
▲ ST's humanoid robot Kai
Composed of over 500 ST components including MCUs, sensors, motor drivers, and power management ICs
All computational edge processing enables real-time AI inference without external servers or additional devices.
ST’s humanoid robot 'Kai,' which organically combines hardware and AI, control and sensing, and safety and connectivity, is ST’s powerful message toward the next stage of industrial automation.
In the middle of the exhibition hall at the Smart Factory & Automation Industry Exhibition (SF+AW 2026), which opened at COEX in Seoul on March 4, there was a booth that was constantly bustling with visitors.
The star of the show is 'Kai,' a humanoid robot demo unveiled by STMicroelectronics (hereinafter ST).
It was an upper-body demo unveiled for the first time at a domestic exhibition since its debut at CES 2026, and it was an iconic display that showed the future of industrial automation and robotics at a glance.
Kai is more than just a simple robot demonstration; it is the result of integrating ST's semiconductor technology and embedded AI capabilities.
Composed of over 500 core ST components, including MCUs, sensors, motor drivers, and power management ICs, this demo clearly conveys the message of a 'complete integration of hardware and AI software.'r />
An official from the ST booth explained, “Kai is not a specific finished product, but a technology platform and reference for realizing humanoid robots.”

▲ ST's humanoid robot Kai is displaying natural movements in response to a gesture from an ST official.
The most eye-catching part of the booth was Kai's intelligent motion recognition demonstration.
Kai recognizes human movement in real time through a camera and precisely identifies movements by analyzing landmark data from 17 joints of the human body.
This information leads directly to the control of the robot's dual arms, enabling natural movements that respond to human gestures.
It is impressive that all computations are processed at the edge, enabling real-time AI inference without external servers or additional computing devices.
At the heart of this performance is ST's next-generation MCU, the STM32N657.
This chip, operating at up to 1 GHz, is equipped with a 600 Gops-class Neural-ART AI accelerator to simultaneously process high-resolution vision data and complex pose recognition calculations.

▲ Functional blocks of ST's humanoid robot Kai
The ST engineer I met on-site was &ldq“For humanoid robots, latency is directly linked to safety,” he emphasized, adding that “achieving this level of performance in edge AI is a key competitive advantage.”
The control system supporting Kai's stable operation is also noteworthy.
The STM32MP257 processor, based on 64-bit Cortex-A7 and Cortex-M33 dual cores, performs sensing data processing and motor control in parallel.
This ensured precise motion control and real-time responsiveness of the dual arm simultaneously.
In addition, by applying a Linux OS-based environment and CodeSys runtime optimizations, it also satisfies the real-time performance and scalability required in industrial settings.
Connectivity and safety are also essential elements for humanoid robots to be deployed in actual industrial sites.
Kai supports three Gigabit Ethernet ports and CAN and RS-485 interfaces, taking into consideration interoperability with existing industrial equipment and control systems.
In addition, it emphasized that reliability has been secured even in environments collaborating with humans by applying functional safety certified MCUs and security solutions.
Through this exhibition, ST also presented the challenges facing humanoid robots.
Price competitiveness, energy efficiency, and safety and cybersecurity compliance are factors that must be addressed for commercialization.
ST's strategy is to address these challenges step by step through highly integrated semiconductors, power management technology, and integrated sensing solutions.
An automation industry official who toured the booth said, “Rather than a robot for immediate mass production, Kai is a blueprint showing the direction in which industrial humanoids will evolve,” adding, “It clearly demonstrated the role semiconductor companies can play in the robotics ecosystem.”evaluated it as uo;