The next wave of artificial intelligence value creation will perhaps play out in the software that designs airplanes, nuclear reactors, electronic components or data centers. This is the bet of NP Company, a young Parisian start-up from Inria which announces a fundraising of 6 million euros in pre-seed.
The roundtable is led by Partech and brings together several leading figures in the French artificial intelligence ecosystem, including Guillaume Lample and Cédric O, co-founders of Mistral AI, Florian Douetteau, founder of Dataiku, as well as Vincent Luciani, founder and CEO of Artefact. The Peugeot family office is also participating in the round table.
Founded in 2025 by Emmanuel Menier and Matthieu Nastorg, two doctors from Paris-Saclay University specializing in artificial intelligence applied to simulation, NP Company is developing a new generation of simulation software intended for the aeronautics, defense, energy, electronics, data centers and automotive sectors.
The company’s ambition is to tackle one of the main obstacles in modern engineering: the calculation time necessary to validate a new design. Even today, certain physical simulations can mobilize computing infrastructures for several hours, even several days. Each design modification then requires restarting a new series of calculations, considerably slowing down industrial development cycles.
To get around this constraint, NP Company transposes the principles that enabled the emergence of major language models to physical simulation. The company develops foundation models trained on industrial physics data and based on a Transformer architecture, similar to that used in generative artificial intelligence systems.
“Engineering is the next great frontier of artificial intelligence, not the next chatbot,” explains Emmanuel Menier, co-founder and CEO of NP Company. “For decades, simulation has been the primary bottleneck in industrial design. We want to remove this constraint so that engineers spend their time solving complex problems rather than waiting for a calculation to finish. »
According to the startup, this approach already makes it possible to obtain performance gains of around 1,000 times on certain industrial benchmarks, particularly in aeronautics. Ultimately, NP Company estimates that it will be able to achieve speedups of up to 50,000 times on simulations of complete assemblies, among the most complex to process in the industry.
Beyond simply reducing calculation time, the startup seeks to transform the way products are designed. Where engineers are today forced to limit the number of variants studied for reasons of computational cost, these models would allow thousands of design options to be explored simultaneously.
“We were not looking to build a faster version of existing software,” emphasizes Matthieu Nastorg, co-founder and CTO. “With a pre-trained foundation model, it becomes possible to explore thousands of designs in the time it previously took to analyze just one. This opens the way to a new way of designing industrial systems. »
This approach is part of a broader movement that sees artificial intelligence tackling physical and industrial problems. After having automated part of the intellectual work, the foundation models are now beginning to be applied to the equations which govern planes, batteries, electrical networks, semiconductors and even digital infrastructures.
The funds raised will allow NP Company to accelerate the recruitment of researchers and engineers specializing in artificial intelligence and digital physics. In the longer term, the company plans to develop automated design tools capable of proposing optimized architectures themselves, as well as real-time simulators intended to control complex industrial infrastructures.
