Since late October, the physical AI industry chain has continued to surge, becoming a new hotspot for the AI industry and a new focal point for A-shares.
Zhang Xiaoguo, fund manager of the Truvalue Xingxuan Industrial Trends Mixed Fund, stated that the strong performance of the physical AI industry chain is backed by a triple resonance at the industry level: significant improvement in model capabilities, increased efficiency in the toolchain, and accelerated deployment of embodied carriers. The large-scale mass production phase of humanoid robots from 1 to N is gradually approaching. Zhang Xiaoguo believes that this rally is not a switch in the main AI investment theme, but rather a continuation and evolution of that theme; the sector's pricing logic is shifting from trading the 0-to-1 phase of humanoid robots to trading the 1-to-N phase of embodied intelligence, and the investment methodology is also transitioning from concept-driven investing to industry-driven investing. He pointed out that the physical AI industry is currently on the eve of large-scale validation, standing at the critical point of accelerated development, with high allocation value. He is bullish on three directions: the data collection industry chain, humanoid robot hardware components, and body/robot manufacturers. He expressed strong optimism about key targets in the physical AI industry chain, the Tesla humanoid robot industry chain, and high-quality targets in the domestic humanoid robot industry chain with strong positioning advantages and competitiveness.
1. The physical AI industry chain has performed remarkably since October. What is the main logic behind this?
Zhang Xiaoguo, fund manager of the Truvalue Xingxuan Industrial Trends Mixed Fund: The strong performance of the physical AI industry chain since October is backed by a triple resonance at the industry level. On the model side, GPT-6 Astra has demonstrated the ability of a general foundation model to directly serve as a policy model on embodied datasets, with performance significantly improved compared to previous versions. The model was fed a large amount of embodied intelligence EGO (first-person view) video data during training. The improvement in GPT-6 Astra's physical AI capabilities indicates that the scaling law still holds in the physical AI field, and the physical AI industry is expected to usher in a qualitative leap in performance. The biggest bottleneck in the phased development of the humanoid robot industry is gradually loosening. On the toolchain side, NVIDIA's GTC conference strung together simulation, training, and deployment into an open-source full stack (Isaac OS, Isaac Lab 3.0, GR00T, Agile ONE SN reference body, Human 1M dataset), compressing development environment setup from weeks to days. On the embodied carrier side, OpenAI has officially entered the development of complete robot units, with prototype launch expected in 2027; Tesla announced at its Q3 earnings investor call that Optimus has started early production at the Fremont factory, with capacity expected to ramp up to 20,000 units per week (1 million units annualized) by the end of 2027 at the earliest, and will officially go on sale to the public before the end of 2027. The large-scale mass production phase of humanoid robots from 1 to N is gradually approaching.
2. How would you define this rally? Is it a switch in the main AI investment theme, or a phase of rotation driven by capital moving from high to low?
Zhang Xiaoguo, fund manager of the Truvalue Xingxuan Industrial Trends Mixed Fund: This rally is not a switch in the main AI investment theme, nor is it a simple phase of capital rotating from high to low. It is a continuation and evolution of the main AI investment theme. Recently, the AI hardware sector has undergone significant correction. As one of the most promising important AI application directions for the future, embodied intelligence is experiencing strong rebound in the sector against the backdrop of significant changes in the fundamentals of the physical AI and humanoid robot industries, which is entirely logical.
3. Among recent industry events such as OpenAI entering robotics, NVIDIA's GTC upgrading the physical AI toolchain, and Optimus ramping up production, have there been signals that truly change the pricing logic, rather than just sentiment catalysts?
Zhang Xiaoguo, fund manager of the Truvalue Xingxuan Industrial Trends Mixed Fund: This round of the embodied intelligence industry chain is not a simple sentiment catalyst. The pricing logic of the industry chain has undergone a fundamental change. The focus of sector pricing is shifting from "trading the process of humanoid robots going from 0 to 1" to "trading the process of embodied intelligence going from 1 to N." In the embodied intelligence sector represented by humanoid robots, the market's early focus was on the product development progress of leading humanoid robot brands such as Tesla's Optimus. The speed of product performance improvement (especially intelligence level) and mass production progress expectations were the center of market trading. During the 0-to-1 phase of the industry, sector performance had obvious characteristics of concept-driven investing. Recently, GPT-6 Astra proved the applicability of the scaling law in the physical AI field, providing a new paradigm for the rapid development of physical AI. It is expected to quickly resolve the bottleneck in the intelligence level of embodied intelligence, and the embodied intelligence industry is expected to enter a phase of accelerated development. The humanoid robot industry will officially usher in accelerated mass production, and embodied intelligence is transitioning toward the 1-to-N phase. The investment methodology for this sector is also shifting from "concept investing" to "industry investing."
4. At what stage of development is the physical AI industry currently? What is its allocation value?
Zhang Xiaoguo, fund manager of the Truvalue Xingxuan Industrial Trends Mixed Fund: Currently, physical AI is on the eve of large-scale validation, with industrial scenarios leading the way, entering a period of technology closed-loop validation. It is still far from mass consumer adoption, belonging to the late 0-to-1 phase and the early 1-to-10 phase of the industry. Multimodal VLA models and world models (WM) have already achieved convergence of technical routes in laboratories and small-scale settings, but there are still issues with insufficient general scenario capabilities. GPT-6 Astra proved that general foundation models can also have the ability to directly serve as policy models, which is expected to further accelerate the evolution of physical AI performance. At the fundamental level, physical AI specialized large model companies continue to make efforts, and general large model companies are entering the field one after another; data demand is surging. Leading robot companies both domestically and internationally are increasing their investment in physical AI model development. General AI large model giants such as OpenAI, NVIDIA, and Google are entering the physical AI field one after another. GPT-6 Astra achieved a leap in general model capabilities in physical AI by feeding EGO data, igniting the demand for data in embodied models. Currently, the physical AI industry is in a state where models are converging, data is scarce, investment is front-loaded, and performance is back-loaded. The industry is currently standing at the critical point of accelerated development, with high allocation value.
5. Which sub-sectors are you bullish on, and what is the main allocation strategy?
Zhang Xiaoguo, fund manager of the Truvalue Xingxuan Industrial Trends Mixed Fund: Bullish on three directions. First, the data collection industry chain, including data collection hardware (industrial cameras/CIS, IMU, force/tactile sensors, teleoperation equipment) and data services. Currently, embodied intelligence data is severely insufficient, and demand is exploding. Second, humanoid robot hardware components, including planetary roller screws, harmonic reducers, dexterous hands and micro motors, sensors, lightweight materials, etc. Particularly focus on leading component companies with deep cooperation with Tesla and leading domestic robot customers and strong competitive strength. Third, body/robot manufacturers. The rapid progress of physical AI is expected to bring rapid improvement in the product capabilities of embodied intelligence and release downstream demand, thereby accelerating the mass production of body manufacturers. In terms of allocation strategy, based on an industry perspective, deeply cultivate the industry, carefully select individual stocks, and strive to capture the industry beta investment opportunities brought by the development of the humanoid robot industry while pursuing higher alpha. Strongly bullish on key targets in the physical AI industry chain, the Tesla humanoid robot industry chain, and high-quality targets in the domestic humanoid robot industry chain with strong positioning advantages and competitiveness. Adopt a long-term mindset, closely follow the industry, and accumulate positions on dips, with firm confidence and active positioning.
6. Where is the biggest expectation gap in the physical AI industry currently? What potential risks should investors pay attention to?
Zhang Xiaoguo, fund manager of the Truvalue Xingxuan Industrial Trends Mixed Fund: The real-machine interaction data collection direction has the biggest expectation gap. Currently, the market generally recognizes the importance of data in the physical AI industry chain, but the current focus is on EGO data. However, high-quality real-machine interaction data is more important for the evolution of physical AI. Real-machine interaction data is multimodal data returned from robots executing tasks in the real physical world. In addition to first-person view operation-related data, it also includes data information related to cameras, depth, force control, touch, IMU, joint torque, end-effector status, etc. These are the highest quality and most valuable physical AI training data. The data collection direction has great potential. Risks to watch: changes in technical routes, technology development progress may fall short of expectations, downstream demand release pace may fall short of expectations, intensified industry competition, etc. MACD golden cross signals have formed, and these stocks are performing well!
Comments