CES 2026 Observation: Computing Power Seeks Hardware Carriers, AI Reconstructs the Traditional Consumer Electronics Industry Chain
From January 6th to 9th, Pacific Time, the 2026 Consumer Electronics Show (CES) was held in Las Vegas. Different from previous years when the tech circle focused on the competition of large - model parameters, the core agenda of this year's exhibition has shifted to a specific industrial direction: "Physical AI".
"The 'ChatGPT moment' of Physical AI has arrived." Jensen Huang, the founder and CEO of NVIDIA, made such a judgment in a speech before the exhibition. He pointed out that machines are beginning to have the ability to understand the real world, reason, and take action.
This change stems from the overflow of upstream computing power and the need to find an outlet.
At this year's CES, chip giants continued their aggressive release rhythm. AMD released the Helios AI rack integrated with 72 MI455X accelerator chips. NVIDIA introduced the new - generation Rubin architecture and released the Alpamayo open - source autonomous driving model. Intel launched the 3rd - generation Intel® Core™ Ultra processors, and Qualcomm released the Dragonwing IQ10 general - purpose robot architecture.
Chip manufacturers are competing to increase the supply of computing power, but the huge model capabilities in the cloud urgently need specific terminal hardware to undertake and digest.
Chen Jun, the executive deputy general manager and chief analyst of Sigmaintell, told a reporter from Cailian Press that in previous CES exhibitions, exhibitors were mainly from the consumer electronics industry, while this year's CES more reflects the concept of "Physical AI", that is, the integration of AI and the real world. "The gradual integration of virtual AI and the real world, and AI's better understanding of the physical world will become the technological direction for some time to come." Chen Jun said.
Under this trend, China's supply chain will become the core force to undertake the implementation of this round of "Physical AI" thanks to its manufacturing experience in fields such as robotics and optical displays.
From the participation of more than 20 Chinese robot companies in the exhibition to smart glasses manufacturers' attempts to break away from the dependence on mobile phones, China's supply chain is promoting the transfer of AI from the "brain" to the "body".
Chinese robots move from the laboratory to industrial mass production
In the "Physical AI" landscape, robots are the highly - anticipated general carriers. According to incomplete statistics by a reporter from Cailian Press, more than 20 Chinese companies in the robot industry chain participated in this year's CES, covering the entire link from core components to complete machine manufacturing.
Different from previous static displays, this year's exhibitors focused more on demonstrating mass - production capabilities and specific industrial parameters.
At the exhibition, Zhongqing Robot displayed the full - size general - purpose humanoid robot T800. It is understood that this product is equipped with a full - stack integrated joint module, with a peak torque of 450N·m in coordinated output and an instantaneous peak power of 14000W at the joints. Fourier showed the new - generation GR - 3 humanoid robot and exhibited a desktop - level concept prototype using the same IP. Agibot released the new - generation simulation platform Genie Sim 3.0 and launched the Qiyuan Q1 robot that can be stored in a backpack. Yuejiang displayed the world's first household intelligent agent robot Rover X1.
In addition, several robot companies such as Unitree, Deep Robotics, and Pudu Robotics also made a collective appearance. Behind this large - scale collective move overseas is the upgrade of the supply - chain iteration efficiency.
An analyst told a reporter from Cailian Press that relying on the experience accumulated by upstream manufacturers in fields such as motors and sensors, the iteration speed of the domestic robot industry far exceeds that of robot companies in other regions such as Europe.
This efficient R & D rhythm enables technical indicators to be quickly converted into industrial standards.
In the upstream component segment, Zhaowei Machinery told a reporter from Cailian Press that the B20 series dexterous hand exhibited by the company has 20 degrees of freedom, with a total hand weight of only 600 grams, and its size is basically the same as that of an adult male's palm. It can respond to external pressure in combination with visual capture and flexible electronic skin technology.
In the perception aspect, Pacini Perception Technology displayed the multi - dimensional tactile sensor PX - 6AX - GEN3. Its humanoid robot TORA - ONE demonstrated the whole - process ice - cream making at the CES site, and grasped fragile ingredients through force - control accuracy.
In terms of radar, Hesai Technology announced at CES 2026 that it would cooperate with South Korea's Motion - AI company MOVIN to provide core sensors for its lidar - driven real - time motion capture system.
Oufeiguang told a reporter from Cailian Press that the company brought the latest progress in robot optical imaging technology. The company's R & D investment in the first three quarters was 801 million yuan to support the full - chain capabilities from optical design to system integration.
In addition to the "all - around" narrative of humanoid robots, the commercialization of specialized robots is progressing faster. Data provided by Chen Jun to a reporter from Cailian Press shows that the global sales of complete robot machines are expected to be about $45 billion in 2026 and are expected to exceed $60 billion in 2030.
In vertical scenarios, Ecovacs released the pool - cleaning robot ULTRAMARINE and the new GOAT series of lawn - mowing robots at this year's CES.
Public data shows that as of now, the global cumulative sales of Ecovacs' roller - type live - water floor - mopping robots have exceeded 1.6 million units. This kind of specialized equipment for specific scenarios (such as underwater and lawns) has achieved commercial returns of "Physical AI" first by stripping away the complex humanoid appearance.
Another cleaning - robot giant, Roborock, exhibited the world's first wheel - legged structure floor - sweeping robot. It is reported that this product is highly subversive in form - each wheel - leg has the ability to independently stretch, lift, and adjust the height, and can adjust its posture in real - time according to the ground conditions to keep the body stable in complex environments.
Nakul Duggal, an executive of Qualcomm Technologies, said at the CES forum that the technological innovations accumulated in multiple industries over the past decade are accelerating the integration between different fields. Physical AI has the potential for high - level "popularization", and the industry has the conditions for large - scale development.
AI reconstructs the traditional consumer electronics industry chain
Beyond robots, traditional consumer electronics and automobile manufacturers are using mature manufacturing supply chains to accelerate the implementation of AI technology in both "new" and "old" hardware products.
Currently, AI has almost become a standard feature of PCs. In addition to AI PCs, traditional PC manufacturers are also telling new stories.
A reporter from Cailian Press learned from Lenovo Group that at CES 2026, Jensen Huang and Yang Yuanqing, the chairman and CEO of Lenovo Group, jointly announced that they would cooperate to launch the "Lenovo Artificial Intelligence Cloud Super Factory". NVIDIA's newly released Vera Rubin will be an important part of this cooperation.
Home appliance manufacturers are also "fully involved in AI".
TCL Industries told a reporter from Cailian Press that the company comprehensively demonstrated various AI applications this time, including products such as the TCL Xiaolanyi fresh - air AI air conditioner and the TCL AiMe companion robot.
Dreame Technology exhibited "whole - house intelligence", that is, in a home environment, devices such as TVs, washing machines, refrigerators, and air conditioners can be automatically adjusted according to the environment and usage status.
On the other hand, traditional consumer electronics giants are trying to build smart glasses into a new terminal independent of mobile phones.
Thunderbird Innovation, a subsidiary of TCL Technology, released the Thunderbird X3 Pro Project eSIM. Thunderbird Innovation said that this product integrates an eSIM communication module and supports 4G protocols. It can independently complete functions such as calls, AI conversations, and real - time translation without connecting to a mobile phone or WiFi. Previously, Thunderbird Innovation announced the completion of a new round of financing of over 1 billion yuan, with investors including funds under China Mobile and China Unicom, indicating the communication industry chain's attention to the independent nature of smart glasses.
XGIMI, a home - projector manufacturer, released the AI glasses brand MemoMind at CES 2026, choosing the MicroLED + optical waveguide solution for its technical path. Rokid launched the Rokid Style, which weighs only 38.5 grams, trying to solve the problem of wearing comfort. Lenovo Group also displayed its new concept - level AI glasses.
Sigmaintell predicts that the global sales of complete AI glasses machines will be about $10.7 billion in 2026, a year - on - year increase of more than 80%.
The technological maturity of upstream component manufacturers provides a foundation for traditional terminal manufacturers to quickly enter the market. Guangfeng Technology participated in the exhibition with three LCoS AR optical engines such as the Dragonfly G1. The company said that its technical solution aims to push the price of consumer - grade AR glasses below $1500, competing with international mainstream products such as Meta Ray - Ban. Goertek released the AI + display smart glasses, the reference design Rubis, using a full - color resin optical waveguide module, with the module weight reduced to 4 grams.
In the automotive field, the upgrading of each link in the supply chain is accelerating the process of integrating large AI models into vehicles.
TCL CSOT, a subsidiary of TCL Technology, displayed the world's first printed OLED automotive screen, and the production line started construction in October last year. Tianma globally launched a new - generation intelligent cockpit solution, including a 49.6 - inch panoramic immersive "Tianxuan" screen that integrates the functions of the instrument panel, central control, co - pilot, and rear - view mirror.
Geely Auto released the all - domain AI 2.0 technology system at this year's CES. Its "Eva intelligent agent" integrates the Xingrui large AI model and the end - to - end large voice model. Great Wall Motor released the ASL 2.0 intelligent agent and the VLA full - scenario large model, emphasizing the role of AI in vehicle environment understanding, decision - making execution, and cross - scenario collaboration.
When computing power, screens, and communication capabilities converge towards the edge side simultaneously, hardware is no longer just a cold device. As Cristiano Amon, the president and CEO of Qualcomm, said in his speech, the next - generation personal AI devices will integrate the physical world and the digital world. "Whoever can master edge data and transform it into the ability to provide highly relevant services to users will be the winner in the AI competition."
This article is from the WeChat official account "Growth Enterprise Market Observation", author: Wang Biwei. It is published by 36Kr with authorization.