Dassault Systèmes unveils virtual companions and industrial AI strategy at 3DEXPERIENCE World 2026

French software giant positions knowledge as currency in the generative economy, announces three AI assistants, and deepens its partnership with NVIDIA.

Thousands of SOLIDWORKS and 3DEXPERIENCE platform users gathered in Houston from 1-4 February for Dassault Systèmes’ annual 3DEXPERIENCE World conference, where the French engineering software company unveiled a fundamental shift in how manufacturing organisations will work with artificial intelligence. The event’s central message: in an AI-driven economy, who makes a product matters less than who knows how to make it.

“In the last century, industry was about producing objects,” explained CEO Pascal Daloz during his keynote. “Now, industry produces knowledge and know-how that generates those objects.” That knowledge – captured in CAD models, simulations and production processes – becomes the real value. And the virtual world where that knowledge lives determines what gets built physically.

Virtual companions arrive mid-2026

The company’s AI strategy materialises through three ‘virtual companions’. Last year saw the introduction of Aura; this year brings Leo and Marie, scheduled for summer 2026 release. Unlike ChatGPT and similar large language models that excel at text, Dassault is building what it calls Industry World Models – AI that understands how the physical world actually works.

“The real world doesn’t consist of text and images,” Daloz stated. “It consists of physics, materials, energy and constraints. In the digital world, many things look great, but in the physical world, they have to work.”

SOLIDWORKS CEO Manish Kumar demonstrated how these AI assistants will support engineers in practice. The three companions each bring distinct expertise: Aura focuses on business and strategic choices, connecting knowledge inside and outside the organisation. Leo concentrates on manufacturability and system integrity – validating whether designs can actually be produced. Marie is the scientist, specialising in physics and materials research.

The naming isn’t arbitrary. Leo references Leonardo da Vinci, the engineer and inventor. Marie honours Marie Curie, the scientist. Aura chose her own name. “We asked our AI agents to come up with a name,” Kumar recalled with a smile. “Aura emerged. When we asked how the system arrived at that, it explained it’s an acronym for ‘Assisting You to Realise your Ambitions’.”

Crucially, these companions train on customers’ specific data. Every company has its own design rules and standards – often unwritten, but embedded in their CAD models and workflows. “What customers want is for the next design to follow these rules,” Kumar explained. The models remain exclusive to each customer; data isn’t mixed between different companies, Dassault claims.

Companions make a real difference.

The time savings are substantial. Kumar demonstrated how a 2D construction drawing can be converted into a fully parametric 3D model within seconds. Another example: a complete steel structure for a water tank, ready in five minutes – without clicking a single command, purely by typing requirements. Engineers can interrogate their designs via chat: how many components are needed, which materials are optimal, and where potential bottlenecks exist.

“AI is here to augment what you do, not replace you,” Daloz emphasised in conversation with Kumar during the keynote. The AI assistants function as sparring partners, but ultimate decisions rest with engineers. “The human making the choice bears the responsibility.”

NVIDIA partnership intensifies

To deliver the immense computing power these AI models require, Dassault Systèmes is intensifying its 25-year collaboration with NVIDIA. “This is the biggest partnership our two companies have ever had,” declared NVIDIA CEO Jensen Huang on the main stage.

The partnership encompasses deep integration of NVIDIA’s technologies – CUDA-X, RTX, AI and Omniverse – into the 3DEXPERIENCE platform. This should accelerate simulations by a factor of 100 to 1,000.

Huang sketched a future where everything is software-defined, from cars and the robots that build them to the factories where that happens. “What used to be pre-rendered or offline simulations becomes the virtual twin vision we always envisioned. Everything now happens in real-time.”

The collaboration is reciproca. NVIDIA, in turn, uses Dassault’s Model-Based Systems Engineering (MBSE) software to design its own highly complex AI factories.

The IP governance challenge

Alongside the virtual companions announcement, Dassault launched SOLIDWORKS 2026, introducing hundreds of improvements in design, simulation and data management. Generative AI accelerates drawing creation and automatically recognises fastener-like components in assemblies. An AI-driven virtual assistant unlocks knowledge directly by summarising community posts, wikis, questions and ideas.

But this digitalisation of knowledge immediately creates a new problem. Once knowledge sits in AI models, who owns it? And how do you prevent your own intellectual property being used against you?

“In the AI era, the biggest problem isn’t so much the algorithms, because they’re reasonably available. The data is what matters,” stated Morgan Zimmerman, CEO of 3DEXPERIENCE at Dassault Systèmes.

Zimmerman illustrated the urgency with French automotive supplier Valeo, which supplies components to Volkswagen, BMW and Renault. Previously, Valeo shared headlight designs safely with these customers – they could only use the headlight and install it in their cars. But now a car manufacturer can use all headlight data Valeo ever shared to learn how to build headlights themselves. “That’s a fundamentally different risk,” Zimmerman explained. “And many companies are only realising this now. That’s why IP lifecycle management becomes crucial.”

Dassault Systèmes has therefore expanded Product Lifecycle Management to what it calls Intellectual Property Lifecycle Management. The 3DEXPERIENCE platform tracks in detail which data may be used for what purposes. “We do very detailed IP tracking, so we understand what happens with this data,” Zimmerman explained. “If the IP owner hasn’t given permission, we don’t use that data for learning processes.” Organisations can specify per dataset whether it may be used for machine learning.

Cloud transition inevitable

That governance and AI capability requires enormous computing power, forcing companies to the cloud whether they like it or not. “No customer can afford the GPU power needed for AI on their own infrastructure,” Zimmerman stated. A GPU costs £200,000 and becomes outdated within 18 months to two years. There’s also innovation speed: “There’s a technological breakthrough every week. On-premise means weekly upgrades. No company can bear those IT costs.”

Dassault Systèmes is therefore building its own AI Factory to offer this infrastructure as a scalable SaaS. That forces consideration of new business models, too. For forty years, Dassault sold licences per user. But how do you count AI assistants? “Virtual companions aren’t real users,” Daloz noted. “Do we keep counting users, or must we measure something else?”

The answer lies in what Daloz calls ‘units of know-how’ and ‘units of work’. Instead of paying per logged-in user, you might soon pay for expertise and availability. Virtual companion Leo can be deployed as a mechanical engineer, chemist or manufacturing specialist – each with certified knowledge. “The same companion has different know-how for automotive than for aerospace,” Daloz explained. And there’s a second dimension: Leo works 24 hours daily. Where a human engineer is available 40 hours weekly, an AI companion delivers 168 hours.

The first release, MD01, is scheduled for March/April 2026. Functionality like image-to-mesh and parametric feature recognition will roll out then. Text-to-CAD and the virtual companions follow in summer 2026.

Enterprise Times: What does this mean

For manufacturing customers, the clock is ticking. Virtual companions arrive mid-2026, but the architecture enabling them – cloud infrastructure, IP governance, new licensing models – must be shaped now. Organisations waiting until the software is available will be too late. By then, technology no longer determines who leads, but who has their IT foundations in order.

The shift has profound implications for IT procurement and governance. Where IT traditionally managed on-premises applications, the role shifts to orchestrating cloud services, monitoring data sovereignty, and implementing IP governance that determines which data may be used for AI training. The question is no longer “can you read it”, but “may you learn from it”.

For the channel, this represents both opportunity and disruption. Traditional license sales give way to consumption-based models around expertise and capacity. Partners who can guide customers through IP governance frameworks and cloud migration strategies – not just implement software – will capture value.

For the wider industry, Dassault’s approach signals a fundamental repositioning. While competitors race to add AI features, the French company is redefining the very economics of engineering software. Knowledge becomes currency, virtual companions become billable resources, and intellectual property governance becomes as critical as cybersecurity.

The race isn’t about having the best AI. It’s about having the best foundation to deploy it. And that race has already started.

The post Dassault Systèmes unveils virtual companions and industrial AI strategy at 3DEXPERIENCE World 2026 appeared first on Enterprise Times.