Paris-based Mistral AI has signed a major deal with automaker Stellantis to deploy customized industrial AI models, following a €1.7 billion funding round led by ASML, with which it is also co-developing industrial LLMs.
In sum – what to know:
Industrial AI – Mistral AI expands further into Industry 4.0 with Stellantis deal for enterprise-wide domain-specific gen AI deployment.
Funding round – it follows a €1.7 billion Series C round led by Dutch chip specialist ASML, which will go to fund its industrial AI push.
Domain data – ASML is partnering with Mistral in a co-creation deal of its own to fine-tune models on proprietary manufacturing data.
Paris-based startup Mistral AI is ramping up its new focus on Industry 4.0 in a bid to embed generative AI in complex enterprise environments, starting mostly with IT-facing operations, moving eventually into more sensitive OT affairs. It has just signed with Italian automotive manufacturer Stellantis to start a “company-wide” deployment of a custom industrial version of its large language model (LLM) system, trained on domain-specific data from Stellantis.
The deal follows a year of pilots at the Italian firm, and comes on the heels of a new €1.7 billion funding round, headed by Dutch semiconductor equipment manufacturer ASML. The new money will finance its R&D work around “custom decentralized frontier AI solutions that solve the most complex engineering and industrial problems”, it said. Others in the Series C round included DST Global, Andreessen Horowitz, Bpifrance, and Nvidia, among others.
Mistral AI is in the process of embedding its own architects, engineers, and scientists within enterprise customers to develop domain-specific generative AI systems for them, which in turn embed their industrial and proprietary data and processes – as reported in The Wall Street Journal last week. Arthur Mensch, co-founder and chief executive at the firm, told the Journal: “At some point, the capabilities of the frontier model can only be increased if we partner.”
The point is that general-purpose frontier models from the likes of Mistral AI, alongside higher-profile US firms like OpenAI and Anthropic, have reached a “saturation point”, having practically exhausted publicly-available training data in a bid to “compress human knowledge and make models increase across the board”, explained Mensch. The new frontier, and the real monetization opportunity, is inside enterprises – and Mistral AI is going direct.
As well as providing much of its new Series C funds, ASML is also a direct customer of Mistral AI. The French firm, the tech pin-up for European AI, has installed its engineers at AMSL. They have a ‘co-creation’ deal, reports the Journal, where ASML provides its own proprietary technical data from its semiconductor manufacturing operations, and Mistral SI provides its AI models, expertise, and engineers to help ASML fine-tune or “post-train” those models.
This post-training domain adaptation discipline sees its existing LLM refined on new industrial data – in this case, ASML’s internal manufacturing logs and technical documents, presumably. Which means ASML gets a customised Mistral AI model that understands its terminology and workflows, and answers questions and automates documentation accordingly, plus a bunch of dedicated staff, and some kind of exclusive usage rights.
Mistral AI gets revenue from the co-creation services and engineering support, plus experience and data insights to improve its own foundational models – based on agreed and aggregated privacy terms between the two firms. Christophe Fouquet, chief executive at ASML, said in a press note about the Series C round: “The collaboration… aims to generate clear benefits for ASML customers through innovative products and solutions enabled by AI.”
As such, there is a joint services play to the arrangement as well. Mensch said: “We have the ambition to help ASML and its numerous partners solve current and future engineering challenges through AI, and ultimately to advance the full semiconductor and AI value chain.” The notice also stated: “For the last two years, we have advanced AI through cutting-edge research and strategic partnerships with corporate and industrial champions.
“We will continue to develop custom decentralized frontier AI solutions that solve the most complex engineering and industrial problems. It empowers enterprises, public sectors, and industries a competitive edge through state-of-the-art models, tailored solutions, and high-performance compute infrastructure. This funding round reaffirms the company’s independence.”
Meanwhile, the new contract with Stellantis – which owns Alfa Romeo, Chrysler, Citroën, Dodge, DS Automobiles, Fiat, Maserati, Opel, Peugeot, Vauxhall, among other brands – starts with AI work on sales and after-sales and extends to engineering and other domains, they said. The deal, announced at Italian Tech Week, follows 18 months of work, and “deepens a collaboration that has already delivered tangible results in automotive innovation”, they said.
Stellantis is already using generative AI in a range of use cases, they said – “from “next-generation in-car assistant to AI-powered business and engineering workflows”. They will establish two joint AI departments, they said: an ‘innovation lab’ to co-develop “highly customized AI solutions to tackle complex use cases and embed intelligence directly in Stellantis’ business processes” (sales and after-sales); and a ‘transformation academy’ to scale AI across the whole business, to “new use cases that drive measurable business impact” (engineering and other domains).
The press note talks about “improving quality and unlocking innovation across all functions”; it does not say explicitly if or how generative AI will be used in production processes on the factory floor, but, directly or indirectly, the ambition is that generative AI will make a difference here too. “This partnership turns distributed intelligence into a competitive advantage and strengthens Stellantis’ long-term leadership in the mobility sector,” they said.
Ned Curic, chief engineering and technology officer at Stellantis, said: “Our work with Mistral AI is helping us move faster and smarter. What makes this partnership unique is Mistral AI’s ability to work closely with Stellantis to deliver meaningful results. Together, we are shaping intelligent, adaptable systems that bring real value to our customers and help Stellantis stay ahead.”
Mensch said: “Stellantis is a unique partner for Mistral AI because of its scale, engineering depth and determination to apply AI across every part of the enterprise Our highly customizable AI solutions and our applied engineering teams will be instrumental in delivering practical solutions to improve Stellantis’ customer experience and operations. Together, we’re setting a new benchmark for how AI can transform the most complex industries.”