NVIDIA's partner network — including Advantech, FANUC, Solomon, Techman Robot, ITRI, Delta, Universal Robots, and Spingence — is showing physical AI and edge robotics at Automation Taipei 2026, held at Taipei Nangang Exhibition Center Halls 1 and 2. Jetson Thor, Jetson AGX Thor, Jetson Orin, Isaac Sim, Isaac Lab, and Omniverse are powering humanoid robots, AMRs, quadrupeds, and digital-twin factories across the show floor.
When and where is Automation Taipei 2026 held, and which NVIDIA partners are exhibiting?
Automation Taipei 2026, Taiwan's robotics and smart-automation trade show, is running at the Taipei Nangang Exhibition Center Halls 1 and 2, though the two available releases list slightly different dates. NVIDIA's official blog states the show runs from August 19 to 22 CITE:E1, while a separate release lists the dates as August 20 to 23 CITE:E2; both agree on the venue, Taipei Nangang Exhibition Center Halls 1 and 2 CITE:E1CITE:E2.
Exhibiting partners include Advantech, Antai International, ASUS, Delta, Kuang Yun Machinery Engineering, Litai Technology, Maison, MSI, Nexcobot, New Being Group, Pegatron, Siemens, Solomon, Spingence, Techman Robot, Choicecom, and Adlink Technology CITE:E12.
How do Jetson Thor, Jetson AGX Thor, and Jetson Orin power real-time inference for humanoid, collaborative, and quadruped robots?
Three tiers of NVIDIA's Jetson edge-AI chip family are each running a different class of on-robot inference on the show floor. NVIDIA describes Jetson Thor as a high-performance edge AI platform purpose-built for real-time robot inference, combining control, vision, and language models into a compact, energy-efficient system that acts as the robot's "brain" CITE:E11. Solomon integrates the NVIDIA Isaac GR00T 1.7 open robot foundation model, NVIDIA Cosmos, and NVIDIA NemoClaw on a Jetson AGX Thor compute platform for its humanoid robots and autonomous mobile robots CITE:E5. Techman Robot's TM Xplore I is driven by Jetson Thor for AI-computing-heavy tasks such as warehouse material handling CITE:E6. ITRI's bio-inspired quadruped robot is driven by Jetson Orin and trained through Isaac Sim and Isaac Lab to move across rugged terrain, sand, and varying load conditions CITE:E7.
| Partner | NVIDIA edge chip | Robot / application |
|---|
| Solomon | Jetson AGX Thor (with Isaac GR00T 1.7, Cosmos, NemoClaw) | Humanoid robots and AMRs with long-range perception CITE:E5 |
| Techman Robot | Jetson Thor | TM Xplore I for warehouse material handling CITE:E6 |
| ITRI | Jetson Orin (trained via Isaac Sim, Isaac Lab) | Bio-inspired quadruped robot CITE:E7 |
What are the real-world applications of humanoid robots, AMRs, and collaborative robots at the show?
Solomon, Techman Robot, and Universal Robots are each deploying edge-AI robots for distinct industrial tasks, from long-range humanoid perception to warehouse logistics and safety-critical sim-to-real transfer. Solomon is showing humanoid robots and autonomous mobile robots with what it describes as advanced long-range sensing capability CITE:E5. Techman Robot's Jetson Thor-driven TM Xplore I targets applications such as warehouse material handling CITE:E6. Universal Robots, working with Inventec (英業達) and 立普思, is using the real-time computing of NVIDIA IGX Thor to address the sim-to-real transfer challenge, meeting low-latency and safety requirements CITE:E9.
How do NVIDIA Omniverse, Isaac Sim, and Isaac Lab enable virtual factory modeling, AI training, and robot deployment?
FANUC, ITRI, and NVIDIA's own Omniverse-Isaac Sim pairing form the simulation pipeline that partners use to build virtual factories, generate training data, and validate robots before real-world deployment. FANUC uses NVIDIA Isaac Sim to build a virtual factory for simulation training, and uses Isaac Lab to support reinforcement-learning workflows for robot applications CITE:E4. ITRI trains its Jetson Orin-driven quadruped robot through the same Isaac Sim and Isaac Lab combination to learn movement across rugged terrain, sand, and different load conditions CITE:E7. NVIDIA Omniverse, combined with the Isaac Sim reference robot simulation framework, provides a physically accurate simulation environment that generates large-scale synthetic data for training AI models and for testing and optimizing robots before they are deployed in real environments CITE:E13. Universal Robots, Inventec, and 立普思 apply NVIDIA IGX Thor's real-time computing specifically to solve the sim-to-real transfer problem under low-latency and safety constraints CITE:E9.
How does Delta use NVIDIA technology for real-time 3D mapping, dynamic obstacle avoidance, and AI-based defect detection?
Delta is using NVIDIA Isaac ROS Nvblox and cuMotion to give its robots real-time 3D mapping and dynamic obstacle avoidance, plus NVIDIA PAIDF AnomalyGen to sharpen automated optical inspection. Through Isaac ROS Nvblox and cuMotion, Delta's robots can build 3D maps in real time and perform dynamic obstacle avoidance CITE:E8. Delta is also applying AI to factory-scale digital twins, and it uses NVIDIA PAIDF AnomalyGen to generate diverse defect imagery that improves the defect-recognition rate of automated optical inspection (AOI) CITE:E8.
How are NVIDIA partners building end-to-end smart manufacturing solutions, from industrial GPU servers to full-stack AI operations platforms?
Advantech and Spingence are each building end-to-end smart-manufacturing stacks that pair NVIDIA's industrial GPU hardware with full-stack AI operations software. Advantech is showing AI-driven smart manufacturing built on NVIDIA Omniverse, along with intelligent operations solutions using industrial GPU servers built on NVIDIA MGX and NVIDIA RTX PRO 6000 Blackwell GPUs CITE:E3. Spingence is showing its SPAK full-stack AI operations platform, built on the NVIDIA Factory Operations Agent Blueprint CITE:E10.
What this means
Across the eight partners covered in NVIDIA's own release, the same technology stack recurs at every layer: Jetson Thor, Jetson AGX Thor, and Jetson Orin for on-robot inference CITE:E5CITE:E6CITE:E7CITE:E11; Isaac Sim and Isaac Lab for training and virtual-factory simulation CITE:E4CITE:E7CITE:E13; and Omniverse-based software plus industrial GPU servers for factory-level operations CITE:E3CITE:E10. Delta's use of Isaac ROS Nvblox, cuMotion, and PAIDF AnomalyGen CITE:E8, and Universal Robots' use of IGX Thor for sim-to-real transfer CITE:E9, show the same underlying platform being applied to two distinct physical-AI problems — real-time navigation and simulation-to-reality accuracy. The one unresolved detail across the sourced material is the show's exact run dates, which differ by a day at each end between NVIDIA's blog and the other release CITE:E1CITE:E2.
Author's Take・林紀旭 James Lin
The pattern that stands out is standardization: humanoid robotics (Solomon), collaborative robots (Techman Robot, Universal Robots), factory simulation (FANUC), quadruped research (ITRI), and industrial computing (Advantech) are all converging on the same NVIDIA edge stack — Jetson Thor, Jetson AGX Thor, and Jetson Orin for on-robot inference, paired with Isaac Sim and Isaac Lab for training before deployment. Delta and Universal Robots are pushing that same stack into the two hardest physical-AI problems — real-time 3D mapping with dynamic obstacle avoidance, and sim-to-real transfer under low-latency constraints — using Isaac ROS Nvblox/cuMotion and IGX Thor respectively. None of the exhibitor claims include production yield, accuracy, or latency figures, so the metric worth watching next is whether any of these partners eventually publish measured performance data — AOI defect-detection rates, sim-to-real transfer success, or throughput — once these systems move past the show floor.