Expectations for production expansion have strengthened.
On Tuesday, China's A-share market trended lower with fluctuations. The Shanghai Composite Index closed down 0.81%, the ChiNext Index fell by 1.16%, while the STAR 50 Index rose by 0.29%.
The high demand for computing power and storage from AI and other sectors has driven rapid expansion in advanced process capacity, directly boosting global silicon wafer shipments and capital expenditures by wafer fabrication plants.
As third-quarter earnings continue to be disclosed, expectations for semiconductor production expansion next year are rising based on the current level of tight supply.
Despite the differentiated performance structure within the semiconductor sector since October, the fundamental logic supporting the long-term development of domestic semiconductors remains unchanged.
Driven by AI and localization substitution as dual main themes, the semiconductor industry chain is witnessing structural opportunities characterized by 'capacity expansion' and 'supply chain security'.
01
Comprehensive price increases for AI-driven storage solutions
In terms of sector performance, there was a divergence within technology stocks. AI applications and the semiconductor sector showed gains, while the lithium battery industry chain led declines. Similarly strong performers such as steel, chemicals, coal, and non-ferrous metals also saw pullbacks recently.
Specifically, AI application themes continued to strengthen, with Rongji Software achieving a two-day limit-up, and Fullgoal Holdings and Shiji Information hitting the daily price cap. Other stocks such as Zhixing Technology also surged significantly.
Yesterday, Alibaba officially announced the 'Qwen' project, launching the public beta version of the Qwen app. Adopting a 'free and accessible to everyone' approach, it is making an all-out push into the AI-to-consumer market. The competition surrounding personal AI assistants has quietly intensified. As of now, the Qwen app rapidly climbed to fourth place on Apple's App Store free apps overall ranking list on the second day of its public beta release, surpassing DeepSeek in ranking.
Moreover, although concerns over a tech bubble in U.S. equities have been intensifying recently, news over the weekend that Berkshire Hathaway initiated a position in Google has also gained attention. Google is expected to become a new challenger to NVIDIA’s current market capitalization, with its Gemini 3.0 large model set to be released this week. Analysts believe that the business model for AI applications is rapidly transitioning from proof-of-concept to revenue generation, and demand for domestically produced AI applications is expanding at a high speed.
As the upstream sector of the AI industry chain, semiconductors, which had been undergoing adjustments in recent months, are showing signs of recovery. Longxun Co. achieved a 20% price increase limit, while Amlogic and Dongxin Co. rose more than 10%. Companies like Advanced Micro-Fabrication Equipment (AMEC), NAURA Technology Group, and Hua Hong Semiconductor also rebounded strongly, driving the E Fund Semiconductor Equipment ETF (159558) up by 2.1%, and the STAR 50 ETF (588080) increased by 0.36%.
On the news front, the super logic behind memory chips has been further reinforced. According to Reuters, Samsung Electronics raised the prices of its server DDR5 memory chips this month due to supply shortages amid a global surge in AI data center construction. The increased chip prices rose between 30-60% compared to September levels.
It’s not just Samsung taking this action. In addition to DRAM, NAND flash memory prices are also surging sharply, resulting in a comprehensive price hike across the board.
Previously, major NAND flash manufacturers including Samsung Electronics, SK Hynix, and Kioxia implemented production cut strategies in the second half of the year, driving NAND prices higher. Currently, Samsung is negotiating next year's supply volume with overseas major clients and internally considering raising prices by more than 20-30%.
According to analysts, severe downturns in parts of the memory industry during 2023 and 2024 led to insufficient investment. Although new capacity construction has begun, it will take time for production to come online, meaning supply constraints may persist longer. More concerning, if high prices continue, consumers may bear the cost of memory shortages, with prices for products ranging from smartphones to automobiles facing upward pressure.
As semiconductor companies disclosed their Q3 earnings, the capacity utilization rate of domestic wafer manufacturers has risen to a high level, indicating robust industry demand. A significant number of urgent orders for analog and memory chips have also been received.
The core bottleneck behind this wave of memory chip price hikes lies in the supply-side capacity gap. Under the mid- to long-term AI-driven trend, the domestic substitution of computing power and high-performance memory still requires time to catch up. Capital expenditures are expected to grow at an exceptionally steep rate, driving continuous growth in equipment company orders.
Driven by the dual themes of AI and domestic substitution, the semiconductor industry chain is witnessing structural opportunities characterized by 'capacity expansion' and 'supply chain security.'
02
Expectations for production expansion have strengthened.
Since the fourth quarter, the overall performance of the semiconductor sector has been lackluster, but the fundamental logic supporting the long-term development of the semiconductor sector remains unchanged.
On the demand side, high computing and storage demands driven by AI are propelling rapid expansion in advanced process capacity, directly boosting global silicon wafer shipments and fab capital expenditures.
Taking memory chips as an example, Samsung's price increase for DDR5 clearly reinforces the supercycle logic of memory. The surge in storage demand brought about by AI training and inference is driving explosive growth in demand for the most advanced DRAM, HDDs, and high-performance SSDs. In this process, companies like Winbond and Nanya Technology, which hold a majority share of DDR4 products in niche markets, are also experiencing 'demand spillover.'
Amid sustained AI-driven demand and rising memory prices, the supply-demand dynamics in the memory industry may persist. To meet the growing storage needs, manufacturers may increase capital expenditures, thereby driving up demand for semiconductor equipment.
As China’s leading NAND flash memory chip manufacturer, Yangtze Memory Technologies Co., Ltd. (YMTC) has built a third factory in Wuhan, which is expected to commence operations in 2027, while also expanding the capacity of its second factory. Research firm Omdia estimates that YMTC's capital expenditure in 2025 will be more aggressive than its global peers, accounting for approximately 20% of the total global NAND flash investment, with further acceleration expected.
Last week, SMIC, the leader in A-share wafer foundries, also disclosed its third-quarter report. According to the announcement, SMIC achieved revenue of 17.162 billion yuan in the third quarter of 2025, representing a year-on-year increase of 9.90%; net profit attributable to shareholders of listed companies was 1.517 billion yuan, an increase of 43.10% year-on-year.
Notably, SMIC stated in its investor relations activity record that its production lines are actually running very full, with capacity utilization reaching 95.8% in the third quarter. One reason for the lack of a significant jump in fourth-quarter guidance is that the mobile phone market is currently experiencing a particularly acute shortage of memory, and prices have risen sharply.
Secondly, an imbalance of just 5% in the current memory market supply and demand could trigger a doubling of price fluctuations. However, even if new manufacturers want to enter the market, products such as NOR Flash, NAND Flash, and MCUs require at least 16 months from tape-out to mass production. This means that in the foreseeable future, the market positions of existing suppliers will remain stable and difficult to quickly replace.
For foundry enterprises, the storage cycle has a positive impact. In addition to SMIC, Huahong's capacity utilization rate in the third quarter even reached 109.5%, and the capacity of factories under construction is also accelerating its ramp-up.
Benefiting from the opportunities brought by the replacement of the industrial chain, the company’s customers have increased their terminal shares in areas such as analog, MCU, memory, CIS, and display drivers, which has also brought more foundry business to SMIC. If this part of the demand grows more optimistically, it is expected to drive the company to accelerate its capacity expansion pace.
In addition, Changxin Technology has issued a completion report for its IPO tutoring process, with expectations for memory expansion in 2026.
The capacity construction of downstream wafer fabs brings high-certainty growth opportunities to upstream equipment, components, and material sectors. At the same time, achieving domestic control over core equipment and materials upstream is crucial for meeting domestic computing and storage demands, as true capacity release requires all machines in the entire production line to be fully equipped.
The logic of domestic substitution remains a key investment theme in the semiconductor sector, widely supported by the market. Investing in domestic semiconductor substitution equates to investing in leading companies in critical segments. This year, the annual increase of the Semiconductor Equipment ETF (159558) managed by E Fund reached 45.98%.
The Semiconductor Equipment ETF (159558) managed by E Fund aligns closely with the domestic substitution theme, tracking an index focused on two major 'chokepoint' themes in semiconductors—semiconductor equipment (60.9%) + semiconductor materials (22.5%). It is regarded by the market as playing the role of 'selling picks' in the industry chain. The constituent stocks cover key segment leaders such as lithography machines, etching machines, thin-film deposition equipment, silicon wafers, and photoresists, with nearly 30% weight attributed to AMEC and NAURA.
From a performance perspective, the sustained release of domestic AI computing power demand has gradually begun to marginally reflect in the earnings of companies within the chip supply chain. For instance, Cambricon’s revenue surged nearly 24-fold in the first three quarters, with its attributable net profit regaining growth momentum. The company invested RMB 720 million in research and development, accounting for approximately 15.51% of its total revenue.
Revenue growth among domestic semiconductor equipment companies in the third quarter was equally impressive. Companies such as NAURA Technology, Advanced Micro-Fabrication Equipment (AMEC), and Topcon Science & Technology showed notable quarter-on-quarter increases, positioning them to benefit from the expansion of advanced logic and memory production lines next year.
03
Epilogue
After more than a month of pullback, the semiconductor sector, bolstered by high growth momentum and fundamental support, is now positioned within an expectation window looking ahead to AI-driven demand and capacity expansion in the coming year.
Since October, there has been significant divergence in investment within the technology growth sector, including within semiconductors themselves. As a result, the A-share semiconductor index has underperformed compared to the strong performances of the Philadelphia Semiconductor Index and Taiwan Semiconductor Index.
On one hand, capital has focused more on AI storage and electric power themes that provide higher marginal information, such as memory chips, gas turbines, solid-state transformers, and energy storage. Meanwhile, the domestic computing power market continues to await new catalysts in the short term.
On the other hand, the return of A-share technology growth style is highly correlated with the spillover effect of U.S. stock AI technology and marginal changes in liquidity. The uncertainty surrounding the Federal Reserve's interest rate cut in December poses resistance to a full follow-through in the A-share market.
The return of A-share technology style is only a matter of time. Moving forward, the signals released by the earnings reports of tech giants such as NVIDIA, Alibaba, and Baidu will serve as a direct response to the 'U.S. stock AI bubble theory' and also lay the groundwork for subsequent rebounds in A-share technology stocks.
The catalytic effects brought about by downstream domestic AI applications are also expected to become new 'ignition points.'
In addition to the Qwen app, recent updates to domestic models (MiniMax M2/GLM-4.6/Kimi K2-thinking) have been released intensively, with some players already preparing for their initial public offerings. Moreover, artificial intelligence is expected to be included as an independent industry in the five-year planning scheme for the first time early next year, further elevating its strategic importance.
In addition to the Semiconductor Equipment ETF E Fund (159558), attention can also be paid to the STAR 50 ETF (588080), which has achieved a year-to-date increase of 37.56%. Last Friday, it attracted 2.39 billion yuan against the market trend, with its latest scale reaching 69.702 billion yuan.
The STAR 50 Index it tracks is a typical broad-based index for hard-core technology, with semiconductors accounting for a weight of 65.5%. Its constituent stocks include leading companies such as Cambricon, Hygon Information Technology, SMIC, Montage Technology, and Advanced Micro-Fabrication Equipment. The index features a price fluctuation limit of 20%, high elasticity, and a comprehensive fee rate of 0.2%, providing investors with a convenient tool to deploy leading STAR Market companies at low cost.
(End of article)
