NVIDIA GTC and Tesla Earnings Converge as A-share AI Narrative Enters Industry Verification Window

Recently, several major events converged in the same window—NVIDIA's GTC Berlin conference and Tesla's third-quarter earnings call, among others—bringing multiple catalysts to the AI supply chain in A-shares. Wind data shows that the Wind AI Concept Index has risen more than 2% cumulatively since October 19.

Analysts believe that NVIDIA's system-level layout for AI factories and physical AI, combined with Tesla's commercialization push in Robotaxi and humanoid robots, is strengthening the AI industry logic from both ends—computing infrastructure and scenario deployment—and supporting segments in the A-share supply chain such as liquid cooling, high-speed interconnect, and robot actuators are expected to benefit from event-driven catalysts. However, whether this rally can shift from event-driven to fundamentals-driven still depends on order visibility and the pace of scenario fulfillment.

On October 21 Beijing time, NVIDIA's GTC Berlin conference opened, and NVIDIA CEO Jensen Huang delivered a keynote speech, showcasing the company's latest layout in areas such as AI infrastructure, agents, open models, and physical AI.

Unlike past events centered on chip launches, this speech focused on standardized delivery of AI factories and deployment of physical AI scenarios. NVIDIA announced the DSX POD reference design, allowing partners to complete system integration and validation inside modular containers; once on-site utilities are ready, AI factory deployment cycles are expected to be shortened from about six months to 14 days. Huang said the DSX AI factory platform can reduce cooling water use to near zero and cut energy consumption by up to 30% while achieving the same AI output. In robotics, NVIDIA launched the Agile ONE SN reference humanoid robot platform based on Isaac GR00T, with 31 degrees of freedom, expected to launch in 2027, aiming to lower the barrier to robot development.

Shortly afterward, Tesla disclosed its third-quarter financial results and held a conference call on October 22 Beijing time. Tesla CEO Musk said the Robotaxi service is continuing to expand its operating scope and plans to launch the service in Dallas next month. He expects Robotaxi's contribution to gradually emerge starting in the second quarter of 2027 and possibly become an important business in the third or fourth quarter of 2027. On humanoid robots, Musk cautioned the public not to expect large-scale deliveries too early, but confirmed that Tesla has begun early production preparation work for the Optimus high-capacity production line at its Fremont, California factory, and expects to begin deliveries to companies other than Tesla by the end of 2027. Tesla also revealed that the Cybercab has about 50% fewer parts than other models Tesla has previously produced, which helps reduce manufacturing complexity.

Deep shift in industry logic

Behind the catalyst events, the narrative focus of the AI industry is undergoing a noteworthy migration. Previously, market pricing of AI mainly revolved around cloud computing expansion, with core variables being GPU shipments, data center capital expenditure, and optical module speed upgrades. The signals released by NVIDIA and Tesla this time have shifted part of the pricing anchor toward the link between AI and interaction with the physical world.

In his speech, Huang likened the AI industry to a system composed of five layers: energy, accelerated computing, AI factories, AI models, and digital intelligence embedded in various applications. He said that before AI models emerged, infrastructure had to be in place first, and as AI applications expand, power supply, data center construction, and computing deployment efficiency are becoming key factors in industry development. This statement means that NVIDIA's competitive boundary is extending from GPU chips to a full-stack ecosystem composed of hardware, software, models, and industry applications, and the core of its business model is tilting from "selling computing power" to "selling AI production capacity."

In physical AI, NVIDIA demonstrated agents' applications in industrial design, laboratories, and factories: in Siemens factories, agents test production plans for new products through digital twins; in Dassault Systèmes' engineering design scenarios, agents use simulation to identify vehicle design issues and optimize solutions. In autonomous driving, Volvo Cars is adopting NVIDIA's Hyperion computing platform and Alpamayo software stack, and next-generation models from Volkswagen Commercial Vehicles, Jaguar Land Rover, and Mercedes-Benz will be compatible with the Hyperion platform.

This narrative shift is not limited to the corporate level. The AI Research Institute of the China Academy of Information and Communications Technology recently released the "AI Native Research Report (1.0)" (2026), summarizing the current stage as a paradigm shift from "internet native" and "cloud native" to "AI native," and defining "AI native" as systems, products, or organizations that take AI as their core operating mechanism and source of value from the outset of design, rather than layering AI capabilities onto existing forms. The report also distinguishes between two forms, "+AI" and "AI native": "'+AI' adds recognition, recommendation, or Q&A steps on top of existing processes, while the processes themselves continue to operate as before"; "AI native" requires that "artificial intelligence be intrinsic to architecture and processes, and that existing forms be reconstructed because of intelligence." This judgment is consistent with NVIDIA's shift from "selling computing power" to "selling AI production capacity" and with Tesla's change in direction of pointing to AI architecture as the bottleneck for robots; the focus of industry competition is shifting from "whether to connect to AI" to "whether AI can be embedded as a core component of system operation."

The changes on Tesla's side are equally telling. On the earnings call, Musk explicitly attributed Optimus's current main bottleneck to the artificial intelligence architecture rather than external data collection. He said Tesla hopes to first produce about 10,000 to 20,000 robots for large-scale autonomous training, and is building a "robot academy" where robots practice a variety of tasks and feed the training results back to the model. This statement reveals that the industrialization bottleneck of humanoid robots is converging from hardware manufacturing toward "brain" capabilities, and this is precisely the core link that AI cannot bypass when extending from the digital world to the physical world.

Hua Xiaowei, chief analyst of new-quality small and mid-caps at Guosheng Securities, said that the market pricing logic of the AI industry will switch to fundamental verification. With the concentrated disclosure of third-quarter reports, the AI industry chain is entering an important window for testing its prosperity, and stocks with earnings exceeding expectations are expected to become the structural main line during market adjustments.

Order visibility and scenario fulfillment may become core variables

Facing the switch in AI industry logic, several brokerage analysts have provided investment strategy frameworks from different dimensions. Unlike the earlier recommendation logic that simply revolved around computing power elasticity, the current investment foothold is more concentrated on earnings verifiability and the pace of order fulfillment in segmented links.

Xiao Qunxi, chief analyst of the machinery industry at Guotai Haitong Securities, made specific judgments from the perspective of the robot industry chain. She said that with the continuous iteration of large models, motion control and perception technologies, the robot industry is upgrading from single-machine automation to intelligence and flexibility. Humanoid robots are expected to gradually move from technical verification to industrial scenario applications, driving demand growth for core components and intelligent manufacturing equipment.

Liu Chenming, chief strategy analyst at GF Securities, provided cross-validation from the perspective of export chain prosperity. Among the high-prosperity directions for the third quarter that he sorted out, AI hardware involves memory chips, optical modules, optical chips, optical cables, PCBs, electronic cloth, MLCCs, and liquid cooling; electromechanical equipment involves gas turbines, industrial robots, freight containers, and others. Liu Chenming believes that AI hardware remains the direction with the highest prosperity, while non-AI categories show divergence, with pace changes mainly stemming from factors such as base effects and working-day misalignment; the policy window may be concentrated in November, and focus can be placed on tracking overseas cloud vendors' capital expenditure guidance in late October.

Fang Zhengtao, a strategy analyst at Huatai Securities, judged from the perspective of trading structure that since mid-September, the performance divergence of global AI assets has intensified, and market focus has switched from demand certainty to return elasticity; capital expenditure may no longer be the single pricing anchor for AI assets. Current AI trading may shift from the extensive allocation stage of piling up computing power to the stage of carefully selecting individual stocks along industry bottlenecks and cash flow certainty. It is recommended to focus on directions with high earnings visibility in the third-quarter reports, such as PCBs, servers, liquid cooling, and power supplies.