They have a business concept, prestigious French engineering degrees, stints at US tech groups and now €105mn to play with. Thirtysomethings Arthur Mensch, Timothée Lacroix and Guillaume Lample became the EU’s new AI darlings this week, with a record seed fundraising round.
Their spectacular financing success — it was Europe’s largest-ever seed round — values the one-month-old business, Mistral, at more than €240mn on not much more than hope: that the company, which is aiming to build an open-sourced language model with B2B applications, will become a European AI champion and give the bloc a strategic stake in how the industry is shaped and regulated globally.
The hype surrounding the French trio of former Meta and Google researchers is evident. The great and the good of Paris, including telco billionaire Xavier Niel, shipping magnate Rodolphe Saadé and the Decaux family, have rushed to buy a ticket alongside leading investor Lightspeed Venture Partners. Jean-Charles Samuelian-Werve, founder of French health-tech unicorn Alan — worth more than €2.7bn — is advising the start-up. Meta AI chief scientist Yann LeCun, a Frenchman, is an enthusiastic supporter.
As is often the case in France, the state is not far away: BPI France, a state investment vehicle with a mission to nurture “national champions”, is also a backer.
The high-profile fundraising shows that France’s top universities, including École polytechnique and École normale supérieure — from which the Mistral founders graduated — continue to produce alumni with cutting-edge skills. The same is true of the other European countries: the UK is where nearly half of the 130 or so European AI start-ups are based, according to data compiled by Sifted, followed by Germany and the Netherlands.
It is hard not to see an element of Fomo in the Mistral frenzy. But it is worth noting the founders have agreed to a big dilution — more than 40 per cent, when it is typically about 10 per cent to 20 per cent at the seed stage — to fund the expensive computing power needed to train their model. And investors in the region don’t have a lot of options. AI fundraising targeting European start-ups was a meagre $4bn this year, compared with $25bn in the US, according to Dealroom.
Regulators, meanwhile, don’t want to be caught out by a technology with revolutionary potential: what could hurt these start-ups and their investors is the Artificial Intelligence Act being cooked up in Brussels.
Draft legislation, which is expected to be finalised by the end of the year, imposes tough restrictions on privacy, transparency and would ban some technology.
Developers of generative AI models like ChatGPT would have to disclose content that was generated by AI and publish summaries of copyrighted data used for training purposes. The US and China have already published their own rules. Britain will host a global summit on AI safety this year.
Interviewed by the Financial Times this week, Mensch issued a warning to Brussels: the EU draft legislation “in its current state [would] make it very hard to actually innovate in a field whose implications we don’t really understand yet”.
Questioned by the 30-year-old newbie entrepreneur on stage at a Paris tech gathering, French president Emmanuel Macron on Wednesday said he shared his “worries” about the EU AI act.
“The worst scenario is a Europe that would invest much less than the US or China [in AI] and that would not be able to create big champions, but that would start by regulating,” he said. “This scenario is possible . . . I don’t want [the EU] to set things in stone too early.”
He went on to outline a plan to encourage innovation by promoting AI “clusters” funded with a mixture of public and private money.
Regulation did need to happen, he said, but not without collaboration with Washington and London, a reference to the need to align western partners in the context of tensions and rivalry with China. The French president received a polite answer from Mensch: “These are good first steps.”