I was listening to Andrew Anagnost’s keynote at AU 2025 and caught myself thinking about the next step in the transformation led by Autodesk.
The setting in Nashville was impressive, but what stayed with me wasn’t the stage production—it was the continuity of Autodesk’s long-term strategy. For me, it echoed back to AU 2011, when Carl Bass stood on stage and described a vision of moving away from documents, files, and silos, toward a future where data would be connected, searchable, and semantically meaningful – check my old blog – Carl Bass Autodesk Vision.
At the time, Autodesk went further and introduced a cloud PLM strategy. I wrote about it back then in Autodesk Enters PLM – Everything Changes for Autodesk. The industry reaction was skeptical, even dismissive: “Manufacturing companies will never put their data in the cloud.” Yet Autodesk pushed ahead, and the following years proved that the shift to cloud was not only possible but inevitable. I’m writing a separate Beyond PLM article about Autodesk PLM Summit, so stay tuned.
From Cloud Vision to Established Platform
Sitting in the audience in 2025, I couldn’t help noticing that the bold ideas from 2011 are now simply infrastructure. Autodesk Platform Services (APS), data models, and industry clouds like Fusion, Forma, and Flow are no longer “strategy slides.” They exist, they scale, and they form the baseline of Autodesk’s business.
In other words, the question of whether cloud can work for engineering and manufacturing has been answered. Companies are using it daily. Data lives in APS, APIs connect services, and workflows extend across organizational boundaries.
Autodesk’s cloud gamble is now the industry’s working reality.
Adding a New Layer – AI Native Autodesk
Andrew Anagnost framed Autodesk’s next step as adding another layer to the familiar stack. The foundation remains APS. On top of that are the industry data models, and above them, the domain-specific applications. What’s new is the addition of AI as a layer that cuts across all of it.
Autodesk’s updated description—cloud-based, AI-native, end-to-end—is less about marketing language and more about positioning. It signals that AI is not a plug-in feature or an isolated assistant. Instead, Autodesk intends to make AI part of the platform fabric, accessible to every application and every workflow that runs on top of APS.
During the keynote, Andrew Anagnost presented real examples of Autodesk applications powered by AI assistants—for instance, generating geometry from a text prompt, creating a Microsoft PowerPoint presentation for a design review, and several others.
A Timeline of Transformation
In a dedicated keynote, CTO Raji Arasu put this shift in historical context. She started with the Manchester Baby in 1948, the first stored-program computer, then moved forward through the rise of APIs, the adoption of cloud computing, and now the acceleration of AI. The next stage in Autodesk’s roadmap is what they call MCP Servers, with private betas planned for later this year.
The MCP Servers are designed to provide AI-native services in a standardized way—search, reporting, analysis—embedded directly into engineering and design processes. Instead of one-off AI integrations, Autodesk is aiming for a consistent architecture that can grow into a foundation, much like APS did for cloud.
Arasu described the company’s AI priorities as threefold: automate routine tasks, connect currently disconnected processes, and provide real-time insight. While simple in phrasing, each of these implies significant technical depth. Automation requires access to reliable, structured data. Connection means interoperability across tools and organizations. Real-time insight demands scalable data infrastructure and models trained on relevant domain knowledge.
Toward Neural Geometry
The longer-term picture came from Mike Haley, who leads Autodesk’s AI Lab. His focus was on what he called neural geometry. The idea is to develop foundational AI models for design, similar in concept to large language models, but applied to geometry.
This is not about a single feature like generative design—it’s about rethinking how CAD itself can function when AI understands geometry in a more fundamental way. Haley described a future where AI models could propose design alternatives, detect structural issues, or interpret intent directly from geometric context.
The ambition is to move CAD from a command-driven environment into a more collaborative interaction between human engineers and AI models. That’s a profound change, one that requires not only advanced algorithms but also enormous, high-quality datasets.
Why This Strategy Feels Grounded
It is easy to dismiss AI announcements as hype. What made Autodesk’s messaging more credible to me was the way it ties back to the platform architecture they’ve been building for more than a decade.
APS provides the necessary data infrastructure. Unlike traditional PLM systems, which often create silos, APS is inherently designed to connect. MCP Servers give AI a standardized entry point into workflows, rather than scattered experimental features. And Autodesk’s dual presence in both AEC and manufacturing creates opportunities to build multi-disciplinary applications, something competitors with narrower focus may struggle to match.
There is also a matter of scale. Autodesk has the financial and market position to pursue long-term development of foundational AI models for engineering—something that requires both investment and access to domain expertise.
Parallels with the 2011 Cloud Shift
The similarities with 2011 are hard to ignore. Back then, Autodesk bet on cloud when many doubted it. Today, Autodesk is betting on AI at platform scale while skepticism still lingers about whether AI can be trusted in engineering workflows.
History suggests that the doubts will fade, and what seems risky today will become ordinary infrastructure tomorrow. Just as cloud went from “impossible” to “indispensable,” AI in design and manufacturing is likely on the same trajectory.
Opportunities and Open Questions
The opportunities are clear: AI can reduce manual effort, accelerate design, and connect fragmented processes. But the open questions are equally important.
How will Autodesk ensure reliability and trust in AI-generated outputs? How will MCP Servers interoperate with tools outside the Autodesk ecosystem? And how will AI workflows adapt to regulatory, safety, and compliance constraints in industries like aerospace, medical devices, or construction?
These are not minor challenges, and Autodesk will need to address them as the platform evolves.
Nashville Music City Fun
To me, AU is strongly associated with the Las Vegas Venetian Hotel. But Nashville Music City has its own charm.
What is my conclusion?
For me, AU 2025 was less about flashy demos and more about continuity of vision. The same company that once argued for the cloud in an industry that didn’t believe in it is now making the case for AI-native design and manufacturing.
Autodesk is positioning AI not as an add-on, but as an architectural layer. That distinction matters. If the approach works, AI will become part of the infrastructure of engineering software, just as the cloud did over the last decade.
Whether Autodesk succeeds will depend on execution, adoption, and the ability to solve practical challenges. But the direction is clear, and the momentum is real.
Just my thoughts…
Best, Oleg
Disclaimer: I’m the co-founder and CEO of OpenBOM, a digital-thread platform providing cloud-native collaborative and integration services between engineering tools including PDM, PLM, and ERP capabilities. With extensive experience in federated CAD-PDM and PLM architecture, I advocate for agile, open product models and cloud technologies in manufacturing. My opinion can be unintentionally biased.
