At the CES 2025 event, robots undeniably stole the spotlight as a key attraction. A multitude of manufacturers exhibited their diverse robotic offerings. Across various vertical applications, multifaceted service robots capable of handling intricate tasks—ranging from lawn-mowing robots, floor-cleaning robots, snow-clearing robots, pool-maintenance robots, to companion robots—displayed a variety of forms and functionalities. Notably, the renowned “Four Little Dragons” of floor-cleaning robots—Ecovacs, Roborock, Narwal, and Dreame—gracefully premiered their latest innovations. These robotic products are on the cusp of seamlessly integrating into both residential and professional environments.

Within the realm of humanoid robotics, the 14-member “contingent” of humanoid robots accompanying Mr. Huang Renxun at the CES exhibition garnered significant attention. STAR1, distinguished as the first humanoid robot to ascend the Great Wall, and Galbot G1, an embodied large-scale model robot from Yinhe Universal, were among the highlights. These biomimetic humanoid robots, characterized by their high-fidelity aesthetics and agile maneuvers, excel not only in executing fundamental tasks like walking and object manipulation but also in advanced capabilities such as natural language communication with humans, environmental awareness, and mission planning facilitated by integrated AI algorithms. This showcase has profoundly illustrated the monumental progression from science fiction to tangible reality in the evolution of humanoid robots.

At this year’s CES, AI technology has seamlessly integrated into an array of household appliances, resulting in substantial upgrades in both functionality and user experience. These appliances, spanning from AI TVs, refrigerators, air conditioners, to washing machines, have all undergone AI-driven enhancements. The AI refrigerator boasts capabilities such as displaying ingredient details, monitoring food storage conditions, suggesting recipes, and interconnecting with other smart devices. The AI washing machine provides tailored washing solutions to suit individual needs. The AI TV intelligently optimizes picture quality, analyzes users’ viewing habits, and offers personalized, bespoke viewing experiences. Meanwhile, the AI mirror aggregates and analyzes health data, integrating telemedicine functionalities. From leveraging AI technology to unveiling novel hardware products, tech giants are converging their focus onto the burgeoning field of smart homes.

At this year’s CES, AI has evolved beyond a mere concept, seamlessly integrating into diverse products, marking a profound milestone in its journey from theory to practice. Edge AI has emerged as a critical catalyst in this evolution. By harnessing Edge AI, wearable devices such as AI glasses can now deliver swift AI responses locally, negating the need for complex cloud computations. Smart home devices, leveraging this technology, can swiftly provide precise and personalized intelligent actions tailored to users’ daily routines and immediate contexts. Furthermore, Edge AI has vastly improved robots’ capabilities in perception, comprehension, decision-making, and execution, enabling them to accurately decipher humans’ intricate languages, gestures, and facial expressions, thereby facilitating seamless and precise real-time interactions. If last year’s CES underscored the consensus that “AI hardware heralds the future,” then the undeniable consensus emerging from CES 2025 is that “Edge AI represents the future of AI hardware.“

As detailed in our prior article, the successful deployment of Edge AI functionalities relies critically on low-power, high-performance chips. In the realm of computational power, chips must possess heightened processing capabilities to manage intricate AI algorithms and model inference efficiently. Given that numerous edge devices rely on battery power, chips must exhibit ultra-low power consumption to maximize device longevity. In terms of storage, chips necessitate ample memory to accommodate AI models and operational data, with swift read/write speeds being pivotal for seamless AI task execution.

To cater to these Edge AI prerequisites, chip manufacturers have incorporated specialized AI acceleration modules, notably Neural Processing Units (NPUs), to augment the processing efficiency of AI algorithms. Notably, iStarChip Semiconductor has emerged as a trailblazer in the industry by integrating an autonomous NPU into control chips. iStarChip’s Fusion Controller Unit (FCU), following the integration of the NPU, is capable of executing AI models directly within the chip, supporting a robust AI computing power of up to 1 TOPS (Tera Operations Per Second). This facilitates efficient AI processing on edge devices, thereby bolstering the performance of diverse AI hardware. Additionally, iStarChip Semiconductor’s proprietary FCU leverages multi-core heterogeneous technology, consolidating CPU, DSP, and NPU cores onto a singular chip. By enhancing chip integration, it not only minimizes costs and power consumption but also elevates system stability.

CES 2025 served as a pivotal exhibition, providing a vivid glimpse into the evolving trends of chips within the AI landscape. The surge of Edge AI has presented chip manufacturers with expansive market prospects and fostered a new, promising frontier, driving relentless advancements in performance, power consumption, and storage. Despite the conclusion of CES 2025, as stakeholders in the semiconductor industry, it is our duty to stay abreast of this trend, actively invest in research and development innovations, and collectively anticipate the dawn of the new AI epoch.