iREX 2025 was held in Tokyo from Dec. 3 to six. Credit score: Georg Stieler
The Worldwide Robotic Exhibition, or iREX 2025, earlier this month set new information: 673 exhibitors and 156,110 guests — even with components of the Tokyo Large Sight exhibition floor below renovation.
The temper on the occasion was good, as market order consumption is recovering after two years of decline (notably pushed by exports). What did I see there that I haven’t seen someplace else earlier than?
1. Sensible AI: From lab to manufacturing facility ground
Whereas there have been loads of dancing humanoid robots, primarily of Chinese language origin, what caught my eye had been AI-enabled industrial robots to deal with the slight variations and complexities beforehand reserved for human dexterity.
For instance, Yaskawa confirmed the MOTOMAN NEXT-NHC 10DE, an autonomous dual-arm robotic that packs objects right into a field with human-like delicacy.
In line with Yaskawa, the robotic’s motions had been discovered by imitation of a human demonstration. Engineers first had an individual put on motion-capture markers on their arms and recorded the individual rigorously packing a field on digicam. Utilizing this captured knowledge, the NEXT-NHC 10DE replicated the human’s packing motions.
Yaskawa’s iREX exhibit emphasised AI controls (developed with NVIDIA) for adaptive, exact dual-arm coordination. This implies the robotic wasn’t simply pre-programmed; it used AI algorithms working on a high-performance GPU to carry out real-time changes, akin to visible inspection of merchandise placement and delicate dealing with.
The MOTOMAN NEXT-NEX35 is able to unloading, unpacking, and feeding variable uncooked supplies. This method has an autonomous management unit and fuses AI judgment with robotics, enabling real-time recognition and resolution making as environment change.
We had suggested a beverage producer with related wants about its automation technique six years in the past, however they had been inconceivable to fulfill at the moment.
2. Platform play: Opening up the walled gardens
Carefully associated to AI was FANUC’s stunning pivot to extra openness. Traditionally recognized for closed, proprietary robotic controllers, Japan’s largest robotic producer introduced a strategic partnership with NVIDIA to speed up bodily AI in industrial robotics simply earlier than iREX.
FANUC launched a ROS 2 driver on GitHub, permitting builders to regulate its robots through the open-source Robotic Working System. The corporate additionally launched official Python and ROS 2 help on its newest controllers.
This open-platform strategy is supposed to “improve connectivity with exterior AI environments” and pace up implementation of bodily AI – basically connecting robotic arms with fashionable AI software program. FANUC’s ROS 2 driver helps 1-millisecond management loops and works throughout fashions, from small cobots to its 2.3-ton heavy robotic.
The driving force nonetheless is dependent upon particular FANUC controller software program choices (J519/R912 or S636), so this openness is targeted on developer workflows, whereas key controller capabilities stay licensed. Moreover, FANUC built-in NVIDIA’s Isaac Sim simulation platform, contributing official 3D fashions of its robots for digital testing and AI coaching.
At iREX, FANUC demonstrated a AI-powered dual-arm cable set up system with Japanese AI startup Fingervision, in addition to a collaboration with Inbolt from France for monitoring and dealing on shifting components (see video under). It additionally confirmed a robotic responding to and executing pure language voice instructions.
It’s fascinating to see that functions that had been experimental 5 years in the past are actually so steady that they’re turning into mainstream.
The development towards extra openness was additionally mirrored by Kawasaki’s introduction of its “open” controller idea. A single controller synchronously managed two force- and power-limited or collaborative robots from Kawasaki, an electrical gripper from Oriental Motor, and the XPlanar magnetic floating linear conveyor system from Beckhoff Automation.
Kawasaki additional confirmed the advances of its ROBO CROSS cloud platform, the place finish customers can obtain functions and instruments developed by system integrators as modules. Taking part corporations up to now embrace Kawasaki Heavy Industries, NTT Enterprise Options, Daihen, FingerVision, Yasukawa Electrical, Yamaha Motor, and Japanese startup ugo.
Whereas these strikes are essential to speed up innovation safety, these methods additionally pose new challenges. Safety, stability, and AI focus are dangers as superior AI entry is more and more turning into depending on a small set of suppliers — notably NVIDIA — with corresponding bargaining-power implications.
The successful robotics corporations most likely gained’t be essentially the most “open” or “closed” however the ones that outline a transparent differentiation layer: software libraries, high-quality job knowledge, validated security, and lifecycle help.
3. Chinese language gamers proceed worldwide enlargement at iREX
Chinese language companies turned out in pressure, with 84 exhibitors (up from 50 in 2023) making up the vast majority of abroad corporations at iREX 2025. Such high-profile occasions are an necessary platform for them to current themselves not solely to a Japanese, but in addition to a global viewers.
Whereas we haven’t but seen Chinese language industrial robots making inroads in Japan, they’ve turn into formidable opponents in different components of the world. Again in 2016, Japan managed roughly a 3rd of world robotic exports. By 2024, its share had slipped to round 21%, whereas China’s share climbed from about 3% to 4% to just about 10%.
At iREX, essentially the most seen Chinese language exhibitors got here from segments the place Japanese gamers are much less seen or much less aggressive on value – low-cost cobots, humanoid robots, and superior tactile sensing. Their methods differ. As an illustration, Dobot has began to rent extra native staffers, which looks as if the proper strategy, as this allows it to poach extra high expertise in native markets such because the U.S. and Japan.
Bodily AI would require quite a lot of real-world knowledge. A rising share of the {hardware} used to gather that knowledge – particularly low-cost sensing and humanoids – is coming from China. PaXini confirmed its PMEC Hyper Assortment System, consisting of a digicam and gloves with multidimensional tactile sensors.
AgiBot, one in all China’s main humanoid robotic makers, used iREX to announce its entry into the Japanese market and confirmed its vision-language-action (VLA) mannequin ViLLA.
4. European deep-tech startups scale via globalization
It was encouraging to see European deep-tech startups presenting themselves in Tokyo. The aforementioned Inbolt confirmed functions on the cubicles of each FANUC and Common Robots (UR).
Swiss startup AICA confirmed an AI-powered gearbox meeting with associate UR. Regardless that they’re nonetheless younger, the trail to scale for these corporations is thru world partnerships with main producers within the U.S., Japan, and South Korea.
5. iREX 2025 provides a glimpse into the longer term
One in every of my favourite reveals at iREX got here from SOLOMON, a machine imaginative and prescient specialist from Taiwan.
It tuned a Unitree G1 humanoid with enhanced, extra succesful arms and an onboard NVIDIA Jetson AGX pc. Utilizing NVIDIA’s GR00T platform, the robotic was skilled to see as much as 5 m (16.4 ft.), perceive orders in pure language, and plan the bodily steps to select from an outlined set of objects.
The applying was nonetheless shaky and did generally have hassle greedy the objects on the first try, nevertheless it supplied a glimpse into the longer term. A ChatGPT-style second for embodied intelligence — when robots will be capable of generalize natural-language directions to new, non-preprogrammed duties — appears doable within the not-too-distant future.
In regards to the writer
Credit score: Georg Stieler
Georg Stieler advises a few of the world’s largest robotics corporations. He spent greater than 10 years dwelling in China and now splits his time between Switzerland and the Folks’s Republic of China.
Due to multi‑yr, shut collaboration with AI startups in Silicon Valley, Stieler can also be deeply conversant in its tradition.
