Threedy’s 10-Year Bet: Rethinking Industrial 3D Data from First Principles

Most B2B software companies optimize existing approaches. They make things faster, cheaper, or easier to use. They take what works and make it work better.

A decade ago, Christian Stein and his team at Fraunhofer research took a different approach: they decided to fundamentally rethink how 3D data should work in a connected world.

That question—what if we stopped treating 3D data like files and started treating it like streaming information—became Threedy, an industrial 3D platform that’s now raised over $11 million and serves major automotive manufacturers.

In a recent episode of Category Visionaries, Christian explained the technical insight that made it all possible, and why first-principles thinking in research environments can create competitive advantages that last for years.

The Two Traditional Approaches (And Why They Both Have Limits)

To understand Threedy’s innovation, you need to understand what they’re replacing. Christian explains that industrial 3D data has historically been handled in two ways, both with significant limitations.

The first approach: “people build software that kind of generates images on a server side and then streams images to the client, abstracting about data size distribution, et cetera, on that way.”

This server-side rendering approach solves the distribution problem—clients don’t need to handle massive 3D files. But it sacrifices interactivity. Users are essentially watching a video stream, not manipulating actual 3D data. For industrial applications where engineers, designers, and technicians need to interact with models, this creates fundamental constraints.

The second approach: “three, data is utilized by exporting it from complex management systems, by converting it, by reducing it to match the capabilities of an application of an end user device.”

This client-side approach preserves interactivity but creates nightmares around data management, version control, and distribution. You’re constantly converting files, reducing fidelity to make them usable, and managing dozens of derivative versions for different applications and devices.

Both approaches made sense when they were designed. But both were built on assumptions about how data needs to work—assumptions that Christian’s team decided to question.

The First-Principles Question

The breakthrough came from asking a deceptively simple question: what if we designed for the world we’re moving into rather than the world these systems were built for?

“Here’s exactly where we differ, that what we did roughly 10 years ago is starting to think about how we think that data should be utilized in a modern software world where data is typically progressively streamed based on user demands, based on interactions in the application,” Christian explains.

This is first-principles thinking in action. Instead of asking “how do we make the existing approaches better,” they asked “if we were designing this today, knowing where technology is heading, what would we build?”

The insight: consumers expect data to stream progressively based on what they’re actually doing. Netflix doesn’t download the entire movie before you start watching. Spotify doesn’t require you to store every song locally. Google Maps doesn’t force you to download city data before you can navigate.

Why should industrial 3D data work differently?

The Always-Connected Paradigm Shift

The key assumption that enabled Threedy’s approach: we can design for an always-connected world.

“Obviously Threedy is a lot different than your standard 2D information that we consume every day. So our approach was to kind of rethink how standard algorithms are working and if we do the paradigm shift to think that we’re always connected, we can distribute workload between client and server, then we can kind of redesign standard algorithms to work in an always connected world a little bit better than these approaches.”

This paradigm shift—assuming constant connectivity and distributed workload—allows for fundamentally different algorithm design. Instead of choosing between “render everything server-side” or “download everything to the client,” you can intelligently split the work based on what users are actually doing.

Looking at a low-detail overview? Stream minimal data. Zooming into a specific component? Progressively load higher-resolution detail. Rotating the model? Predict what data you’ll need next and preload it.

This approach combines the benefits of both traditional methods: interactivity of client-side rendering with the scalability of server-side processing.

Why This Required Research Time

Building this required more than good engineering—it required reimagining how fundamental algorithms work. You can’t just modify existing 3D rendering engines to work this way. You need to redesign from the ground up.

This is why the work happened at Fraunhofer, a research organization, over many years before becoming a company. “I have a history as a computer scientist working a lot with computer graphics software engineering, did my first steps in the games industry and then by chance decided to master and joined the Fraunhofer research organization,” Christian shares.

Research environments provide something that startups rarely have: time to work on hard technical problems without immediate commercial pressure. The team could spend years developing the core technology, validating the approach, and refining the algorithms before worrying about product-market fit or revenue models.

By the time they were ready to spin out, roughly a decade of research had gone into the core technology. “What we did roughly 10 years ago is starting to think about how we think that data should be utilized,” Christian notes.

That decade of research created a moat. Competitors can’t easily replicate this approach because it’s not about adding features to existing systems—it requires fundamentally different technical architecture.

The Market Shift That Made It Matter

Even brilliant technology needs the right market conditions to matter commercially. For Threedy, that shift came from how industrial companies started using 3D data.

“While in the past, maybe utilization of complex heavyweight industrial data was mostly all about engineering and design, today there’s a rising demand of utilizing that data all across the organization and downstream along, let’s say, the whole product life cycle of the physical products that are manufactured,” Christian explains.

This democratization of 3D data created the market opportunity. When only engineers needed to interact with CAD models, traditional approaches worked fine. Those users had powerful workstations, specialized software training, and patience for complex workflows.

But when sales teams need to configure products, service technicians need to visualize assemblies, and manufacturing operations need to optimize production—suddenly you need 3D data that works like consumer applications: fast, intuitive, accessible from any device.

This is where Threedy’s architecture becomes commercially valuable, not just technically interesting.

What “Visual Computing Infrastructure” Actually Means

Christian describes Threedy as “visual computing infrastructure” and “middleware”—deliberately positioning it as foundational technology rather than end-user application.

The company enables “the development of fast and scalable industrial 3D and mixed reality applications.” Notice: they’re not building the applications themselves. They’re building the infrastructure that makes those applications possible.

This positioning is strategic. Rather than competing with CAD systems, visualization tools, or mixed reality applications, Threedy becomes the layer that makes all of them work better. It’s infrastructure that other software can build on.

“I think what’s totally unique about us is how we build our core technology,” Christian emphasizes. The competitive advantage isn’t features—it’s fundamental architecture.

The Long-Term Vision: From Files to Data Spaces

The decade-old technical bet connects to an even bigger vision about how industrial software should evolve.

“One of our core concepts of putting a data space into the center and moving away from files to APIs on shared spaces is something that is going to be very valuable in building these industrial applications of the future,” Christian explains.

This is the natural endpoint of their first-principles thinking. If you’re streaming 3D data progressively based on user needs, you’re already treating it more like a data service than file storage. The next step is moving the entire industrial software ecosystem away from file-based workflows toward API-driven data spaces.

This vision positions Threedy not as a product company but as infrastructure for how industrial companies will manage spatial data as 3D, mixed reality, and AI become ubiquitous.

The Research-to-Startup Advantage

Threedy’s origin story reveals something important about technical competitive advantages. The company had paying customers “roughly four years before we founded the company,” Christian notes. Those customers were buying decade-old research—and willing to pay for it before it was even a commercial product.

That’s the power of solving hard technical problems that create genuine differentiation. You can’t quickly replicate a decade of research. You can’t easily copy fundamental algorithmic innovations. And you can’t fake the kind of deep expertise that comes from years of focused work on a specific technical challenge.

For founders trying to build defensible technology companies, Threedy’s story offers a template: find a fundamental assumption that everyone accepts, question whether it still makes sense given where technology is heading, and be willing to spend years building the right solution rather than the quick solution.

Ten years ago, Christian and his team asked: what if we stopped thinking about 3D data as files? That question became a research project, then paying customers, then a spinout company, and eventually an $11.5 million platform serving major industrial companies.

Sometimes the longest bets create the strongest moats.