As AI accelerates its transition from the cloud to the physical world, smart glasses, humanoid robots, smart vehicles, and high-density data centers are reshaping the underlying architecture of the electronics industry. This transformation poses extreme challenges for hardware design, demanding high power density, low power consumption, miniaturization, and high precision.
At electronica China 2026, TDK Corporation showcased its core technical data and hardware solutions in edge computing, automotive-grade sensing, and new energy power supply, providing a clear technical pathway to address current industry pain points.
Energy Consumption and Practicality Challenges of Smart Glasses
Smart glasses are widely recognized as the most natural hardware entry point in the AI era, but their technical bottleneck lies in achieving all-day contextual awareness under extremely limited battery capacity. Traditional vision solutions suffer from high power consumption due to continuous video recording and the operation of general-purpose CPUs or NPUs.
To address this, TDK adopted a different hardware logic and introduced the AIsight visual perception solution. This architecture utilizes a proprietary customized AI DSP chip positioned between an ASIC and an NPU, specifically optimizing power consumption at the hardware level for machine learning and object detection algorithms.
Figure | TDK smart glasses equipped with AI technology; Source: TDK
In terms of camera configuration, the system consists of one front-facing low-power camera for contextual perception and two eye-tracking cameras for detecting gaze points (ROI intent areas).
According to TDK engineers, the overall power consumption of this vision subsystem during model operation is only 5-10mW. The system does not record video normally; it only performs object detection at the edge, capable of simultaneously recognizing 20 specific targets such as faces and QR codes. At this time, the main processor and cloud Agent remain in sleep mode and are automatically awakened only when the AI DSP captures key useful information, thereby achieving 24/7 monitoring.
Beyond vision, "false wake-ups" in voice interaction have long been a pain point for smart glasses and TWS (True Wireless Stereo) earphones. TDK introduced Voice Vibration Detection (VVD) technology in its SmartMotion solution. Without adding any hardware such as microphones, this technology directly utilizes the accelerometer in the 6-axis motion sensor to detect the bone conduction vibrations caused when the wearer speaks.
Figure | SmartMotion for TDK smart glasses; Source: TDK
TDK engineers pointed out that since environmental noise or others speaking cannot cause the wearer's bone conduction vibrations, this technology can completely eliminate false triggers. It has been mass-produced and applied in smart earphones, enabling the function of automatically lowering the music volume as soon as the wearer starts speaking.
Vibration Suppression for Embodied Robots and Industrial AI
Humanoid robots and industrial equipment maintenance face similar physical challenges: high-frequency vibrations from motors can severely interfere with sensor outputs, and traditional large models are highly prone to "hallucinations" in harsh industrial scenarios, leading to decision-making failures.
The head, torso, and feet of humanoid robots need to track posture in real time for balance adjustment, acting as the "inner ear" of the machine. The 6-axis IMU specifically designed by TDK for this purpose adopts a MEMS structural design of BalancedGyro, which can effectively suppress linear vibrations generated by motors at the sensor source, preventing zero-bias shifts.
TDK engineers revealed that the temperature stability (temperature drift performance) of this 6-axis IMU is 1 to 2 times better than that of industry competitors, significantly reducing the long-term integration error of the gyroscope.
In scenarios where vision fails (such as transparent glass, dark night lighting, and high-vibration environments), ultrasonic sensors become a crucial safety redundancy. The ultrasonic sensor module launched by TDK embeds the AI training model directly inside the module. It can automatically identify grass, concrete, or stairs based on the behavioral patterns of sound wave reflections, assisting embodied robots in anti-slip preprocessing or fall prevention.
Figure | TDK humanoid robot solution & ultrasonic sensor
According to TDK engineers, this module also supports invisible integration (under development), allowing it to be hidden behind the enclosure. This eliminates the cost for customers to drill holes or apply color-matched painting on the device surface, and it can stably handle complex industrial conditions. Currently, TDK is developing technology with ASIC suppliers to achieve 5cm blind spot detection with a single module, which will reduce the sensor integration cost at the system level by nearly half in the future.
For the predictive maintenance of industrial equipment, TDK SensEI adopts a converged architecture of "real underlying physical mechanisms + ISO industrial standards + professional vibration analyst know-how". This edgeRX solution collects data through front-end wireless temperature and vibration dual-mode sensors, employs low-latency edge intelligence processing at the edge computing gateway, and builds lightweight, industry-equipment-specific small models on localized servers. Data and intelligent computing can be fully deployed and completed locally, ensuring data never leaves the factory.
Figure | TDK edgeRX solution; Source: TDK
It is reported that this solution has been deployed and validated first at TDK production bases and has been promoted to multiple industries such as precision manufacturing and ports, successfully solving the pain point of the difficult implementation of traditional large models in industrial scenarios. In addition to edgeRX for equipment operation and maintenance, TDK has simultaneously launched the edgeRX vision visual inspection solution for product quality control on the production side, forming a dual-dimensional industrial AI empowerment system of "equipment health + product quality".
Power Supply Revolution for AI Data Centers and Blackwell Platforms
The proliferation of new-generation AI server platforms such as NVIDIA Blackwell has brought severe extreme dynamic power perturbation (EDPP) challenges to data centers. The power supply must maximize energy storage at the busbar end to smooth out instantaneous load fluctuations and protect the external power grid.
The "etched foil" process used in traditional aluminum electrolytic capacitors—which utilizes chemical etching to create pores on pure aluminum to increase surface area—has approached its physical limits after decades of development. TDK has now introduced an innovative "laminated foil capacitor" technology, switching to an ultra-thin aluminum substrate where aluminum powder is uniformly deposited and sintered to form. The microstructure changes from the original pore-like shape to spherical particles, significantly increasing the effective surface area of the foil. Under the exact same volume, the capacitance of the laminated foil capacitor is directly increased by 20%-30%.
Figure | Aluminum electrolytic capacitors for AI server power supplies; Source: TDK
TDK engineers revealed that, coupled with this material innovation, TDK has jointly developed a 12kW high-power-density high-frequency AI server power supply unit (PSU) with partners. Meanwhile, to adapt to new topologies such as the capacitor bank unit (CBU) for future Sidecar configurations and optimize space utilization, TDK is expanding the capacitor length from 100mm to 120mm and even 140mm. It is also widely adopting chip hybrid capacitors and MLCCs (Multi-Layer Ceramic Capacitors) at the output end to replace traditional capacitors, fully supporting the high-density power supply needs of data centers "from grid to core".
Process Evolution in Automotive Electronics and Miniaturized Components
In the new energy vehicle (NEV) sector, competition has deepened into the efficiency of the three-electric system and extreme safety protection. In the high-voltage distribution box, when the vehicle encounters overcurrent or overvoltage, ordinary disconnection methods struggle to quickly extinguish high-voltage arcs. TDK's high-voltage direct current (HVDC) contactors (HVC) are specifically designed for the disconnection and arc extinguishing protection of high-voltage circuits, serving as essential safety components for every electric vehicle.
Figure | TDK solutions for xEVs; Source: TDK
In terms of driving safety, TDK's automotive-grade IMUs cover a complete layout ranging from QM level, ASIL-B level, to the highest ASIL-D level (IAM-20689 series). Through the posture angle information output in real time by the automotive IMU, the vehicle's adaptive headlight system can dynamically adjust the angle of the automatic headlights when driving on bumpy roads, keeping the headlight illumination area horizontally stable at all times. This balances driving visibility and prevents dazzling oncoming vehicles.
Figure | TDK automatic headlight leveling technology solution & thermal management solution; Source: TDK
Beyond ensuring driving safety, TDK's one-stop supporting capability also extends to the field of automotive thermal management. Through the synergy of temperature sensors, pressure sensors, and angle sensors, data is fed back to the thermal management ECU in real time, precisely driving motors to adjust the refrigerant valve angle and fan louvers. Through this real-time closed-loop control, the system can implement precise energy allocation among the cabin, motor electronic control, and battery, significantly reducing overall vehicle energy consumption.
In terms of hardware miniaturization, TDK's material processes also extend to the consumer electronics and smart wearables fields. Its TMR (Tunneling Magnetoresistance) sensor measures only 1.8x1.8x0.4mm, comparable in size to a standard SMD capacitor. It has been widely applied in the joints of robot dexterous hands and under the magnetic axis mechanical keyboards for esports.
Figure | TDK TMR sensor; Source: TDK
For highly constrained spaces such as smart rings and Bluetooth earphones, TDK has launched a microchip antenna based on LTCC (Low Temperature Co-fired Ceramic) technology. Measuring only 1.6x0.8x0.4mm, it reduces the space of traditional antenna systems by at least half. As a standard component, this antenna can be directly mounted via SMT (Surface Mount Technology) and soldered together with conventional electronic components, significantly reducing processing costs and assembly time.
Figure | Ultra-small chip antenna; Source: TDK
Conclusion
The advancement of physical AI and high-density power supply essentially depends on extreme breakthroughs in underlying hardware regarding edge power consumption control and physical materials.
From the 5-10mW visual perception architecture of AIsight and the BalancedGyro technology that reduces temperature drift in humanoid robots, to the laminated foil sintering process that increases server capacitor capacity by 20%-30%, these technological advancements indicate that hardware optimization is shifting from minor tweaks to the integration of underlying materials and customized chip architectures. By transferring its top-tier thin-film processes accumulated in its traditional stronghold—hard disk drive heads—to sensors, and leveraging decades of expertise in electronic ceramics and precision sintering technology to empower a new generation of high-performance components, TDK provides concrete implementation pathways at the hardware foundation for smart glasses, embodied robots, and new-generation data center power supplies.