AI Education Intelligent Hardware in the First Half of 2025: A Battle to Break the Deadlock in Technological Iteration and Ecosystem Reconstruction
In the first half of 2025, driven by both technological iteration and demand upgrading, the domestic AI education intelligent hardware market demonstrated innovative vitality. From the hardware upgrades of educational companies such as Gaotu and Xueda, to the breakthroughs in scenario applications by technology enterprises like Youdao and Tomcat, and then to the ecological co - construction by underlying technology providers such as Biren Technology and Huawei, the entire industry is experiencing a transformation from single - tool products to intelligent ecosystems.
Technological Breakthrough: Large - scale Models Reconstruct the Underlying Logic of Educational Hardware
The in - depth application of large - scale AI models became the most prominent technological feature in the educational hardware industry in the first half of 2025. Different from the software - access - based model in 2024, major manufacturers integrated large - scale model capabilities into the hardware's underlying layer in the first half of this year, achieving a shift from "function stacking" to "intelligent nativization".
Gaotu's Civil Service Exam Passing Pad, released in February, is a typical example of this trend. Based on Huawei's Qingyun AI C5e (2nd generation) hardware platform, it deeply integrates the HarmonyOS operating system with Gaotu's self - developed Zhixue Exam - Passing System, realizing "data - driven evolution". According to Luo Bin, the co - founder of Gaotu, the differentiating competitiveness of the Exam Passing Pad lies in the "implantation of educational DNA". Its built - in "famous teacher resource library + Zhixue Exam - Passing System" constructs a complete learning cycle from goal setting, real - time Q&A to weak point tracking. The data from students' usage can be fed back in real - time to optimize teaching and research, forming an intelligent evolution path of "the more you use it, the better it understands you". This model, which deeply combines educational scenario data with large - scale model training, breaks the limitations of traditional hardware as mere "tools".
Xueda upgraded its intelligent hardware product line twice in February and June. Its learning machines and practice machines are fully connected to the DeepSeek large - scale model, and it has self - developed the Jiuzhang large - scale model to form a dual - core drive. Zhao Puzheng, the person in charge of Xueda's intelligent learning, emphasized that against the backdrop of the new curriculum standard reform, learning machines have been upgraded from "knowledge carriers" to "ability - training tools". The newly added 260,000 synchronized exercises not only focus on real - life background design but also build a "test frequency - difficulty" matrix through large - scale model analysis to help students obtain core scores in limited time. Notably, in the P, S, and T series of learning machines released in June, Xueda introduced the "filtered learning" and "filtered practice" functions for the first time. It automatically generates course mind maps through a large - scale model - driven knowledge graph and intelligently filters out unmastered questions to improve learning efficiency.
Youdao's SpaceOne dictionary pen, launched in February, created a new form of "AI - native learning hardware". As the first domestic hardware product pre - installed with the DeepSeek - R1 deep - thinking model, its "Youdao Xiaop" function can provide interactive Q&A and in - depth explanations across all school levels, subjects, and question types. In actual tests, SpaceOne's in - depth thinking process for high - school math problems on geometric progressions demonstrated the reasoning ability of the large - scale model, while its analysis of International Mathematical Olympiad questions showed its advantage in cultivating logical thinking. Its 3.6 - cm ultra - wide pen tip for scanning and an 8 - megapixel AI camera for capturing questions, combined with DeepSeek's multi - round interactive Q&A, aim to initially achieve an "answering and tutoring experience that surpasses that of human teachers".
Product Innovation: Scenario Segmentation Drives Diversification of Hardware Forms
In the first half of 2025, product innovation in AI education hardware showed obvious characteristics of scenario segmentation, expanding from traditional learning machines and dictionary pens to more diverse forms, covering the full - range of scenarios from K12 to vocational education, and from learning tools to emotional companionship.
In the K12 field, the Smart Star AI Learning Desk jointly launched by Dukang and Mengkubao in February represents an innovative direction. It integrates an 8 - inch screen into the learning desk. Through the "integrated desk lamp" design and AI intelligent light - sensing technology, it combines the three functions of eye protection, learning companionship, and emotional support. Its voice assistant "Xiaolang Xiaolang", which supports the DeepSeek large - scale model, can achieve functions such as fingertip - based instant query and photo - based Q&A, meeting the learning needs of preschool to primary - school students.
Educational hardware for adult and vocational learning scenarios witnessed growth in the first half of the year. Gaotu upgraded its Postgraduate Entrance Exam Pad in June. Based on Huawei's hardware platform, it added intelligent correction and learning diagnosis functions. Through AI technology, it can correct mistakes with one click and generate monthly learning reports. Zhonggong Education's AI Employment Learning Machine, launched in June, focuses on the entire chain of "career choice - ability improvement - employment achievement". Its core interactive assistant "Teacher Xiaolu" can assist in career assessment, job planning, etc. It integrates course resources for 8 major employment directions such as civil service recruitment and public institution examinations and provides intelligent tools such as AI essay correction and AI job - person matching. Zhengbao Accounting Online School's accounting learning machine is driven by a self - developed large - scale finance and taxation model and DeepSeek. It has 8 major accounting certificate courses and over 150,000 question banks built - in. Through the dual - mode Q&A of "AI tutor + human", it can improve learning efficiency by 30% - 50%.
Emotional - companionship hardware became one of the most - watched innovative tracks in the first half of the year. Tomcat's AI voice - based emotional - companionship robot, which was trial - launched at the end of 2024, was fully rolled out in the first half of 2025. In June, it announced the collaborative voice - control function with Unitree's robotic dog, exploring the application scenarios of a companionship robot that combines voice interaction and mobility. Luobo Intelligence, a startup in the AI companionship hardware field, received tens of millions of investment from well - known venture capital firms such as Shanghai Film Group and GSR Ventures. Its AI - nurtured trendy toy "Fuzozo" targets the emotional needs of Generation Z. Through "personality evolution" and "long - term memory" technologies, it creates a plush key - chain - shaped companionship product that supports "tap - to - share" social fission, showing new possibilities for the integration of trendy toys and AI. "Fuzozo" was launched at the Barcelona Consumer Electronics Show in March 2025 and officially put on the market in June.
Ecosystem Co - construction: Business Exploration of Software - Hardware Collaboration and Cross - border Integration
In the first half of 2025, the competition in the AI education intelligent hardware industry has upgraded from single - product competition to ecosystem competition. Cross - border cooperation among hardware manufacturers, educational institutions, and technology companies has become the main theme, forming various innovative ecosystem co - construction models.
The cooperation model of "hardware manufacturer + educational institution" is the most common. For example, as an underlying hardware technology provider, Huawei carried out in - depth cooperation with multiple educational enterprises such as Gaotu and Biren Technology in the first half of the year. As mentioned above, Gaotu's Exam Passing Pad is developed based on Huawei's Qingyun AI C5e (2nd generation) tablet, using the open - customization capabilities of HarmonyOS to achieve functions such as learning situation tracking and wrong - question management. Biren Technology, in cooperation with ZTE and Zhejiang University, launched the Zhihai AI Education All - in - One Machine supported by domestic computing power. It integrates the DeepSeek course system and the Zhihai platform, forming an integrated solution including AI general courses, professional courses, and practical training. It has been demonstrated and applied at Zhejiang University. In this model, hardware manufacturers provide computing power and underlying technology support, while educational institutions or schools are responsible for content and scenario construction, achieving complementary advantages.
The cross - border integration of "educational IP + AI technology" shows vitality. Mobee, a brand under Xueda, held an offline pop - up event in May in cooperation with the Peppa Pig IP. Its Mobee Thinking Machine, in the form of a "screenless learning machine + touch - sensitive paper cards" and combined with the influence of the Peppa Pig IP, creates an immersive "learning - while - playing" scenario. Tomcat, based on its IP with 24 billion downloads, upgraded the "Talking Tom Cat" to the "Chatting Tom Cat AI Robot", strengthening emotional connection through the IP persona and differentiating itself from standardized products of large manufacturers. This model uses the user recognition of mature IPs to reduce product education costs and endows the IP with new interactive capabilities through AI technology, expanding business monetization paths.
The ecosystem model of "technology platform + industry application" is taking shape. DeepSeek, the most active large - scale model technology provider in the first half of the year, has been connected to nearly ten leading educational technology enterprises. In addition to Gaotu, Xueda, and Youdao mentioned above, products such as the Smart Star AI Learning Desk jointly launched by Dukang and Mengkubao and Biren Technology's Zhihai AI Education All - in - One Machine are all supported by DeepSeek as the core technology. This model of "general large - scale model + industry - specific application" lowers the technical threshold for small and medium - sized manufacturers and accelerates the popularization of AI education hardware.
Industry Perspective: In - depth Analysis of Growth Drivers and Potential Challenges
Although the domestic AI education intelligent hardware market showed a booming development trend in the first half of 2025, in - depth analysis reveals that while the industry growth is driven by technological progress, it also faces multiple challenges such as market competition, user habits, and technological ethics.
In terms of growth drivers, the dual drive of policies and demand forms the core logic of the industry's development. This year's government work report once again proposed to promote the "AI +" initiative, and local governments' policy support for education technology has been continuously increasing, creating a favorable policy environment for AI education hardware. At the same time, the upgrading of consumer demand also provides broad space for the industry. According to market forecasts, the domestic education intelligent hardware market will continue to grow in 2025, with the market scale expected to exceed 100 billion yuan. Parents' dual concerns about their children's learning efficiency and eye health drive the demand for learning machines and eye - protecting tablets. The demand for emotional companionship from Generation Z and the elderly creates a new market space for AI companionship robots.
The experience upgrade brought about by technological progress is another important growth driver. The improvement of the reasoning ability of large - scale models enables education hardware to shift from "answer provision" to "thinking cultivation", as demonstrated by Youdao's SpaceOne in its in - depth thinking process for math problems. The development of multi - modal interaction technology endows hardware with more natural human - machine interaction capabilities, as seen in the synchronized presentation of expressions, movements, and voices of the Tomcat robot. The progress in hardware craftsmanship, such as the ultra - light 27.6 - gram body of Huawei's Smart Glasses 2 and the paper - like eye - protecting screen of an education technology company's learning and practice machine, significantly improves user experience and promotes product penetration.
However, the industry also faces many challenges. Intensified market competition has led to serious homogenization. Many new products were launched in the first half of the year, and currently, the market revenue of the education intelligent hardware vertical mainly comes from traditional categories such as learning machines and dictionary pens, with little functional difference and heavy reliance on price competition. The competition between Zhonggong Education's AI Employment Learning Machine and Gaotu's Postgraduate Entrance Exam Pad in the vocational education scenario, and the competition among education technology companies, large technology manufacturers, and traditional hardware manufacturers in the K12 learning machine market all face problems such as high user acquisition costs and difficulty in breaking through market share.
Technological ethics and user privacy issues are becoming increasingly prominent. While AI education hardware collects students' learning data, ensuring data security has become an industry consensus. Products such as Gaotu's Exam Passing Pad and Xueda's learning machines emphasize data encryption and privacy protection. However, with the expansion of hardware functions, such as the long - term memory of the Tomcat robot and the personality evolution of Luobo Intelligence's Fuzozo, clearer definitions of the boundaries of user data use and ethical norms are needed. In addition, for functions such as AI essay correction and intelligent learning path recommendation, the industry needs to consider how to avoid the degradation of learning ability caused by over - reliance.
Cultivating user habits and market education still take time. Although the functions of AI education hardware are becoming increasingly rich, some parents, especially those of young children, still resist "screen - based learning". Dukang's Smart Star AI Learning Desk tries to alleviate this contradiction through the "learning desk + screen" form, but its market acceptance still needs to be observed. In addition, vocational education hardware faces challenges such as low learning motivation among adults and an incomplete payment habit. Enterprises such as Zhonggong Education and Zhengbao Accounting Online School need to pay more attention to motivating learning and visualizing learning effects in product design.
Outlook for the Second Half of the Year
Looking ahead to the second half of 2025, the AI education intelligent hardware industry is expected to enter a new development stage with the dual opportunities of technological innovation and new product categories. The following three major trends are worth noting:
The popularization of edge - side large - scale models will reshape the hardware architecture. With the progress of edge - computing technology, AI education hardware will transform from the "cloud - based large - scale model + terminal hardware" model to the "edge - side large - scale model + lightweight hardware" model. Tomcat's plan of "evolving from cloud - based large - scale models to edge - side large - scale models" revealed in June and Biren Technology's local deployment plan for the Zhihai AI Education All - in - One Machine indicate the increasing importance of edge - side computing power. This trend will bring two - fold changes: first, hardware design will focus more on computing power integration, such as the Qualcomm Snapdragon AR1 chip installed in Huawei's Smart Glasses 2; second, user privacy protection will be strengthened as sensitive data can be processed without being uploaded to the cloud. It is expected that more manufacturers will launch education hardware products powered by edge - side large - scale models in the second half of the year.
Scenario integration and ecosystem openness will become the keys to competition. The integrated forms such as "learning desk + tablet" and "dictionary pen + AI assistant" that emerged in the first half of the year will further evolve to form a cross - scenario learning ecosystem. Xueda's T - series learning machines, launched in June, have attempted to break the boundaries of learning scenarios, supporting multi - scenario switching such as homework tutoring, focus mode, and exploration mode. The parent - child learning large - screen jointly launched by Namihe and Hisense combines the functions of a "learning machine + best - friend machine + fitness mirror", meeting the needs of multiple family members. In the future, education hardware will be interconnected with smart home devices and wearables, forming a full - scenario ecosystem of "learning + living". Huawei's "1 + 8+N" ecosystem and Xiaomi's AIoT layout provide possibilities for this integration.
Internationalization and vertical markets will become new growth points. As competition in the domestic market intensifies, going global has become a strategic choice for leading enterprises. Tomcat plans to expand its AI emotional - companionship robots to overseas markets such as Europe, America, and Southeast Asia. Gaotu's civil - service - exam training model of the Exam Passing Pad may find an entry point in the Southeast Asian Chinese market. In addition, in - depth development of vertical markets will be a breakthrough for small and medium - sized manufacturers. For example, Biren Technology focuses on the AI education market in universities, Luobo Intelligence targets the emotional - companionship needs of Generation Z, and Zhengbao Accounting Online School delves into the accounting vocational education market. Although these vertical markets are limited in scale, users have a high willingness to pay, and the competitive landscape is not yet solidified, presenting great opportunities.
In the first half of 2025, the domestic AI education intelligent hardware industry completed an important iterative upgrade through the collision of technology and scenarios. From the "software - hardware collaboration" of Gaotu's Exam Passing Pad to the "AI nativization" of Youdao's SpaceOne, from the "emotional companionship" of the Tomcat robot to the "computing - power popularization" of Biren Technology, the industry is answering the question of "how AI can reshape education" with innovation. Although challenges remain, the wave of technological progress and the rigid demand for education ensure the long - term value of this track. For industry participants, only by grasping the possibilities of technological innovation, understanding user needs, and building ecological barriers can they gain an edge in this educational technology revolution.
This article is from the WeChat official account "Edu Guide" (ID: EduZhiNan). Author: Edu Guide. Republished by 36Kr with permission.