Air Liquide, the École normale supérieure (ENS-PSL), and the ENS Foundation announced the creation of a research chair on September 14, 2026, focused on artificial intelligence for chemistry and through chemistry. The chair is the first launched under normalesup.ai, ENS-PSL's new interdisciplinary AI initiative, and aims to accelerate scientific discovery while addressing industrial, energy, and environmental challenges.
The chair will combine machine learning advances with fundamental chemical knowledge to build AI models that require less data, respect physico-chemical laws, and help discover new molecules and materials. Beyond research, it will train students, doctoral candidates, and early-career researchers at the intersection of AI and chemistry - a field both institutions see as strategically important for the future of research and industry.
Why chemistry needs a different kind of AI
AI is already changing how scientists approach chemistry. Predictive methods can now anticipate molecular properties and design new materials before a single experiment is run. But standard machine learning models often demand enormous datasets that simply don't exist for many chemical problems. The chair's research will tackle that gap directly, developing models grounded in physical and chemical laws rather than relying on data volume alone.
For Air Liquide, the motivation is practical. The company uses AI across its value chain - from production to customer services to innovation - where it compresses the time needed to bring new technologies to market. Armelle Levieux, a member of Air Liquide's Executive Committee who supervises Innovation & Technology activities, said: "Artificial intelligence is profoundly shifting scientific and industrial innovation processes. Through this chair, Air Liquide strengthens its collaboration with one of the world's most renowned academic institutions in the field of AI and enriches its own innovation approach regarding AI applications."
Inside normalesup.ai's interdisciplinary push
ENS-PSL launched normalesup.ai in January 2026 to unite its researchers across mathematics, computer science, physics, chemistry, biology, cognitive sciences, and the humanities. The goal is fundamental research that produces scientists capable of working on AI's hardest problems - not just in one silo, but across disciplines. The chemistry chair is the first concrete research program to emerge from that structure.
Frédéric Worms, Director of ENS-PSL, said the chair "embodies the very heart of this project, dedicated to addressing the challenge of artificial intelligence across all disciplines and in their dialogue with one another." He added that Air Liquide's support helps "generate new knowledge, train the talents of tomorrow, and bolster independent French research and its international standing."
The industrial stakes
Air Liquide serves 4.3 million customers and patients across 59 countries, with roughly 65,000 employees and revenues close to 27 billion euros in 2025. The company supplies gases like oxygen, nitrogen, and hydrogen to industries ranging from electronics to healthcare. Faster materials discovery and process optimization through AI could shorten development cycles for technologies in all three sectors.
For researchers following the AI for Science & Research space, the chair's structure is notable. Rather than applying off-the-shelf AI tools to chemistry problems, it will build new models from the ground up - models that incorporate domain knowledge rather than treating chemistry as just another data problem. Researchers interested in similar approaches can explore an AI Learning Path for Research Scientists that covers techniques relevant to this kind of cross-disciplinary work.
Why this matters for research scientists
The chair signals a shift in how AI research is funded and structured at the academic-industrial boundary. For chemists and materials scientists, it means access to models built with chemical constraints baked in - not black-box predictors that ignore the periodic table. For AI researchers, it offers a testbed where data scarcity is the norm, forcing model architectures that can reason from physical principles rather than statistical correlations alone. The emphasis on training early-career researchers also suggests a pipeline of talent specifically equipped to work across both fields, which could shape hiring and collaboration patterns in the years ahead.
Your membership also unlocks:
