AI is creating work for robots that don’t even use AI, and I love what that says about where this industry could be heading!😀
I spend a lot of time talking about Physical AI, so naturally I’m excited about machines becoming more capable of understanding and acting in the world, but the latest robot-order numbers have me thinking about several things happening at once... and they don’t all require a breakthrough in intelligence. The Association for Advancing Automation’s Q3 2026 report shows that North American companies ordered 10,989 robots worth $723.5 million, with unit orders increasing 24.8% from a year earlier. The industry breakdown makes that growth more interesting, with year-over-year increases of:
97.6% in semiconductors, electronics, and photonics.
90.4% in automotive components.
67.7% in plastics and rubber.
38.7% in life sciences, pharmaceuticals, and biomedical.
Those are some serious increases!!! Food and consumer goods, meanwhile, fell 23.7%, which is a useful reminder that different industries are making different investment decisions. Percentage growth also doesn’t tell us which sector bought the most robots, and these figures represent orders rather than completed installations.
The First Growth Engine
What catches my attention is the connection between AI spending and the physical work required to support it. A3 points to semiconductor investment and electronics demand supporting AI among the factors accompanying growth in that sector, yet we often talk about AI as though it lives somewhere vaguely above us, in “the cloud.” Somebody still has to manufacture the chips and equipment that make it possible, which means the digital revolution requires quite a bit of physical stuff. (Everyone in manufacturing is probably wondering why this needs explaining.) That is the first growth engine I see: AI creates demand for things that established automation can help make, allowing a robot doing a familiar job to benefit from the investment cycle without becoming any smarter itself. I think we sometimes let the excitement about what’s coming distract us from the value of what’s already possible, even though manufacturers still need reliable output and have good reasons to invest in technology that already works.
The broader market supports that view. The International Federation of Robotics’ World Robotics 2026 release reports that factories installed more than 600,000 industrial robots in 2025, up 11%, bringing the global operating stock to five million. IFR also points to labor shortages, manufacturing relocation, and supply-chain resilience as demand drivers, which helps explain why the case for automation extends beyond enthusiasm for any particular AI technology.
The Second Growth Engine
The second growth engine gets me REALLY excited because it could change which work becomes practical to automate. Physical AI can help machines perceive their surroundings and respond to changing conditions, including the variation that makes so many industrial applications difficult. Picking a part that always arrives in exactly the same place is one challenge; handling a different part that has been turned sideways or partially obstructed adds requirements that the system has to accommodate. Learning and generalization could reduce how much task-specific engineering that requires, and there is research evidence behind the direction. In Google DeepMind’s 2023 RT-2 research, performance on previously unseen scenarios improved from RT-1’s 32% to 62%, with the work involving more than 6,000 robotic trials. I’m encouraged by that result, although I’d want to understand failure recovery and actual cycle times in an industrial application before treating a research benchmark as evidence that a system is ready for production.
I’ll admit that my own perspective has changed since joining Belden, particularly as I’ve realized how much I had taken infrastructure, networking, and data architecture for granted. Physical AI has made those dependencies harder for me to overlook, because even a highly capable robot needs an environment in which it can work reliably. I don’t want companies discovering those requirements AFTER they’ve invested in the exciting part, when the data, connectivity, security, and people needed to support the application should have been part of the discussion from the beginning.
That’s why I keep coming back to the whole system and the right level of autonomy for the application. More autonomy isn’t automatically a better investment, just as adding an AI label doesn’t answer the operational questions. A3’s report also doesn’t separate Physical AI adoption from conventional robotics, so I’m not attributing the order increase to Physical AI; my argument is that AI-related demand and improving robot capabilities could reinforce each other over time.
For the industry, that creates an opportunity across the integration work, vision systems, and software needed to make additional applications useful. IFR specifically notes that easier programming and integration are reducing deployment costs and opening automation to more applications and customers, and I keep thinking about the manufacturer who could automate a task but can’t justify the complete project. If deployment and changeovers become easier, that calculation may look very different... and THAT is how the market could get bigger.
I’d pay attention to companies making those projects easier for the people who actually run the equipment, with questions about affordability and day-to-day support sitting right alongside the robot demo. Can a smaller manufacturer afford the complete system, and what happens when something goes wrong on a shift without the specialist standing nearby? I’m excited about more robots, but I’m even more excited about the work that could become practical to automate and how many more businesses could participate.
References:
Association for Advancing Automation. (2026, October 28). North American robot orders increase 24.8% in third quarter of 2026. https://www.automate.org/robotics/news/north-american-robot-orders-increase-24-8-in-third-quarter-of-2026
Chebotar, Y., & Yu, T. (2023, July 28). RT-2: New model translates vision and language into action. Google DeepMind. https://deepmind.google/blog/rt-2-new-model-translates-vision-and-language-into-action/
International Federation of Robotics. (2026, September 24). Five million robots now operate in factories globally. https://ifr.org/ifr-press-releases/news/five-million-robots-now-operate-in-factories-globally
