On production bottlenecks in additive manufacturing and whether AI can solve them.
The funny thing about 3D printing is that it’s both faster and slower than its conventional counterparts. Need a single prototype or a couple hundred parts for bridge production? Sure, you could wait the 6-18 months it takes these days to get the necessary molds or tooling before you can actually start making those parts in earnest, or you could get them 3D printed and have your parts in hand in under a week. Conversely, if you need a couple hundred thousand or a million parts, it’s most likely faster (and almost certainly cheaper) to use injection molding or stamping, even with pre-production lead times.
So where does that leave additive manufacturing (AM)?
It’s an issue the AM industry has been struggling with for over a decade: trying to convince the broader manufacturing community that 3D printing is useful for more than just prototypes and short runs. Granted, there are examples of AM done right in production environments but it’s a tricky formula to get right and there’s a reason the poster child AM applications tend to be concentrated in aerospace components and medical devices.
In isolation, getting 3D printing up to speed for production (so to speak) is a challenge, but the need is arguably being made all the more pressing by the apparent productivity gains AI is contributing to other areas of engineering, including simulation, quality assurance, and root cause analysis. Could AI make 3D printing more productive and thereby accelerate additive manufacturing? Roman Arkhangelskiy, founder and managing partner of Upside Parts, says it can, and he’s all too familiar with the demand for shorter lead times in manufacturing.
The Amazon Effect in manufacturing
“Every week, we get a call from someone after hours,” Arkhangelskiy explains, “And the call is basically the following: ‘I’m coming right now and I’ll be there in 30 minutes. I’ll give you a 3D model and then I’m going to wait in your parking lot because I need this part in a few hours.’”
Arkhangelskiy refers to this as “The Amazon Effect” and it’s likely familiar to many contract manufacturers. As consumers, we’ve grown accustomed to being able to order something online and have it show up the very next day. To the consternation of many job shops, that expectation seems to have carried over into manufacturing, no doubt encouraged by services offering instant quoting on uploaded designs. Of course, as anyone who’s worked in manufacturing knows, there’s an enormous gulf between a part being quoted and its being manufactured. Instant quoting, alas, does not entail instant manufacturing.
According to Arkhangelskiy, it is possible to reduce the friction between when a part’s design is finalized and when it’s shipped to customers—and the good news is that it doesn’t involve ethically questionable labour practices. “The friction right now is not about one particular thing,” he says. “Printers are getting faster and better every year. Materials seem to be evolving every week. Really, it’s about combining the processes in a way that makes everything more efficient.” In the case of Upside Parts, that means using AI to onboard new projects, route them to the appropriate “paths of production” as well as monitoring production capacity and printer availability from moment to moment. The result isn’t quite instant manufacturing, but Arkhangelskiy claims it’s getting close:
“One month ago, we had a four-minute threshold. Now it’s two or three minutes. In two or three minutes from when you order, we start printing.”
The human factor in AI-driven manufacturing
Whenever the subject of AI-driven efficiency gains comes up these days, the obvious question to ask is what that means for the human element in the equation. Between the potential impacts on knowledge workers and concerns about humanoid robots displacing physical labor, it’s only natural to wonder whether faster AM necessarily implies fewer people. Predictions in this vein run the gamut from “Nothing will change” to “Everything will change” depending on whom you’re asking. For his part, Arkhangelskiy seems to lean toward the latter group.
“A year ago, I thought the era of professional engineers will last much longer,” he admits. “I was hoping we’d survive for a longer period of time, but I changed my mind.” That’s because he says he started to see more and more AI-generated CAD files being sent to Upside Parts with fewer and fewer mistakes in their designs. Between those CAD files and his company’s use of AI in manufacturing, “We’re moving towards fully robotized and AI-driven manufacturing cells, and that will start a discussion about how people can still maintain control over AI and what it will be producing in those cells.”
It’s a concern that’s been raised in various forms throughout the AI discourse, most recently in the news surrounding OpenAI’s claims about its agents “escaping containment” and the alarm bell rung by a former Anthropic employee. While these events could be described as the latest examples of the criti-hype that seems to infuse the AI industry, Arkhangelskiy’s reference to manufacturing cells at least makes these worries more tangible, if not more plausible. It also brings up another point about the future of additive manufacturing that parallels a frequent talking point about AI.
The China question
There’s a familiar line of argument amongst AI boosters that goes like this: Whatever the present or future risks of developing AI, they pale in comparison to the risks of the West falling behind China in its AI development. It’s a rationale with a more general form that goes at least as far back as the Cold War and it’s as intractable now as it was then—though perhaps less persuasive. What’s interesting about it in this context is the overlap with the additive manufacturing competition that’s emerged from China over the past decade.
In fact, according to Arkhangelskiy, competing with China was one of the original drivers for Upside Parts. “Two years ago, our customers were telling us that China was shipping parts in three weeks, even with the delivery barriers,” he explains. “We wanted to be faster than China, so our initial target was for us to be better than three weeks.” In Arkhangelskiy’s telling, that turnaround time has only gotten shorter.
So, since the trade barriers have only gotten higher, does this mean we can reshore manufacturing via companies like Upside Parts? Arkhangelskiy is hopeful but he also recognizes the limitations of his perspective, as well as the fact that domestic manufacturing capabilities aren’t gained or lost in a day, or even a year. “Do we need another 40 years to bring it back, or can we make it in five or 10 years?” he asks rhetorically. “I’m not the one to answer that.”
What he is confident about is the necessity of AI: “I’m not sure how to shift this gap [between China and the United States] but I know that if we don’t start to implement more AI and robotics in our manufacturing, this gap will only increase.”
Can AI resolve the speed paradox?
These days there’s a tendency to see AI as either the solution to the majority of our problems or as the cause of them. Arkhangelskiy appears inclined much more toward the former view than the latter. While he acknowledges the advantages of advancements in additive materials and the potential production efficiency gains from hybrid additive and subtractive machines, it’s clear that AI and the level of automation it enables is at the heart of his vision for the future of his company.
“We see people breaking limitations with cross-material and cross-technological machines,” he says, “and hopefully they will end up in our robotized facility, which will be doing everything at Amazon-like speed.” It’s this combination that Arkhangelskiy believes will enable him to satisfy those urgent, after-hours orders. “Just wait five more years,” he predicts, “and you’ll get it.”
