Mistral Acquires Austrian Physics AI Startup Emmi — 30 Researchers, Large Engineering Models, and a Second Deal in Three Months
Mistral AI acquired Austrian startup Emmi AI on May 19, adding industrial physics AI to its platform in a deal that accelerates the French lab’s push into manufacturing and engineering verticals. Terms were not disclosed. Emmi’s backer Serena called the startup’s valuation “significant.”
What Emmi AI Builds
Emmi AI, founded in 2024 and based in Linz and Vienna, developed what it calls Large Engineering Models (LEMs) — foundation models trained on the laws of physics rather than text corpora. The models simulate complex physical processes: computational fluid dynamics (airflow over aircraft wings), thermal analysis, material stress testing in automotive crashes, and semiconductor fabrication.
The core pitch: replace multi-day simulation runs with real-time modelling. Emmi cited work with ASML where their integration cut EUV lithography diagnostic time from hours to eight minutes, reducing waste of costly silicon wafers.
Customers in production include Stellantis, Veolia, and defense drone maker Helsing.
The Numbers
Emmi raised €15 million in seed funding last year — Austria’s largest seed round in 2025 — led by 3VC, Speedinvest, Serena, and PUSH VC. The company had more than 30 researchers and engineers at acquisition. All will join Mistral’s Science and Applied AI teams by end of May.
Mistral’s M&A Pace
This is Mistral’s second acquisition in three months. In February, it acquired Koyeb, a French startup that simplifies cloud application deployment. The Emmi deal confirms what Sifted reported earlier this year: acquisitions are now a strategic priority for the lab, which raised nearly €2.8 billion and reached a valuation of €11.7 billion since launching in 2023.
The pattern is deliberate vertical expansion. Mistral CEO Arthur Mensch has said Europe has two years to avoid becoming a US AI infrastructure vassal state. Industrial manufacturing — a sector where European companies hold century-long expertise — is the most defensible vertical for a European AI lab to dominate.
What Mistral Gets
The acquisition adds three things:
-
Physics AI capability: LEMs can simulate and interact with the physical world in ways general-purpose LLMs cannot. This is not something Mistral can build quickly from scratch — the specialization requires deep domain training.
-
Industrial client relationships: Emmi’s aerospace, automotive, semiconductor, and energy client base is enterprise-grade with long contract cycles. They are exactly the clients Mistral wants for its enterprise AI transformation pitch.
-
European footprint: Linz becomes an official Mistral office, alongside Paris, London, Amsterdam, Munich, San Francisco, and Singapore. This extends the company’s hiring reach into the Austrian and German engineering talent pool.
Guillaume Decugis at Serena framed the strategic shift: “Mistral is interested in having a vertical play, where it has historically been more of a horizontal player. We’ve seen with Anthropic that the specialisation of frontier labs can hugely pay off.”
What It Means
Mistral is assembling a stack: language models (horizontal platform) + cloud deployment (Koyeb) + physics simulation (Emmi). The end-state is a vertically integrated AI suite for industrial enterprises — one where Mistral controls the full pipeline from foundation model to domain-specific inference, deployed on infrastructure it controls through Koyeb.
For industrial engineering customers evaluating AI vendors, Mistral is now the only European option that combines frontier language capability with physics-grounded simulation models in a single commercial relationship.
The financial details were not disclosed, but the pace — two acquisitions in 90 days — signals Mistral is spending its €2.8 billion raise faster than its public statements have suggested.