Autodesk is developing neural CAD, a new class of AI foundation models trained to understand and generate precise 3D geometry and CAD data, fundamentally changing how designers, architects, and engineers interact with design software. Unlike general-purpose AI image generators like Midjourney, neural CAD models are specifically built to reason about the three-dimensional physical world and produce fully editable, professional-grade CAD objects that can be immediately refined in traditional design software.

What Makes Neural CAD Different From Traditional Design Tools?

For over 40 years, parametric CAD software has dominated the design world, offering precision and control but requiring users to master complex interfaces and exact specifications. Parametric CAD engines are deterministic, meaning if you type "2.5 inches," the measurement will be exactly 2.5 inches. However, this rigidity makes it difficult to explore fuzzy design concepts in an open-ended way.

Neural CAD represents a fundamentally different approach. Rather than relying on traditional parametric engines, neural CAD uses machine learning techniques trained on precise geometry and 3D representations. It learns how to structure and analyze geometry in two- and three-dimensional space, similar to how other generative AI systems create text or images, except neural CAD produces 3D CAD geometry.

"Imagine Midjourney for CAD, but with fully editable results and a smoother, more intuitive human-to-computer interface," explained Mike Haley, Senior Vice President of Research at Autodesk.

Mike Haley, Senior Vice President of Research at Autodesk

The key difference is that neural CAD can accept a much broader range of inputs. You could sketch something and ask neural CAD to "make it 3D." You could describe what you want using a combination of written prompts, spoken language, drawings, sketches, and images. The neural CAD AI produces geometry based on all these inputs, whereas traditional parametric CAD requires exact specifications.

How Does Neural CAD Actually Work in Practice?

Autodesk has spent more than 15 years developing neural CAD, building entirely new forms of neural network architectures to handle the complexity of reasoning about 3D geometry. After years of foundational research, neural CAD is now evolving exponentially faster.

With neural CAD, professionals will be able to launch their ideas by speaking about them, typing descriptions, drawing sketches, or uploading reference images. The neural CAD engine then reasons through the request and produces highly detailed CAD objects and assemblies. Users can speak naturally with the AI to make changes as the technology reasons through the surfaces, edges, and topology needed to realize the idea. Rather than evaluating a single design option, professionals can evaluate many 3D design options simultaneously and instantly consider trade-offs.

Critically, neural CAD creates first-class, editable CAD geometry that users can immediately refine in parametric CAD software like Fusion or Forma. For specific tasks, it will even create the full history and sequence of CAD commands required to create the 3D model, making it easier to understand and edit.

Why This Matters for Design Professionals

Design and manufacturing professionals face mounting pressure. They are trying to solve increasingly difficult and complex problems while working with tighter budgets, compressed schedules, and slimmer margins. They must design for climate resilience, evolving regulations, mass urbanization, and changing sustainability goals, all while solving these challenges in less time using fewer resources.

Neural CAD addresses these challenges by raising the ceiling on what professionals can achieve. The technology promises higher precision, earlier performance insight, and fewer downstream errors. The opportunity is not just faster design, but better outcomes across industries.

"Give me digital clay," customers repeatedly tell Autodesk, expressing their desire for tools that feel more natural and intuitive.

Mike Haley, Senior Vice President of Research at Autodesk

Research from Autodesk shows that enabling exploration and decision-making during conceptual design, a neural CAD specialty, radically improves project metrics like cost, sustainability, and suitability. By creating high-quality CAD geometry early on, professionals will move seamlessly from conceptual design to detailed design development while preserving the original design intent.

How Neural CAD Differs From Other AI Approaches to 3D Design

Two general approaches exist for making AI systems that work with 3D CAD. The first approach uses large language models (LLMs), which are AI systems trained on vast amounts of text, to produce code that communicates with a parametric CAD software's existing application programming interface (API). This can automate specific forms or shapes with a traditional parametric CAD engine. However, these results remain problematic because an LLM was not trained to understand the three-dimensional physical world. It does not reason in 3D, so when requests become too complicated, the system shows its limitations.

Autodesk opted for the second, and far more difficult approach. Rather than using an LLM to speak to an API about geometry, neural CAD directly reasons in geometry. The large AI models powering neural CAD were trained from the beginning with "hands-on experience" in the form of real-world CAD objects. This is a fundamental difference that allows neural CAD to handle complex 3D reasoning that LLM-based approaches cannot reliably achieve.

Steps to Understand Neural CAD's Role in Your Design Workflow

• Conceptual Exploration: Neural CAD excels at helping designers explore multiple design options quickly during the conceptual phase, allowing rapid iteration and decision-making before committing to detailed specifications.
• Multi-Modal Input: The technology accepts spoken language, written text, sketches, drawings, and reference images, letting designers express ideas in whatever way feels most natural rather than learning rigid software commands.
• Seamless Refinement: Neural CAD produces editable CAD geometry that integrates directly with traditional parametric CAD software, allowing designers to move from conceptual design to detailed development without losing design intent or starting over.

What's the Relationship Between Neural CAD and Parametric CAD?

Autodesk does not view neural CAD and parametric CAD as competing technologies. Instead, the company believes the future of making anything will consist of a seamless combination of both approaches. Parametric CAD remains valuable for giving architects, artists, engineers, and builders precision and control. Neural CAD allows the free expression and intuitive interaction that parametric CAD lacks.

"We are moving to the future where your design tools understand your spoken language, your sketches, three-dimensional design data, the 3D physics of the natural world, and your organization's industry-specific workflows," Haley stated.

Mike Haley, Senior Vice President of Research at Autodesk

Autodesk is enhancing parametric CAD engines with neural CAD AI models, which the company describes as the first step function change in CAD technology in more than 40 years. Neural CAD features will power capabilities in Fusion and Forma that can reason effectively at the detailed geometry level, making professionals faster, smarter, and more competitive.

The shift represents a fundamental disruption of Autodesk's own tools to make them more accessible and user-friendly for creative people. Rather than requiring users to become software super users, neural CAD enables them to focus on being super designers, super engineers, and super artists, expressing their vision and exploring ideas faster, easier, and more naturally than ever before.