In the early hours of Friday, September 5th Beijing time, OpenAI officially released GPT-6 Astra, which the company describes as its most intelligent and best-aligned model to date. The new system has set new state-of-the-art benchmarks across computer usage, browsing, software engineering, cybersecurity, and professional work tasks. During the launch event, OpenAI President Greg Brockman declared, "Welcome to the age of AGI," signaling a shift where AI evolves from a question-and-answer tool into an autonomous agent capable of executing complex tasks.
This transition is expected to drive demand across the entire AI industrial chain, including computing power, hardware, enterprise software, and cybersecurity. Key highlights of GPT-6 Astra include training on a cluster exceeding 100,000 GPUs at the Stargate facility in Texas, marking OpenAI's largest training effort to date. Additionally, for the first time, AI systems assisted in training the model itself, enabling continuous autonomous operation for extended periods and representing a nascent form of recursive self-improvement. The model also demonstrates advanced cybersecurity capabilities, autonomously identifying unknown vulnerabilities in hardened systems and generating functional exploit code during internal testing, including two previously undiscovered zero-day flaws.
Impact on AI Chip and Computing Hardware Demand
Companies such as Cambricon Technologies, Hygon Information Technology, Loongson Technology, MetaX Integrated Circuits, VeriSilicon Microelectronics, Anlogic Technology, and Fudan Microelectronics are positioned at the forefront of this trend. The launch of GPT-6 Astra intensifies the global large-model arms race, expanding overall computing power requirements. With restrictions on high-end imported chips, the domestic supply-demand gap for computing capacity widens, strengthening the case for domestic AI chip volume growth. Faster model iteration also heightens the urgency for inference-side adaptation of domestic chips, reinforcing a coordinated push for "national models on national silicon."
Accelerating Commercialization for AI Applications and Software
Companies like Kingsoft Office, Transwarp Technology, CapeCloud, Neusoft Newpoint, Cloudwalk Technology, Intellifusion, IntSig Information, and UCloud are active in the AI application and platform layer. If GPT-6 Astra lowers model usage costs and enhances inference and multimodal capabilities, it could accelerate the commercial deployment of domestic AI applications. Scenarios such as office productivity, government services, data analytics, and cloud offerings may see increased willingness to pay and higher average revenue per user. Leading application vendors can leverage stronger model capabilities to upgrade products and unlock new revenue streams.
Advancing On-Device AI and IoT Hardware
Companies including Espressif Systems, Amlogic, Bestechnic, Orbbec, Ezviz, Roborock, Arashi Vision, Luxvisions Innovation, and Opt Machine Vision are focused on edge computing and machine vision. GPT-6 Astra reinforces the industry trend of combining cloud-based large models with on-device small models, creating opportunities for higher adoption rates of edge AI chips and smart hardware. Enhanced AI capabilities are poised to drive product upgrades and demand growth in smart homes, robotics, and industrial vision applications.
Strengthening Cybersecurity and Data Protection Needs
Security-focused companies such as Qi An Xin Technology, Anlogic Technology, and Geovis Technology stand to benefit from increased cybersecurity spending driven by AI-related threats. As models become more powerful, the risks of automated attacks and data breaches grow, elevating the urgency and priority of security procurement for both government and enterprise clients.
The Case for Domestic AI Industry Leadership
The Huabao STAR AI ETF (589520) and its feeder funds (Class A: 024560, Class C: 024561) offer targeted exposure to this domestic AI supply chain. The ETF tracks an index consisting of 30 large-cap STAR Market companies providing foundational resources, technologies, and applications for AI. Semiconductor companies account for 70.4% of the index weight, with GPU-related and AI application concepts representing 41.98% and 23.19%, respectively. With a 20% daily price fluctuation limit, the ETF provides accessible exposure to the growth of the sci-tech innovation sector. As a margin trading target, it serves as an efficient tool for one-click investment in domestic computing power infrastructure.
Note: GPU concept and AI application concept weightings are compared against the GPU Index (8841701.WI) and AI Application Index (980112.CNI). Data sourced from Shanghai and Shenzhen exchanges, as of September 4, 2026. As of the end of August, the top component weights include Hygon Information Technology at 10.36%, VeriSilicon Microelectronics at 9.97%, Montage Technology at 9.69%, Cambricon Technologies at 9.43%, Kingsoft Office at 9.05%, Amlogic at 6.46%, Roborock at 4.14%, Fudan Microelectronics at 3.40%, Loongson Technology at 3.39%, and Orbbec at 3.19%.
The individual stocks mentioned herein are components of the underlying index for the Huabao STAR AI ETF and are shown for illustrative purposes only. Their descriptions do not constitute investment advice in any form and do not represent the holdings or trading activities of any fund managed by the firm. Regarding fees, the ETF does not charge a sales service fee, and authorized brokers may charge a commission of up to 0.5%, which includes fees levied by stock exchanges and registration institutions. For the feeder fund, Class A shares impose a subscription fee of RMB 1,000 for amounts of RMB 2 million or more, 0.6% for amounts between RMB 1 million and RMB 2 million, and 1% for amounts below RMB 1 million, with a redemption fee of 1.5% for holdings under 7 days and 0% for holdings of 7 days or more. Class C shares charge no subscription fee, a redemption fee of 1.5% for holdings under 7 days and 0% thereafter, and a sales service fee of 0.3%.
Risk Disclosure: The Huabao STAR AI ETF passively tracks the STAR AI Index, which has a base date of December 30, 2022, and was published on July 25, 2024. The index's constituent stocks are adjusted in accordance with its compilation rules, and historical backtested performance does not indicate future index performance. The fund manager has assessed the risk level of the Huabao STAR AI ETF as R4 (medium-high risk), suitable for active (C4) and above investors. Any information presented in this article, including but not limited to individual stocks, commentary, forecasts, charts, and indicators, is for reference only. Investors are solely responsible for their own investment decisions. Additionally, any views, analyses, or forecasts herein do not constitute investment advice to readers, and the fund manager assumes no liability for any direct or indirect losses resulting from the use of this content. Fund investment involves risks; past performance of the fund does not represent its future performance, and the performance of other funds managed by the fund manager does not constitute a guarantee of fund performance. Investors should invest with caution.
Comments