Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
French artificial intelligence start-up Mistral is in talks to raise €500mn in a deal that would more than double its valuation to at least €5bn, following proposals from multiple investors seeking to invest in the one-year-old company.
Mistral’s three founders, led by chief executive Arthur Mensch, are considering their options just four months after raising about €400mn in a round that valued the start-up at €2bn.
Venture capital and sovereign wealth funds are among those eager to back the Paris-based company, seen as Europe’s best chance of taking on Silicon Valley groups OpenAI and Anthropic, according to multiple people close to the talks.
Mistral has been approached by investors and has begun discussions over a fundraising deal that would value the company at €5bn or more, according to three people familiar with the situation. If the company pushed ahead, it was likely to raise as much as €500mn, said two of the people.
Mistral declined to comment. The Information first reported on the proposed €5bn valuation figure.
The company’s rapid rise to prominence is striking even by the frenzied standards of an AI boom that has played out over the past 18 months.
Founded in May 2023 by three former Meta and Google AI researchers, Mistral has released several versions of its AI model, unveiled a chatbot called Le Chat that is similar to OpenAI’s ChatGPT, and begun to generate revenue in the past few months.
The company has attracted backing from prominent venture capital funds, including Lightspeed Venture Partners, Andreessen Horowitz and General Catalyst as well as Nvidia, Salesforce and former Google chief Eric Schmidt, over two funding rounds.
“Arthur [Mensch] is being asked for autographs on the street. He’s embodying all the hopes people had for Europe for so long,” said one investor in the company. They said the excitement around Mistral made fundraising so soon after the last round a real possibility, adding however that Mistral had enough money to continue building its latest AI model.
Like its rivals, Mistral is advancing generative AI by building so-called large language models (LLMs), which are capable of producing humanlike text and code responses to natural language prompts in seconds.
Developing “frontier models” — LLMs which push forward the capabilities of today’s best AI tools — requires enormous amounts of computing power and is hugely capital intensive. According to two people with knowledge of the process, new capital raised by Mistral would be spent on securing the chips needed to train and run its models.
Unlike Microsoft-backed OpenAI or Google, Mistral is taking an “open-source” approach, meaning the code underpinning its model is publicly available and allows developers to freely build on top of it.
To fuel adoption of its technology, it has signed partnerships with companies such as Microsoft, Databricks, Snowflake, and Amazon’s AWS, and several of those have also taken small stakes in the start-up.
Whether Mistral can keep up with deep-pocketed tech giants and acquire enough AI chips remains an open question. But it has so far built and trained its models at a lower cost and scale than rivals such as OpenAI, Anthropic and Google. According to Mensch, Mistral launched its first AI model with a team of just 10 people and trained it for less than $500,000.
Mistral’s emergence as a contender in the fast-moving field of artificial intelligence has made it a symbol of the French government’s efforts to ensure the country and the EU do not again get left behind by US and Chinese rivals.
President Emmanuel Macron has advocated for the EU to take a light approach to regulating AI so as not to choke off innovation. He pushed for changes to the AI Act, which was finalised by lawmakers in Brussels in March.
Read the full article here
