Last May, French president Emmanuel Macron gathered hundreds of guests to the Elysée Palace in Paris in an event dedicated to “France’s top AI talent”.
Speaking to the room about his hope to see France become a top player in the AI sector, Macron said: “We’ve got a head start: we’re pretty good in the disciplines that are necessary for AI.”
“Mathematicians and data scientists are rare and sought-after resources. France trains many, and of a high quality.”
But despite the country’s AI pedigree, that talent is now becoming a scarce resource — with French researchers and engineers coming on the radar of an ever-growing number of AI players in the past 18 months.
A host of Paris-based AI startups, including Mistral, Poolside and H, have made remarkable debuts featuring record-breaking fundraises — and are now recruiting en masse. In the face of their new competitors, Big Tech companies that are established in the French capital are also courting top engineers — and are prepared to hand out big cheques to recruit and retain the best candidates.
Paris has become the theatre of an ‘AI talent war,’ founders tell Sifted.
“It’s totally something I have witnessed for the past two years or so,” says Thomas Clozel, the founder of French AI unicorn Owkin. “We’re all in competition with all the companies doing AI in Paris, and the competition has become much harder.”
Solid foundations
France has long prided itself on its strong AI talent, often embodied by the country’s celebrity computer scientist Yann LeCun, who has led Meta’s AI research department for the past decade.
“France has a card to play in AI,” Mathias Frachon, the cofounder of tech recruitment company The Product Crew, tells Sifted, “because the skills and level of excellency that we have in fields like maths is really notable.”
A big reason for this comes from academia. A number of prestigious French universities, including Ecole Polytechnique, CentraleSupelec and Ecole Normale Supérieure (ENS), are specially known for producing some of the country’s most famous AI brains. Mistral’s cofounders, for instance, trained at ENS and Polytechnique.
It’s this long standing pool of talent that led Facebook's AI Research Lab (FAIR) to open in Paris in 2015, and Google DeepMind to follow suit in 2018.
Poaching startup talent
Hiring top-notch technical staff is imperative for companies competing to build state-of-the-art AI models and products –– which is why the profiles that are most coveted are those focusing on R&D and engineering.
Companies are after candidates with excellent academic and professional backgrounds to stay at the forefront of the AI race. These are a minority — and it’s where the competition is at its toughest, says The Product Crew’s Frachon.
“There’s a war only for the top 1%,” he says, “but they are superstars and everyone is fighting over them.”
A research scientist based in Paris working for a French AI scaleup, who chose to remain anonymous, tells Sifted that they’ve personally experienced this over the past few months. They say they’ve been propositioned by Meta’s LLAMA team — which builds the tech giant’s large-language models — twice in recent months.
They declined the offer and were not given a compensation range — but say that they could have probably doubled their salary by joining Meta’s ranks.
The Big Tech exodus
The talent competition between deep-pocketed Big Tech companies’ and startups is not new. But, in Paris, it’s a trend that has accelerated significantly in recent months as the AI scene has become more crowded with buzzy startups, as the city has seen impressive seed rounds for companies like Bioptimus ($35m) and FlexAI ($30m).
Many of the city’s biggest startups were founded by, and have recruited Big Tech alumni. H’s original founding team included four ex-DeepMind researchers, while three of Bioptimus’ founders are former Google scientists. Mistral’s founding team is ex-DeepMind and Google.
An analysis carried out by Sifted of 50 research and engineering profiles listed on LinkedIn as having joined Mistral in the past year shows that eight employees came from Meta and six from Google and DeepMind.
Dali Kilani, the cofounder of FlexAI — which launched in April and has been fast-recruiting since — tells Sifted that he has also been taking in candidates from the ranks of Big Tech, with employees coming from YouTube, Oracle, GitHub and AWS.
He says that startups have a card to play against Big Tech to convince candidates. “Some people had been in comfortable positions for 10-15 years, but we managed to convince them,” he says.
“They were tired of being subordinate to the US, where all the decisions happen. The real argument in our favour was the ambition of the project and the promise to be at the heart of decisions.”
Owkin’s Clozel says that there has been an “exodus” of employees from Big Tech companies to startups. “It’s becoming harder to identify with the mission of Big Tech companies,” he says. “People increasingly want to own what they’re creating.”
Fight amongst yourselves
The AI talent exodus from Big Tech will certainly be sounding alarm bells in the head offices of Silicon Valley’s biggest companies, but the fight is also on among startups themselves.
Flavien Coronini, the talent acquisition lead at French AI scaleup Hugging Face, says that the company’s top profiles have been in high-demand for the past months.
“All these AI startups did astronomical seed rounds,” Coronini says, “and they are now trying to headhunt Hugging Face and other startups' employees, and have the means to make offers with big pay packages."
"However, money is not our employees' main incentive. Having an impact on Hugging Face's growth is often more important to them.”
Delphine Groll, cofounder at Paris-based AI healthtech scaleup Nabla, says that competition from other startups is even fiercer than from Big Tech.
“Our engineers, most of whom come from Big Tech, wouldn’t go back because the focus of those companies no longer appeals to them,” she says. “The real competition is sometimes with other AI startups like Mistral, Poolside, and Photoroom, which can offer high salaries.
“But our AI teams are committed to Nabla for the mission — healthcare is a field where they want to make a real impact"
“People who joined us only four to six months ago are being approached constantly,” says FlexAI’s Dali. “There’s a lot of competition between startups.”
Sifted’s analysis found that Mistral's research team also includes former employees of Hugging Face, Nabla and Owkin.
The price of talent
It’s too early to tell whether the battle for talent is having a lasting impact on AI salary ranges in Paris — but it’s certain that there have been signs of inflation in recent months.
Typically, for a research or engineering role in today’s market, employees can expect between €55k to €130k depending on experience, according to two sources with direct knowledge of pay packages in AI companies. One source says that Mistral offers about 15% more. Mistral did not return Sifted's request to confirm.
The figure seems to be on the way up: in July 2023 Nabla told Sifted that AI companies in Paris could expect to pay salaries of around €90k for senior ML engineers.
And Big Tech companies are able to significantly outbid startups. “I was in a situation once where the candidate was offered three times the salary by a Big Tech player,” says FlexAI’s Kilani.
“That’s the super-power of Big Tech, and in these cases you can’t fight.”
Owkin’s Clozel says that he doesn’t have a choice but to align with some of these offers. “If it’s a profile you need to keep, you align — with a mix of salary and equity,” he says.
Paris is still a way off the salary ranges seen in other regions: some of the engineering roles advertised by OpenAI in San Francisco surpass $400k. But Clozel says that top talent no longer comes for cheap in France.
“In data science, talent is expensive now,” he says, “and that’s normal.”