Where is this AI boom in AM?
RECODE.AM #34
Last week I had a very interesting conversation with Nicolas Mathian, Head of Marketing at Sculpteo, during which we talked about the AM market and its development prospects.
During that conversation - which I hope will be shared with you soon - two questions came up: AI in the context of 3D printing, and the current hype in the industry.
My answer was that while AI is gradually finding its place in 3D printing, with mixed results, I currently don’t see any real hype. And that’s despite the fact that exactly a year ago I was predicting that artificial intelligence would be precisely the next big hype wave.
Considering what happened first in the years 2010-2015 in the context of consumer 3D printing, and then in 2018-2022 in the context of industrial metal 3D printing, it seemed logical to forecast a similar flood of startups, aggressive funding rounds, and strong interest from venture capital funds in the area of AI.
Meanwhile, reality - at least for now - looks much calmer. AI in AM exists, it is developing, and it delivers real value, but we are still far from a mass investment frenzy.
It is difficult to even talk about an “active infiltration” of the 3D printing industry by investment funds.
Yes, a few interesting companies have emerged, such as Backflip AI, 3D Spark, Sparc3D, or Entina3D, as well as investments in more established players like AI Build.
These are, however, rather individual, selective bets than a broad, speculative trend. Compared to the period when every new desktop 3D printer manufacturer was potentially “the next MakerBot,” or when almost every metal AM startup could count on multi-million-dollar funding rounds, today’s landscape is surprisingly restrained.
One of the reasons is the maturity of the AM industry itself. 3D printing is no longer an exotic technology promising a revolution in every garage. It is an industrial tool, often deeply embedded in specific manufacturing processes.
AI, instead of creating entirely new mass markets, is increasingly playing the role of an “amplifier” for existing workflows.
It is also possible that, having already been burned twice, no one is particularly eager to step on the “3D” landmine for a third time.
Interestingly, companies are also starting to notice this trend (or dry lake of money).
The most interesting example of this shift is Backflip AI. The company started with tools that fit perfectly into the current hype around generative artificial intelligence: creating 3D models from text or images, the so-called Idea-to-Mesh approach.
The application was intriguing, especially for people outside the CAD world. A few photos taken with a smartphone were enough to generate a surprisingly good 3D model ready for printing within minutes.
This is exactly the kind of experience that is easy to sell on social media and that supports the narrative of the “democratization of design.”
However, the longer you look at Backflip, the clearer it becomes that the company’s true value lies elsewhere. Backflip seems to be moving from flashy model generation toward a much more down-to-earth, yet industry-critical area: Scan-to-CAD.
Instead of focusing on mass-market applications like 3D figurines from photos or gadgets generated from prompts, Backflip chose to tackle a real problem faced by engineers and manufacturing plants.
Scan-to-CAD is a tool that allows a 3D scan to be transformed into a parametric CAD model within seconds, for example in STEP format, or even directly into a native SOLIDWORKS part complete with a full feature tree.
This is a major qualitative leap. Traditionally, moving from a point cloud or mesh to an editable CAD model meant hours, sometimes days, of manual work. Now this process can be reduced to minutes, which in an industrial environment translates directly into money.
In this sense, Backflip shows where AI in AM has the greatest chances of success. Not necessarily in mass, consumer-oriented applications that are visually appealing but difficult to monetize and computationally expensive.
Generating thousands of figurine models or gadgets from text prompts may be impressive, but it is often economically questionable.
The situation looks very different in the case of narrow, highly specialized applications, where a single model can mean saving a production line worth millions of dollars.
And it is precisely this niche character that may explain the lack of a broad investment hype.
VC funds are accustomed to narratives about markets measured in hundreds of millions of users. Meanwhile, solutions such as Scan-to-CAD are aimed at a much smaller, but very specific audience.
Their value does not lie in the scale of users, but in the scale of the problems they solve. For the AM industry, this is good news, because it means the development of technologies that are truly needed, not just media-friendly.
Ultimately, the question is not whether an AI boom in 3D printing will come, but what form it will take.
Everything suggests that it will not be a repeat of the mass waves of enthusiasm seen in previous decades.
Instead, we are likely to see a slow, methodical adoption of highly specialized tools that quietly, but effectively, increase the efficiency of design and production.
This is an incredibly perceptive analysis of where 3D printing tech is actually headed. The shift from consumer hype to specialized industrial applications mirrors what I've seen in other mature tech sectors - the real value emerges when the noise dies down and engineers start solving actual production bottleneks. Its refreshing to see a company like Backflip pivot from flashy model generation to something like Scan-to-CAD that could genuinely save manufacturing plants from days of manual rework.