For its first summit, this May 28 at the Carrousel du Louvre, Mistral laid down its cards: becoming the full-stack champion of AI on the Old Continent. His argument: sovereignty and efficiency.
D-day this May 28. With great fanfare, Mistral AI chose the Carrousel du Louvre to install its summit soberly named “AI Now”. A French demonstration of force, served by an American-style scenography, stage projection, light effects and saturated rock at the opening, in a mythical place designed to seduce foreign customers and prospects, primarily European, that the French start-up intends to conquer.
Attract foreign companies
“AI Now” was not designed for the Parisian ecosystem alone: in the aisles, between two sessions, we encountered Italians, Germans, Spaniards and Scandinavians, who had come to assess the European champion of AI in person.Generative AI. The format, on the other hand, was decisive. Where the French tech event is often content with a stage and a few round tables, Mistral offered an American-style show: three stages in parallel, a carefully orchestrated casting of managers and slick staging. Mistral no longer only wants to exist in the face of the American giants, it wants to establish itself as the natural interlocutor of European companies and push its models well beyond French borders.
The casting, however, betrayed a more French reality. The vast majority of use cases presented on stage came from French companies. Rodolphe Saadé for CMA CGM, Patrick Pouyanné for TotalEnergies, without forgetting EDF and BNP Paribas (quoted on several occasions to praise Mistral’s growing capabilities in finance). Customer feedback still sounds very French: lots of testimonies of principle, managers who come to say all the good things they think of Mistral, but rarely any concrete demonstration (unlike its American competitors who often offer more concrete and quantified feedback).
Sovereignty and easy deployment
Two arguments come up with striking regularity among Mistral’s French and European customers: sovereignty and ease of deployment of models. Two advantages that American laboratories cannot really offer to a European customer. At CMA CGM, as at EDF, we boast lightweight models that are easy to integrate and operate as close as possible to your own infrastructures. It is the defense which pushes the sovereign argument to its height: Bertrand Rondepierre, director of AMIAD, the ministerial agency for defense artificial intelligence, makes it his number one criterion.
But it is yet another dynamic which irrigates the entire summit, and we we already wrote it earlier : the transformation of Mistral into a full-stack AI company. Where most players limit themselves to one link in the chain, the model for some, the cloud for others, Mistral intends to control everything: the models, the software layer and, now, its own hardware. A rare vertical integration, which changes the very nature of the company and gives it end-to-end control over the AI production chain.
Added to this vertical integration is another force, just as decisive: efficiency. An obsession that runs through all of the start-up’s work. Rather than rushing to arm computing like OpenAI, Anthropic or xAI Mistral is banking on leading models in terms of intelligence without requiring dozens of GPUs. This is even Mistral’s original bet, held since its very first models: to aim for the best intelligence/cost ratio rather than the race for size.
A choice that fits perfectly with European reality, where computing power remains rarer than across the Atlantic. Mastering the models, software and hardware requires flawless execution on three fronts at the same time, where the American giants have disproportionate resources. Above all, Mistral will have to prove that it can convince its historical supporters beyond the reassuring circle of the CAC 40 and the public sector.
