NVIDIA and Dassault Systèmes are expanding a long-running partnership into what both companies promote as a foundational shift in how industrial artificial intelligence is built and deployed. The collaboration centers on a shared industrial AI architecture that combines virtual twins with NVIDIA’s accelerated computing and AI software stack.
The aim is to move industrial AI away from isolated tools and toward what the companies call industry world models, which are large-scale, science-validated systems designed to simulate and optimize real-world products and even biological processes before they exist physically.
NVIDIA CEO Jensen Huang framed the effort as part of a broader transition in computing. In his view, AI is becoming infrastructure, comparable to electricity or the internet, and industry needs models that are shaped by the laws of physics rather than purely by data. Physical AI, as NVIDIA describes it, is meant to understand how real systems behave, not just generate text or images.
Dassault Systèmes CEO Pascal Daloz echoed that theme, arguing that virtual twins should be treated as engines of knowledge rather than mere visualization tools. In manufacturing and engineering, he said, value increasingly comes from what can be learned and validated in the digital world before capital is committed in the physical one.
Extensive Partnership
Dassault Systèmes brings decades of experience modeling complex systems through its virtual twin platforms, while NVIDIA contributes accelerated hardware, AI libraries, and its Omniverse simulation environment. Together, the companies want to enable real-time simulation and AI-assisted decision-making across design and operations.
NVIDIA is already using Dassault’s model-based systems engineering tools to design its AI factories, the company’s data centers built to support large-scale AI workloads. According to Huang, NVIDIA now simulates entire facilities, from networking to cooling, inside virtual twins before construction begins. The same models continue running after deployment, using live data to refine performance over time.
Dassault Systèmes is adopting NVIDIA’s infrastructure and open AI models to power its own platforms. Through its OUTSCALE cloud unit, the company plans to deploy NVIDIA-powered AI factories on multiple continents, with a focus on data sovereignty and intellectual property protection, clearly a growing concern for industrial customers.
The companies claim that this approach addresses a key limitation of today’s generative AI systems. Large language models can summarize information or generate code, but they do not inherently understand physical constraints. By grounding AI in validated engineering and scientific models, the partners believe they can create systems that are trustworthy enough to support mission-critical decisions.
AI as a System of Record
Based on the partnership, in manufacturing, virtual factories could be designed, stress-tested, and optimized entirely in software. In materials science and biology, researchers could explore vast design spaces for new compounds while filtering out options that violate known physical or biological rules. Across engineering disciplines, AI-driven virtual models embedded in Dassault’s platforms are intended to assist professionals with analysis and version updates, rather than replace them.
Analysts see the partnership as a sign of where industrial AI is heading. Instead of standalone AI features layered onto existing tools, companies are beginning to assemble integrated stacks that treat AI as a system of record. The emphasis on physics-based validation also reflects a maturing market, where reliability and accountability matter as much as speed.
