"Over the next 5–10 years, the semiconductor, supercomputer, and artificial intelligence (AI) factory industries are expected to grow explosively at the same time. The design and operation of every factory to be built in the future will take place in a virtual twin (virtual space·Virtual Twin)."
Nvidia Chief Executive Officer Jensen Huang said this on the 3rd (local time) at 3DExperience World 2026, the annual flagship event of Dassault Systemes, held in Houston, Texas. He said, "We are now at the starting point of a new industrial revolution," adding, "AI will become the foundation and infrastructure of every industry, like water, electricity, and the internet."
Huang said, "To make this possible, three massive industries must expand simultaneously: semiconductor factories that make AI chips, computer factories that manufacture the supercomputers the chips go into, and AI factories (AI-specialized data centers) where supercomputers run AI," adding, "These three industries are currently growing explosively."
He said, "It takes about $50 billion (about 72.5 trillion won) to build a 1-gigawatt (GW) AI factory, and dozens of GW-scale AI factories are currently under construction worldwide," adding, "This is the largest infrastructure expansion in human history."
Emphasizing the need for virtual twins, he said, "Before building such complex and massive facilities, designing and operating factories in a virtual environment in advance can significantly reduce time and expense." A virtual twin is a technology that implements a virtual model identical to reality, creating a "twin" of real-world objects in virtual space, running simulations, and using them to predict outcomes in advance to optimize products and services. Dassault Systemes is a leading company in virtual twins.
He added, "In the past, products were designed first and factories were then built, but in the future, how a factory is designed will determine what products can be made," saying, "Even now, every factory is designed with CAD (computer-aided design), but going forward, entire factories will be simulated and operated inside a virtual twin."
He continued, "It is expected that within a virtual twin, manufacturing lines will be laid out, sequences optimized, robots organized, and robot AI executed to validate everything from assembly and transport to safety," explaining, "Even after factory completion, AI will continuously optimize the factory's performance, power, temperature, and cooling."
Furthermore, he said, "Virtual twins are needed for the AI infrastructure expansion that will see $85 trillion–$100 trillion invested over the next 10 years, and high-performance computing is essential for this," adding, "Dassault Systemes' virtual twin technology will be built on Nvidia's computing infrastructure."
Nvidia and Dassault Systemes announced a large-scale strategic partnership that day. Dassault Systemes said it would integrate Nvidia's CUDA-X libraries and AI physics, Nvidia AI, and the Nvidia Omniverse platform into its core platforms. The crux of the collaboration is combining Dassault Systemes' virtual twin technology with Nvidia's graphics processing unit (GPU)-based computing infrastructure.
Huang said, "With this collaboration, we expect to work at scales 100 times, 1,000 times, and up to 1 million times larger than before, with productivity also greatly improved," adding, "It will even be possible to design products in real time and run AI robots in factories in real time to produce those products." He added, "We expect all of these changes to become reality within the next 5–10 years."
Nvidia and Dassault Systemes began collaborating on graphics technology in 2002 and have maintained a partnership for more than 20 years. Dassault Systemes has improved performance by applying Nvidia's GPU-based technology to its CAD (computer-aided design) design and simulation solutions, and Nvidia is using Dassault Systemes' MBSE (model-based systems engineering) to simulate and analyze system relationships in advance, reduce design errors, and shorten lead times.
Dassault Systemes CEO Pascal Daloz also said, "We are building a 'knowledge factory' with Nvidia," adding, "The 20th century was an era when industry produced and used physical 'products,' but today's industry is shifting to an era in which knowledge and know-how are produced, and that knowledge and know-how in turn create products."
He emphasized that this collaboration is a step toward the "industry world model," which Dassault Systemes is focusing on. The world model is the so-called "AI that works in the real world," not a chatbot based on a large language model (LLM) like ChatGPT, but an AI model that understands physical laws and possesses manufacturing knowledge. It is a concept similar to physical AI, which Jensen Huang has cited as the next battleground for AI. Daloz said, "To realize physical AI in the true sense, manufacturing knowledge must be combined," adding, "From the standpoint that AI goes beyond simply understanding the world to understanding 'how to make it,' 'industrial know-how' is essential."