Today (September 12), domestic AI chips experienced a major breakout! The STAR AI ETF (589520), which focuses on domestic AI industry chains, surged as much as 3.78% intraday and is currently up 2.8%, potentially achieving three consecutive daily gains! Real-time trading volume exceeded 60 million yuan, showing active participation. Notably, the ETF is trading at a significant premium with a real-time premium rate reaching 1.77%, indicating strong buying interest!

Among component stocks, VeriSilicon surged 20% to the daily limit, Cambricon rose over 8%, Yuntianlifei gained more than 3%, while Fudan Microelectronics and Chipanalog Technology also posted gains.

On the news front, The Information, citing four sources with direct knowledge, reported that Alibaba and Baidu have begun using self-designed chips to train their AI models, partially replacing chips produced by NVIDIA.

Industry experts highlight three strategic values of self-developed chips:

1. Cost Reduction and Dependency Elimination: The primary motivation for large model companies to develop their own chips is to reduce dependence on NVIDIA and control costs.

2. Performance Optimization and Customization: While general-purpose GPUs have broad applicability, they may not fully meet the optimization needs of specific large models. Self-developed chips can be deeply optimized according to their own model characteristics to improve computational efficiency.

3. Supply Chain Security and Policy Risk: Geopolitical factors are also important reasons for large model companies to consider self-developed chips. US restrictions on Chinese chip technology have made Chinese companies more focused on supply chain security.

Ping An Securities noted that against the backdrop of US restrictions on advanced chip exports to China and suppression of domestic computing chips and domestic large models, the urgency of domestic computing power substitution continues to rise. In the medium to long term, chip security issues will continue to be disrupted by political games, and China's independent design, tape-out, and mass production of high-end AI chips is an inevitable trend. Domestic computing chips have achieved good results in AI training and inference, combined with policy-level support, domestic computing power is expected to achieve large-scale deployment and application, bringing development opportunities to related industry chains.

Regarding limit-up stocks, VeriSilicon announced that as of the end of Q2 2025, outstanding orders reached 3.025 billion yuan, a historical high. From July 1, 2025, to September 11, 2025, the company signed new orders worth 1.205 billion yuan, a significant 85.88% increase compared to the entire third quarter last year, setting a new historical record for new orders, with AI computing-related orders accounting for approximately 64%. Besides new orders reaching historical highs, the company's outstanding orders continue to remain at high levels, which is expected to have a profound impact on the company's subsequent operating performance.

Soochow Securities believes that the previous surge was definitely not the end of the "domestic computing power" market trend, and the capital-level consolidation is to prepare for the next healthier rally. Recent performance has shown "startup" signs with active news and frequent industry developments: including expectations for GPU orders from major internet companies, continuously generating positive sentiment; Cambricon's private placement received CSRC approval; Hygon is advancing Sugon's M&A and released comprehensive equity incentives; Alibaba launched a "combination punch," following the release of the QwenMax Preview trillion-parameter high-end large model, continuing to strengthen its AI ecosystem, including signing strategic cooperation agreements with Honor and launching "Gaode Street Scanning List."

**Domestic Substitution Light, STAR Self-Reliance and Self-Improvement**

From the current standpoint, focus on three highlights of the STAR AI ETF (589520) and its feeder funds (Feeder A: 024560, Feeder C: 024561):

1. Policy Ignition, AI Takeoff: Top-level documents have ignited the market, with AI potentially becoming the leading sector throughout this market cycle. Edge-cloud integration is the core trend of AI development, with component stocks being companies with the largest revenue or best positioning in subdivided segments, benefiting from the accelerated AI transformation of edge-side chips/software.

2. Domestic Substitution, Independent Control: Against the backdrop of technological friction, the importance of information security and industrial security is highlighted. As a core technology, achieving independent control in artificial intelligence is crucial. The target index focuses on domestic AI industry chains with strong domestic substitution characteristics.

3. 20% High Elasticity, Strong Offense: Compared to directly investing in STAR Market individual stocks, ETFs can provide low-threshold positioning, and with 20% price limit restrictions, efficiency is higher when markets explode. The top ten holdings account for over 70% of the weight, with semiconductors as the largest weighted industry accounting for over half, showing high concentration and strong offensive capability.

Note: As of the end of August 2025, the top ten holdings of the STAR AI Index accounted for 71.66% of the weight; according to SW Level 2 industry classification, semiconductors are the largest weighted industry, accounting for 54.1% of industry weight.

Risk Warning: The STAR AI ETF and its feeder funds passively track the SSE STAR AI Index. This index has a base date of December 30, 2022, and was released on July 25, 2024. The composition of index components is adjusted in a timely manner according to the index compilation rules, and its backtested historical performance does not indicate future index performance. Individual stocks and index components mentioned in this article are for display purposes only. Individual stock descriptions do not constitute any form of investment advice, nor do they represent holding information and trading trends of any fund under the management company. The fund management company evaluates the risk level of the STAR AI ETF as R4-medium-high risk, suitable for aggressive (C4) and above investors. Please refer to sales institutions for appropriateness matching opinions. Any information appearing in this article (including but not limited to individual stocks, comments, predictions, charts, indicators, theories, any form of expression, etc.) is for reference only, and investors must be responsible for any independent investment decisions. In addition, any views, analyses, and predictions in this article do not constitute investment advice in any form to readers, nor do they bear any responsibility for direct or indirect losses caused by the use of this article's content. Fund investment involves risks. Past performance of funds does not represent their future performance. The performance of other funds managed by the fund management company does not constitute a guarantee of fund performance. Fund investment should be cautious.

MACD golden cross signal formed, these stocks are showing good upward momentum!