, which are both also investors, were among companies testing Mistral’s models, he added.

The start-up’s executives attended the WEF as part of a delegation of French start-ups led by President Emmanuel Macron. Other investors include French telecoms billionaire Xavier Niel, and state investment vehicle BPI France.

While its founders have rejected job offers from US AI companies and chosen to return to France to start their business, Bressand said the start-up is “keeping public authorities at arm’s length”, fighting the perception that AI is becoming a matter of national interest. He added the French state has “of course, no say in our governance”.