As AI moves from experimentation to large scale industrial deployments, global demand for accelerated computing and generative AI is surging. The ability to scale AI requires the right infrastructure to keep up.
To accelerate the deployment of complex AI factories, Dassault Systèmes, Quanta Cloud Technology (QCT) and NVIDIA are showcasing during Computex Taipei, the world’s largest AI exhibition, a first-of-its kind initiative leveraging Dassault Systèmes’ model-based systems engineering (MBSE) on the 3DEXPERIENCE platform, QCT’s hardware R&D capabilities and the new NVIDIA Omniverse DSX Blueprint to generate AI factory virtual twins.
“An AI factory is inherently a massive, complex industrial system,” said Stéphane Sireau, vice president, High-Tech Industry, at Dassault Systèmes. “Through the engineering intelligence of Dassault Systèmes’ model-based systems engineering, combined with QCT’s hardware foundation and NVIDIA DSX’s robust framework, AI factory platforms become equipped with complete lifecycle governance, scalability and industrialization for maximum token performance per megawatt.”
The demo shown at Computex leverages the NVIDIA Omniverse DSX Blueprint and integrates the 3DEXPERIENCE platform to enable QCT to engineer and validate rack-level AI infrastructure at scale through an AI factory virtual twin, before its physical construction, supply and deployment. This generative configurator establishes Dassault Systèmes’ model-based systems engineering on the 3DEXPERIENCE platform as the core engineering backbone for AI factory industrialization.
Also at Computex, we’ll join critical digital infrastructure leader Vertiv and NVIDIA to demonstrate how Vertiv SmartRun, the company’s AI factory infrastructure, can be represented both as a physical infrastructure system and a configurable virtual twin.
The demonstrator uses our MBSE capabilities on the 3DEXPERIENCE platform to help capture configuration rules, dependencies and engineering intent.
At the French Tech Pavilion at Computex, we’ll further demonstrate benefits of the AI factory virtual twin we’ll deliver with NVIDIA, diving deep into the engineering, equipment production and construction aspects.
What is an AI factory?
AI factories are essential infrastructure that enable organizations to create value out of data by industrializing artificial intelligence. AI factories produce “tokens,” which are units of data processed by AI models.
AI factories and data centers are not the same thing. While traditional data centers are general-purpose facilities for computing tasks, AI factories are purpose-built systems “optimized for artificial intelligence workloads, with a strong emphasis on AI inference performance and energy efficiency,” according to NVIDIA.
By producing tokens, AI factories allow enterprises to monitor and process reams of data, run custom agents and continuously train their AI models. AI factories are enablers of industrial AI.
NVIDIA’s new DSX platform is engineered specifically for AI factories. It “brings together open source, modular software libraries, application programming interfaces, reference designs, NVIDIA accelerated computing platforms and partner technologies into a common, codesigned platform for AI factory design, deployment and operations,” according to a press release.
How is Dassault Systèmes contributing to NVIDIA DSX AI factories?
Together, Dassault Systèmes and NVIDIA are integrating AI factory development and operations to enable the generation of AI factory virtual twins and optimize designs through systems engineering and simulation. This bridges the gap between concept and deployment, improving efficiency and maximizing token performance per megawatt.
The collaboration establishes a continuous AI factory digital thread spanning design, validation, deployment and operations for enterprises, comprising:
- Reusable virtual twin assets and industrialized architecture: Using OpenUSD-based digital assets, complex AI clusters such as the NVIDIA Vera Rubin platform are transformed into standardized, modularly reusable industrialized reference architecture. This enables the precise and scalable replication and expansion of AI factories globally.
- Configuration-driven deployment governance and traceability: Dassault Systèmes’ MBSE enables high-precision configuration management. This ensures end-to-end traceability from virtual design and physical rack cabling to networking configurations, minimizing disconnects and risks during physical deployment.
- Continuous lifecycle twins from engineering to operations: Within a full-lifecycle virtual twin environment, up-front simulation of physical dynamics and computational fluid dynamics (CFD) for the AI factory allows aligning virtual planning with physical execution.
At 3DEXPERIENCE World earlier this year, we announced NVIDIA is using Dassault Systemes’ model-based systems engineering to design AI factories, starting with the NVIDIA Vera Rubin platform, and integrating it into NVIDIA Omniverse DSX Blueprint and open framework for developing digital twins to design, simulate and operate large-scale AI factories.
What is model-based systems engineering (MBSE)?
MBSE is an engineering methodology that uses digital models as the primary means of information exchange and system representation throughout the lifecycle of a project. It allows engineers to create, analyze and validate system designs in a virtual environment before physical implementation.
In the case of AI factories, MBSE ensures end-to-end traceability throughout the lifecycle of highly complex projects, allowing real-time testing, modeling scenarios, minimizing risks, lowering token cost and maximizing token performance per megawatt.
AI factories and the Generative Economy
Historically, industry has mass produced physical products and finished goods, but we’re transitioning to a Generative Economy in which knowledge and know-how has become the product. Contributing our knowledge and know-how to producing AI factory virtual twins is a prime example of the Generative Economy in action.
“The 20th Century was really about the industry using and producing objects,” Dassault Systèmes CEO and Chairman Pascal Daloz said. “In this new century, industry is producing knowledge and know-how, and the knowledge and know-how are generating the objects. This is where the true value lies. This is where the power is.”
A traditional factory is optimized to turn raw materials into a tangible product. AI factories turn raw data into knowledge, which is then used to create the tangible product. AI factory virtual twins allow the planning, design, testing and validation of AI factories to happen in a virtual environment, so they can deliver better knowledge faster and, in turn, lead to more efficient production of the end product.
Building an Industrial AI platform
At Computex, we’re showing how, with our ecosystem, we are industrializing AI to fuel industrial AI.
Earlier this year, Dassault Systèmes and NVIDIA together announced a new foundation for industrial AI with the introduction of “industry world models” that will help companies design, simulate and operate faster, more reliably and at a scale never seen before.
Industry world models are advanced AI systems, built upon decades of industrial knowledge and know-how, that understand the dynamics of the physical world and use input data to create internal simulations that allow AI to learn and understand before recommending or taking action. These models power virtual twins on the 3DEXPERIENCE platform.
“We are entering an era where artificial intelligence does not just predict or generate, but understands the real world. When AI is grounded in science, physics and validated industrial knowledge, it becomes a force multiplier for human ingenuity,” Daloz said in a press release at the time. “Together with NVIDIA, we are building industry world models that unite virtual twins and accelerated computing to help industry design, simulate and operate complex systems in biology, materials science, engineering and manufacturing with confidence. This partnership establishes a new foundation for industrial AI, one that is trustworthy by design and capable of scaling innovation across the generative economy.”
We also introduced three agentic companions—AURA, LEO and MARIE—that virtually personify decades of scientific knowledge, industry experience and technical know-how.
Available on the 3DEXPERIENCE platform on the Cloud, these agents can reason, simulate and act within an industrial context.
They’re smart enough to reason with you, challenge your assumptions and hold you accountable while you work through complex industrial challenges. They combine decades of Dassault Systèmes’ science with your own knowledge and information from static documents. They will operate in a secured environment powered by NVIDIA NemoClaw and OpenShell.
“In the Generative Economy, industry produces knowledge and know-how that generates objects: this is where true value lies,” Daloz said in a press release announcing the agents. “The time has come to put knowledge to work by creating a new kind of teamwork between humans and Virtual Companions to make the invisible visible and the impossible possible before anything physically exists, accelerating innovation cycles while protecting the most critical assets.”
Frequently Asked Questions
A result of the expanded partnership between Dassault Systèmes and NVIDIA, industry world models combine Dassault Systèmes’ virtual twin technologies with NVIDIA’s AI infrastructure, open models and accelerated software libraries to build new, unique world models that will transform the use of enterprise AI across industries.
Combining Dassault Systèmes industrial knowledge and know-how with NVIDIA AI libraries, these industry world models understand the physics, behaviors and consequences of actions, in order to generate optimal designs and make the best decisions.
Dassault Systèmes’ new Virtual Companions are purpose-built, AI-powered experts for complex industrial problem-solving rather than personal assistance. They operate on the 3DEXPERIENCE platform to help organizations create, test and validate innovations.
We ground our Virtual Companions in science and decades of industry-grade knowledge. They represent a combination of Industry World Models, AI and multi-scale, multi-discipline modeling validated by the laws of physics. They also access libraries of trusted news sources organized by industry, in addition to your unique enterprise data.
AI factories and data centers are not the same thing. While traditional data centers are general-purpose facilities for computing tasks, AI factories are purpose-built digital infrastructure that enable organizations to create value out of data by industrializing artificial intelligence.
AI factory virtual twins allow the planning, design, testing and validation of AI factories to happen in a virtual environment, so they can deliver better knowledge faster and, in turn, lead to more efficient production of the end product. By bridging the gap between concept and deployment, AI factory virtual twins improve quality, lowering token cost and maximizing token performance per megawatt.
Related Resources on Industrial AI
Patrick BallPatrick is a Senior Communications Manager on the Corporate Publishing team here at Dassault Systèmes. An experienced journalist, marketer, speechwriter and storyteller, Patrick's words have appeared on pages and stages around the world.