For one Illinois-based robotics company, the emerging era of physical AI has ignited customer interest like never before.
“We’ve been around for 22 years, but the last three to four years have been a perfect moment,” said Anish Shah, chief operating officer at Trossen Robotics based in Downer’s Grove, Ill., a Chicago suburb. “Physical AI is the hottest industry right now.”
Trossen makes robot arms, often called manipulators, used by the company’s 10,000 customers in all manner of industrial settings, from chefs to carmakers who need repetitive work done on a constant basis.
These customers often face a common problem: it is hard to keep workers at jobs that require so much repetition. Hence, a place for robots that are gaining more agility as each year goes by.
Trossen’s robot arms complete with optical sensors were demonstrated at a private Intel venue during Computex 2026 in Taipei. They were used to pick up short rectangular boxes with two-finger prongs and move each box to a tray molded to fit it. The job they were performing was an imitation of a how one unnamed European car company is using the same robots on a production line to move components that will be attached to a new car.
Intel’s role in the demo was to provide the brains for the robots’ work, relying on two new Intel Core Ultra Series 3 processors. Three engineers from Intel were able to train the robots to do the work over a single weekend by moving another pair of the same robots by hand, extending their arms and pinching together their two gripper prongs to pick up similar objects.
With the Intel chip and AI software from Scale AI, the maneuver was automated for other robots to be put work to work later on to repeat over and over, many times.
The Intel team’s training took only a weekend, a training process that previously might have taken weeks and cost hundreds of thousands, according to Trossen and Intel.
The robots being used in the demo came from Trossen’s WidowX line each costing about $4,500, a price that does not include the compute engine.
Shah said Trossen is globally the largest seller of robots in its segment, and has attracted education and research institutions among its customers. Smaller innovators and educators like the Trossen robots because they can be repaired with 3D printed parts, if needed. All the parts are manufactured in the US, a plus for many customers.
Trossen prides itself on bringing together all the components needed for robots to work, including hardware and software and compute. Intel has been a constant provider of processors used with Trossen’s robots. Intel says the Core Ultra Series 3 is built especially with robotics and industrial users in mind.
The processor combines CPU, GPU and NPU for low-latency edge AI with up to 16 cores and 180 TOPS. It is designed to deliver low-latency edge AI and multi-camera integration. Intel also includes Time Coordinated Computer to address jitter and latency spikes. TCC is designed to optimize hardware interfaces to ensure time-sensitive commands, such as immediate braking or steering instructions to an actuator in a car. The same concepts apply for more precise work that Trossen customers need.
Among its many customers, Trossen counts Apple and Meta and 60 of the Fortune 500. Also in that number is Chef Robotics, which sells robot systems to food companies that are desperate to find labor for food preparation.
Chef Robotics needed a robot that could move in a gesture like scooping instead of only chopping, which is where AI became valuable in training its robots, Shah said.