Core Viewpoint - The report from CICC maintains the revenue forecast for Hesai Technology (02525) for 2025, raises the 2026 revenue forecast by 6.4% to 4.53 billion yuan, and introduces a first-time revenue forecast for 2027 at 6 billion yuan, while maintaining an outperform rating for the industry [1] Group 1: Company Current Status - In January 2026, Hesai Technology showcased its latest lidar technology achievements at CES 2026 and updated its delivery volume, production capacity planning, and strategic partnerships [2] Group 2: Production and Delivery Capacity - The company is expected to deliver over 1.6 million units in 2025, with 24 OEMs securing over 120 models for mass production as of January 5. The annual production capacity is planned to increase from 2 million units in 2025 to 4 million units in 2026, achieving a doubling growth. The Bangkok Galileo factory is expected to commence production in early 2027. The company has a strong order backlog and is actively expanding domestic and international production capacity, which is likely to drive future shipment volume and revenue growth [3] Group 3: L3 Vehicle Approval and Lidar Quantity Increase - On December 15, 2025, the Ministry of Industry and Information Technology approved L3 level conditional autonomous driving vehicles for trial operation in designated areas. This is expected to accelerate penetration, with the number of lidar units per vehicle likely to increase to 3-6. The 2026 models of Avita 12 and Lantu Tianshan are expected to be equipped with 4 lidar units. Hesai showcased a new generation L3 automotive lidar solution at CES 2026, which has received the first mass production approval for passenger vehicles, with production planned to start by the end of 2026 or early 2027 [4] Group 4: Next-Generation AI and Lidar as a Key Engine - Global companies are accelerating the large-scale deployment of L4 autonomous driving fleets. The company has partnered with leading firms such as Motional, Baidu, Didi, WeRide, and Pony.ai, and has been selected by NVIDIA as a lidar partner for the "NVIDIA DRIVE AGX Hyperion 10 platform." Additionally, Hesai's JT series lidar is widely used in the robotics and industrial markets, with cumulative shipments exceeding 200,000 units, including applications in mowing robots, smart companion robots, and 3D spatial digitization devices [5]

Group 1: CES Overview - The 2026 International Consumer Electronics Show (CES) took place from January 6 to January 9 in Las Vegas, showcasing innovations in AI, XR Metaverse, smart home technology, digital health, electric vehicles, semiconductor chips, and sustainable technologies [1] - The event featured 4,112 global companies, with major players like Nvidia, AMD, Hisense, and TCL participating, despite Apple's absence [1] - AI and robotics emerged as the central themes of this year's CES, with a focus on comprehensive discussions and competitions surrounding AI technologies [1] Group 2: Nvidia's Developments - Nvidia's CEO Jensen Huang highlighted the demand for AI chips and introduced the open-source physical AI inference model Alpamayo, designed for autonomous driving scenarios [2] - The Alpamayo model operates on a visual-language-action architecture, enabling multi-step reasoning in complex road environments [2] - Nvidia's latest AI superchip platform, Vera Rubin, has entered full-scale production, boasting double the capabilities of its predecessor while improving assembly efficiency and energy control [2] Group 3: AMD's AI Strategy - AMD's CEO Lisa Su presented the company's AI product lineup, noting a surge in active AI users from 1 million to 1 billion since the launch of ChatGPT [3] - Su projected that active AI users could reach 5 billion by 2030, necessitating a 100-fold increase in global computing power in the coming years [3] - AMD introduced a series of new AI processors, positioning AI-driven personal computers as a core focus for future development [3]

Core Insights - The competition among major chip manufacturers is intensifying around the foundational computing power for AI, with native AI hardware accelerating its large-scale implementation, transitioning from laboratory settings to consumer and industrial applications [1][2] - Chinese manufacturers are playing an increasingly significant role at global tech events, showcasing their advancements in AI hardware and capabilities, which are supported by their supply chain and R&D strengths [2][11] Group 1: Computing Power and AI Development - The rapid evolution of underlying computing infrastructure is crucial for the accelerated development of AI large models, with chip leaders emphasizing the exponential growth of computing power and the new application opportunities it creates [3][4] - NVIDIA's CEO highlighted the arrival of the "ChatGPT moment" for physical AI, indicating that machines are beginning to understand and act in the real world, with autonomous taxis being one of the first applications to benefit [3] - AMD's CEO noted that the floating-point computing power for training AI models is growing fourfold annually, with inference token consumption increasing by 100 times over the past two years, necessitating new product offerings to meet Yotta-scale infrastructure demands [4] Group 2: Emerging AI Applications - The potential of edge AI is significant, with advancements in smart wearable devices being highlighted as a new category of mobile terminals that will coexist with smartphones [5][6] - Qualcomm's CEO projected that the market for personal AI devices could reach 100 million units in the coming years, emphasizing the importance of edge data for providing highly relevant user services [6] Group 3: Robotics and Physical AI - The maturity of physical AI was showcased at CES, with Chinese manufacturers presenting advanced robots that demonstrate improved capabilities compared to previous years [7][8] - The introduction of humanoid robots and their increasing commercial viability was noted, with companies achieving substantial progress in motion control and operational precision [9][10] - The integration of hardware, sensors, and environmental perception into scalable systems is seen as essential for advancing physical AI applications across various industries [10] Group 4: Chinese Manufacturers' Competitive Edge - Chinese companies are leveraging their supply chain efficiencies and R&D capabilities to drive rapid iterations and cost optimization in the robotics sector, significantly outpacing European competitors in product development cycles [11][12] - The shift from simple manufacturing to innovative solutions reflects a broader transformation in the perception of "Made in China" to "Created in China," highlighting the technological advancements and better solutions being offered [12]

Core Viewpoint - Major chip manufacturers are engaged in intense competition around the foundational computing power for AI, marking a shift from experimental AI hardware to large-scale deployment in consumer and industrial applications [1] Group 1: Industry Dynamics - The competition in the AI ecosystem reflects a significant restructuring of computing power foundations, indicating a deeper level of competition in the AI landscape [1] - Chinese manufacturers are leveraging their supply chain and research capabilities to transition from "manufacturing" to "creation," showcasing their evolving role in the global AI market [1] Group 2: Technological and Capital Integration - The convergence of technology, capital, and industrial forces is reshaping the competitive landscape of global intelligent technology [1] - The ongoing developments signify a critical moment in the AI ecosystem, where the integration of various elements is pivotal for future advancements [1]

Core Insights - The computer industry is undergoing a "platform reset" that occurs every 10-15 years, with a significant focus on physical AI and the upcoming "ChatGPT moment" [1] - Chinese robotics companies showcased a complete supply chain at CES 2026, with over 30 exhibitors demonstrating various robotic technologies and solutions [1][3] Industry Trends - The presence of Chinese robotics at CES is not new, but this year marks the most comprehensive collective display of the industry [3] - Companies like Yuzhu Technology and others presented innovative robots, highlighting the advancements in the field [3] Company Developments - New entrants like Pasini and Black Sesame Intelligence showcased their technologies, including high-precision tactile sensing and a comprehensive intelligent platform for robotics [4] - Magic Atom reported immediate orders for their products, indicating a shift from mere product display to commercial viability [7] Market Strategy - Companies are focusing on global market expansion, with plans to establish a flexible overseas inquiry and production system to meet international demand [10] - The shift in customer demographics at CES indicates a growing interest from professional buyers, leading to more serious discussions about collaboration [9] Competitive Landscape - The competitive environment is intensifying, with Chinese firms recognizing the need to accelerate R&D and market capture to maintain their lead [9] - The collaboration between foundational technology providers like NVIDIA and Chinese robotics companies illustrates a dual approach to advancing the industry [10]

Core Insights - Nvidia showcases comprehensive AI infrastructure deployment and iteration, emphasizing the application prospects of physical AI [1][3] - AMD and Intel have made significant progress in previously weaker areas, such as rack capabilities and the 18A process node [1][4] Group 1: Nvidia Developments - Nvidia's new AI platform Rubin has entered full-scale production, featuring six new chips: Vera CPU, Rubin GPU, NVLink 6 switch, ConnectX-9 super network card, BlueField-4 DPU, and Spectrum-6 Ethernet switch [3] - Rubin GPU achieves 3.5 times the training speed and 5 times the inference speed compared to the Blackwell architecture, with a peak computing power of 50 Petaflops [3] - The token cost during inference is reduced by up to 10 times compared to the Blackwell platform, and the number of GPUs required for training mixture of experts (MoE) models is reduced to one-fourth [3] - Nvidia introduces a memory storage platform driven by BlueField-4 to address KV Cache issues, enhancing token processing speed by up to 5 times in specific scenarios [3] - Microsoft and Coreweave are set to be the first customers to deploy Rubin in the second half of 2026, with Microsoft's next-generation Fairwater AI super factory scaling to hundreds of thousands of Vera Rubin chips [3] - Nvidia's Alpamayo series VLA open-source AI models and tools are aimed at autonomous vehicle development, with the DRIVE system entering mass production for the Mercedes-Benz CLA, expected to hit the US roads in 2026 [3] Group 2: AMD and Intel Innovations - AMD launches the Helios rack based on the MI 455X, featuring a fully liquid-cooled design with four Instinct MI455X GPUs and one EPYC Venice Zen6 CPU [4] - The MI500 series chips are in development, expected to enhance AI performance by 1000 times by 2027 [4] - Intel introduces its first computing platform based on the 18A process node, the Intel Core Ultra 3 series (codenamed Panther Lake), achieving a total AI computing power of 180 TOPS, with the GPU contributing 120 TOPS [4] - The Core Ultra 3 platform supports running a 70 billion parameter large language model locally with 32k context, with consumer laptops featuring this processor set to pre-sell on January 6, 2026 [4]

Core Insights - Nvidia's AI chip business generated nearly $148 billion in revenue over the past nine months, significantly surpassing the $27.5 billion from the same period in 2023, but the company's transition to an integrated hardware-software platform has faced major setbacks [1] - The Omniverse software, which was intended to be a core tool for creating digital twins in manufacturing and logistics, has seen minimal revenue and a stalled commercialization process, leading to the decision to shut down the Omniverse Cloud service by August 2025 due to lack of demand [1][3] - CEO Jensen Huang expressed frustration over the slow progress of the Omniverse division, criticizing the team for focusing on demonstrations rather than product development, and highlighting the lack of widespread adoption by large enterprises [1][4] Revenue and Market Response - Despite the explosive growth in AI chip revenue, the market has not reacted strongly to the revenue gap from Omniverse, indicating the challenges Nvidia faces in establishing a second growth curve [2] - The inability to address software usability and industry adaptation issues may hinder Nvidia's ambitions in robotics and industrial digitalization for the long term [2] Demand and Service Closure - Omniverse was launched in 2021 as a platform for designers to collaborate on 3D designs, but the reality has fallen short of expectations, with few clients actually signing on for large-scale simulations [3] - Developers have reported that the platform is difficult to use, incomplete, and prone to crashes, leading to the termination of the cloud service project [3] Internal Pressure and Management Concerns - Huang's anxiety over Omniverse's performance is evident, as he has pressured the team to find new revenue sources and has expressed frustration in internal meetings regarding the lack of profitability and the team's focus on demonstrations [4] - The actual outcomes of collaborative projects have also led to dissatisfaction among management, particularly regarding the scale of partnerships with companies like BMW [4] Long-term Challenges and Industry Barriers - Nvidia executives compare Omniverse to CUDA, suggesting that it may take years of investment to fully realize its potential in the "physical AI" market [6] - The company faces intense competition and structural barriers in the robotics simulation field, with many large enterprises preferring to develop their own internal simulation software rather than relying on Nvidia's platform [6] - Industry-specific technical challenges and cost-effectiveness issues also pose significant obstacles to the widespread adoption of Omniverse [6][5] Development and Market Creation - Currently, Omniverse is seen as a horizontal open platform for developers rather than a complete application, indicating that Nvidia's attempt to create a market from scratch will require a lengthy nurturing period [7]

Core Insights - The article highlights the acceleration of native AI hardware development, emphasizing the growing demand for computing power and the expansion of AI applications from cloud to edge [1][2][12] - Chinese manufacturers are increasingly becoming significant players at CES 2026, showcasing a variety of AI hardware, including AI glasses and robots, which have garnered substantial attention [1][12] Group 1: AI Hardware and Computing Power - The rapid evolution of underlying computing infrastructure is supporting the accelerated development of AI large models, with a noted annual increase in floating-point operations for training by four times and a 100-fold increase in inference token consumption over the past two years [2][3] - NVIDIA's CEO highlighted the arrival of the "ChatGPT moment" for physical AI, with autonomous taxis being one of the first applications to benefit from this technology [2] - AMD's CEO discussed the transition to Yotta Scale compute, indicating that AI potential will be fully unleashed across data centers, personal computing devices, and edge AI [2][3] Group 2: Robotics and Physical AI - The maturity of physical AI was emphasized at CES, with a notable improvement in the capabilities of showcased robots compared to previous years, indicating a shift from simple demonstrations to more complex functionalities [7][8] - Companies like UTree Technology and Zongqiong Robotics are showcasing advanced robots capable of performing intricate movements, signaling a move towards mass production and real-world applications [8][9] - The integration of edge AI and real-time situational awareness in wearable devices is expected to create significant market opportunities, with projections of reaching 100 million units in the coming years [5][6] Group 3: Chinese Manufacturers and Global Role - Chinese firms are transitioning from being merely the "world's factory" to becoming key players in global technological innovation, supported by strong supply chain and R&D capabilities [1][12] - The efficiency of China's supply chain allows for rapid product iteration, with a production cycle of about one year compared to three to five years for European companies [12][13] - The article underscores that the advancements in AI technology are not just about manufacturing but also about creating better solutions and technologies, marking a shift from "Chinese manufacturing" to "Chinese innovation" [13]

Core Viewpoint - Qualcomm is showcasing its advancements in personal AI and physical AI at CES 2026, emphasizing the shift from cloud dependency to terminal autonomy through powerful edge computing capabilities [2][6][7]. Group 1: Personal AI Developments - Qualcomm's Snapdragon X series processors are enabling a new generation of AI PCs, demonstrating capabilities such as real-time multimodal AI content creation without latency [2][6]. - The demand for AI PCs has surged, with companies like Lenovo and ASUS reporting a 2.5 times increase in AI PC product offerings within a year [9]. - 62% of users consider "local intelligent agent invocation" a core reason for purchasing AI PCs, highlighting the importance of local processing in enhancing user experience [13]. Group 2: Physical AI Innovations - Qualcomm's advancements in physical AI are evident in the automotive sector, with over 400 million vehicles utilizing Snapdragon digital chassis solutions [19]. - The company is leading in the in-car infotainment and cockpit SoC solutions, with over 75 million vehicles adopting the Snapdragon cockpit platform [19]. - The Snapdragon 8797 platform supports multi-modal large models for cockpit and driving assistance, enhancing the in-vehicle experience [21]. Group 3: Strategic Vision and Market Trends - Qualcomm's strategy focuses on building a comprehensive ecosystem that integrates personal AI and physical AI, aiming for a seamless user experience across various devices [8][26]. - The global AI PC market penetration is expected to exceed 80% in the next three years, indicating a significant shift towards edge computing [27]. - The company is positioned to capitalize on the exponential growth in demand for edge computing capabilities, as AI evolves towards intelligent agents and embodied forms [10][11]. Group 4: Technological Advancements - The Snapdragon X2 Plus platform features an 80 TOPS NPU, setting a new speed record for its class, and offers a 35% performance improvement while reducing power consumption by 43% compared to previous generations [15][29]. - Qualcomm's new processors, such as the Q-8750 and Q-7790, are designed for diverse IoT applications, enhancing the company's ability to support a wide range of smart devices [26][39]. - The Snapdragon Ride platform has been validated in over 60 countries, accumulating extensive data that enhances Qualcomm's capabilities in physical AI applications [46].

Core Viewpoint - The article emphasizes the urgency of the threat posed by the advancement of Physical AI, as represented by NVIDIA, which is pushing AI into real-world manufacturing, potentially reviving the U.S. manufacturing sector and diluting China's engineering and skilled labor advantages [7][20]. Group 1: NVIDIA's Strategy and Physical AI - NVIDIA's CEO Jensen Huang's keynote at CES focused on reducing the development costs of Physical AI, which is essential for AI factories [10][20]. - Physical AI enables autonomous systems to perceive, understand, reason, and perform complex actions in the physical world, contrasting with generative AI that primarily processes language [13][14]. - The training costs for Physical AI are significantly higher than for generative AI due to the complexity of understanding real-world physics [15][16]. Group 2: Technological Advancements and Implications - The introduction of the Vera Rubin platform by NVIDIA significantly enhances inference performance, reducing costs to one-tenth of the previous generation, which will decrease the demand for GPUs in AI enterprises [19][20]. - The Cosmos model allows for pre-trained multimodal models that facilitate the development of Physical AI, enabling virtual training for robots without the need for real-world trials [19][20]. Group 3: Competitive Landscape and Market Dynamics - NVIDIA's shift from a GPU supplier to a competitor in the autonomous driving market poses a significant threat to existing players, particularly in China's emerging electric vehicle sector [22][24]. - The collaboration between NVIDIA and companies like Mercedes for smart driving cars indicates a strategic move to integrate AI systems into manufacturing, potentially disrupting the industry [22][25]. Group 4: Future Directions and Recommendations - The article suggests that China must enhance its AI infrastructure investment to match the U.S. dominance in computational power and data centers, which currently sees the U.S. holding over 70% of global computing power [32][33]. - The need for a unified approach within China's AI industry is highlighted, emphasizing the importance of collaboration to develop competitive alternatives to NVIDIA's Physical AI [31][32].