Embodied Intelligence Standardization Accelerates as Working Group Formed; Institutions Survey Industry Chain

　　The drafting working group for embodied intelligence data standards has been officially established,simulated training, world models, and data standardization resonate together, accelerating the opening of the “model training-real interaction-data backflow” data flywheel, the institution this month Intensive research11 concept stocks.

　　Embodied intelligence data standards

　　On November 10, the National Data Standardization Technology Committee's second “Standard Week” activity of 2026 was held in Xiamen, Fujian.

　　According to the relevant person in charge, the National Data Standardization Technology Committee has officially established the Embodied Intelligence Data Standard Drafting Working Group this time, as well as meteorology, health and hygiene, power...the drafting working group for 4 industry applications including commerce, to accelerate the formulation of data-related standards during the “15th Five-Year Plan” period, and empower the digital development of industries and industrial chains.

　　In addition, the latest data shows that our country's total computing power scale has reached 3.1 million PFLOPS, ranking second in the world.

　　Standardization points directly at the core bottlenecks of the current industry. One of the core bottlenecks of embodied intelligence development is the insufficient supply of high-quality physical interaction data—currently, the amount of high-quality embodied data available on the market is only on the order of 1,000,000 hours, while to benchmark against the capability base of general large models, embodied models are expected to need on the order of 100,000,000 hours of other effective data, the supply-demand gap is very prominent.

　　Industry insiders pointed out that the establishment of data standards means that embodied intelligence data will move from being self-collected and self-used by various companies toward a standardized stage that is measurable, flowable, and reusable.

　　At the industrial level, global tech giants have recently been intensively establishing a foothold in physical AI.

　　Apple The humanoid Robot autonomous control system HALO disclosed by the research team chose to Unitree Technology complete real-machine verification on the G1 body, achieving an average success rate of 91% in 360 single-item task tests across 6 real environments that were not pre-scanned through training with large-scale simulation synthetic data; NVIDIA Continuously expanding Isaac open Robot development platform; Meta and Fourier collaborated to develop a humanoid Robot prototype with the internal code name Max.

　　On the domestic side, PetroChina Kunlun Manufacturing and Kunlun Capital under the group plan to jointly establish with Unitree and Fanjie Intelligence jointly establish a joint venture, focusing on industrial embodied intelligence Robot Solution.

　　Physical AI heading towards the commercialization tipping point

　　Physical AI refers to systems capable of understanding and acting upon the real physical world Artificial Intelligence systems, in order to Robot Embodied intelligence, with carriers such as autonomous driving, is its core form.

　　Fangzheng Securities In the latest industry weekly report, it pointed out that Apple HALO verified the feasibility of large-scale simulation synthetic data-driven Robot skill learning, which is expected to reduce the cost of real-machine data collection and accelerate the integration of robot motion control and environmental perception capabilities; Apple、NVIDIA and other giants are accelerating physical AI technology breakthroughs, while real-scene verification and commercial landing of robots continue to advance.

　　Guolian Minsheng Securities predicts,2026 China Embodied Intelligence Robot User spending will break through 11,000,000,000 USD, with a growth rate of nearly 120%, the industry is moving from a market on the order of 10,000,000,000 towards the on the order of 100,000,000,000 track, humanoid robots Large models are undergoing a paradigm leap, and general intelligence is accelerating to emerge.

　　The mass production pace is also being validated. CITIC Securities points out that Tesla The third-quarter report earnings call released a signal that Optimus mass production exceeded expectations; combined with the earlier deep correction in the robotics sector, industrial catalysis is spreading from the complete machine body toward core components.

　　Institutions are keeping a close eye on multiple concept stocks

　　According East Money“Hot topics” sorting, currently there are 79 stocks in the A-share market involving physical AI concepts, with a total market capitalization of approximately 2.67 trillion yuan. Muxi Co., Ltd.、Huachuan Technology、Huaqin Technology、Rockchip and others are at the forefront in terms of volume.

　　Since the beginning of the year, physical AI concepts have seen mixed gains and losses, but the sector's trend has been strong since November, except for Weichai Power、Foxconn Industrial Internet and 4 other stocks fell slightly, while the remaining concept stocks all rose significantly: Alte The increase within the month is already close to 40%, MoveOne、Rockchip、Suochen Technology all rose by more than 30%, and there are also Fantuo Shuchuang、Jingwei Hengrun、Folai New Materials、Xingchen Technology and 13 other stocks rose by over 20%.

　　From the perspective of capital, East Money Choice data shows that a total of 24 physical AI concept stocks received the favor of financing clients during the month, among which, Rockchip received a large position increase of 429000000 yuan with leveraged funds, Weichai Power、Tianyu Shuke respectively received aggressive acquisitions of 135000000 and 88000000 yuan, Suochen Technology、Haitian Rider、Fengyuzhu、Keli Sensing Each obtained over 40,000,000 yuan in net buying financing.

　　In terms of institutional attention, since the beginning of this month, a total of 12 physical AI concept stocks have received institutional research, among which, Haitian Rider、Xiechuang Data respectively received clustered investigations from 156 and 150 institutions, OptiLight、Tuosta (Rights Protection) both received attention from over 60 institutions, Folai New Materials、Lingyunguang、Weichai Power、CapCloud、Kunlun Wanwei (Rights Protection)'s number of institutional research visits are all above 25.

　　Haitian Rider When accepting institutional research recently, it was stated that the company has already implemented multiple embodied intelligence projects and can provide full-chain services including data solution design, collection, labeling, and evaluation for customers. The customer coverage of physical AI data business humanoid robots includes three categories: enterprises, internal embodied teams of major technology companies, and automakers, with representative customers including AGIBOT, Jiji Shijie, etc.

　　Folai New Materials Developed multi-modal technical routes such as piezoresistive and capacitive, designed three categories of product configurations, mainly serving entities such as enterprises with existing motion capture capabilities, large model enterprises, embodied training grounds, and data service providers. The company also pointed out during the research that flexible sensors Project revenue accounts for an extremely low proportion of the company's overall revenue and has no significant impact on the company's main business.

　　(Article source: East Money Research Center)

Massive information, precise analysis, all within the Sina Weibo APP