Analysis of the Impact of Post-Moore Era Computing Power Breakthroughs on A-Shares Semiconductor and AI Computing Power Industry Chain Investments
Unlock More Features
Login to access AI-powered analysis, deep research reports and more advanced features
About us: Ginlix AI is the AI Investment Copilot powered by real data, bridging advanced AI with professional financial databases to provide verifiable, truth-based answers. Please use the chat box below to ask any financial question.
Related Stocks
I will now provide you with a systematic, comprehensive, and professional investment analysis report.
On January 9, 2026, a research team from Peking University published an important research achievement in Nature Electronics [1]. The team, composed of Researcher Tao Yaoyu from the Institute of Artificial Intelligence and Professor Yang Yuchao from the School of Integrated Circuits, created an innovative
This technological breakthrough’s core innovations include:
- Device Innovation: System integration of “volatile vanadium oxide devices” and “non-volatile tantalum/hafnium oxide devices” under the multi-physics domain fusion architecture
- Computing Paradigm Breakthrough: Enables multiple computing methods to operate in their suitable physical domains (current, charge, light, etc.), significantly improving computing efficiency
- Application Scenario Expansion: Applicable to cutting-edge fields such as embodied intelligence, edge sensing, neuromorphic computing, and communication systems
Even more noteworthy is that the research team led by Researcher Sun Zhong from Peking University successfully developed a
This achievement marks:
- A “century-old problem” that has plagued the global scientific community has been solved
- Analog computing with both high precision and scalability has become a reality
- Has revolutionary application prospects in fields such as 6G communications, large AI model training, and edge computing
The core characteristic of the post-Moore era is that traditional silicon-based devices are approaching physical limits. New devices represented by memristors and optoelectronic devices are regarded as the hope to break the dilemma of computing power and energy efficiency due to their unique computing performance [1]. However, these new devices can only support a single computing method, making them unable to adapt to the diverse computing needs in practical applications, which severely restricts the improvement of computing power and efficiency. The technological breakthrough by Peking University just solves this core bottleneck, providing an innovative path for the expansion of the operator spectrum of post-Moore new devices.
The computing power revolution in the post-Moore era will reshape the entire semiconductor industry chain, and a relatively complete investment map has been formed in the A-shares market:
| Industry Chain Segment | Core Segment | Key Listed Companies | Beneficiary Logic |
|---|---|---|---|
Computing Power Chips |
AI Training/Inference Chips | Hygon Information, Cambricon, Loongson Technology | Surging demand for domestic substitution |
Optical Modules |
High-Speed Optical Communication Modules | Zhongji Innolight, NeoPhotonics, Accelink Technologies | Rigid demand for AI computing power interconnection |
Advanced Packaging |
2.5D/3D Packaging, Chiplet | Changjiang Electronics Technology (JCET), Tongfu Microelectronics, Huatian Technology | Breaking process limitations |
Servers/Infrastructure |
AI Servers, Liquid Cooling Systems | Dawning Information Industry, Inspur Information, Foxconn Industrial Internet | Explosive demand for computing power infrastructure |
Storage and Interfaces |
Memory Interfaces, Memory Chips | Montage Technology, GigaDevice Semiconductor | Upgrade of data transport capacity |
According to Frost & Sullivan forecasts, by 2029, the scale of China’s AI chip market will surge from RMB 142.537 billion in 2024 to
| Company | 2025H1 Revenue | YoY Growth | 2025H1 Net Profit | YoY Growth | Key Highlights |
|---|---|---|---|---|---|
Hygon Information |
RMB 5.464 billion | +45.21% | RMB 1.201 billion | +40.78% | DCU matches international mainstream products; dual-drive of CPU + DCU |
Cambricon |
RMB 2.881 billion | +4347.82% | RMB 1.038 billion | Turned profitable | Officially escaped years of losses; leading domestic AI chip provider |
Loongson Technology |
RMB 244 million | +10.9% | RMB 294 million loss | Narrowed loss | Recovery of traditional security application market; launch of “Three Musketeers” products |
Hygon Information is a leading enterprise in the field of domestic high-end processor R&D, with a product roadmap similar to NVIDIA and AMD:
- DCU Products: Based on a “CUDA-like” general-purpose parallel computing architecture, with low migration costs for CUDA users and operator coverage exceeding 99%. It has established in-depth cooperative relationships with leading internet companies such as ByteDance, Tencent, Alibaba, and Baidu, achieving full adaptation with mainstream domestic large models
- Product Iteration: The Deep Computing No. 3 was launched in 2025, and the R&D of Deep Computing No. 4 is progressing smoothly
- Strategic Integration: Business integration with Dawning Information Industry to promote systematic collaborative development of “hardware - software - ecosystem”
Cambricon achieved a historic breakthrough in its first-half 2025 performance:
- Revenue increased by 4347.82% year-on-year, mainly driven by market expansion facilitating AI application implementation
- Contract liabilities surged by 61,223.22%, indicating sufficient order reserves
- Inventory reached RMB 2.69 billion, accounting for 31.95% of total assets, with strategic stockpiling to prepare for explosive demand
Optical modules are core components of AI computing power infrastructure, and the market showed a trend of both volume and price growth in 2025:
| Company | 2024 Revenue | YoY Growth | 2024 Net Profit | YoY Growth | Gross Margin |
|---|---|---|---|---|---|
Zhongji Innolight |
RMB 23.862 billion | +122.64% | RMB 5.171 billion | +137.93% | 33.8% |
NeoPhotonics |
RMB 8.647 billion | +179.15% | RMB 2.838 billion | +312.26% | 44.72% |
Zhongji Innolight, relying on technological leadership and market dominance, has become one of the core global suppliers of AI computing power infrastructure:
- Technology Layout: Has R&D capabilities for 3.2T products, and is actively deploying cutting-edge directions such as LPO/LRO, CPO, and Optical Circuit Switches (OCS)
- Production Capacity Guarantee: 1.6T optical modules began mass shipments in the second quarter, and the Thailand factory’s production capacity supports zero-tariff delivery to North America
- Market Position: Its stock price surged from RMB 66.70 per share to a high of RMB 448.00 per share in the last 5 months, with its market capitalization once approaching RMB 500 billion
NeoPhotonics focuses on the niche market of short-distance high-speed transmission in data centers:
- Technology Route: Bets on LPO technology, sacrificing transmission distance to reduce costs and improve efficiency, and has obtained NVIDIA’s exclusive LPO certification
- Cost Advantage: Vertical supply chain integration reduces costs by approximately 30%, with a gross margin of 44.72% significantly higher than that of Zhongji Innolight
- Growth Potential: Net profit growth rate of 312.26%, the highest in the optical module sector
According to Yole Group forecasts, the global advanced packaging market will exceed
| Company | Technological Advantages | 2025H1 Performance Highlights |
|---|---|---|
Changjiang Electronics Technology (JCET) |
XDFOI Chiplet Platform, 2.5D/3D Heterogeneous Integration | Automotive electronics business grew 34.2% YoY; industrial and medical businesses grew 38.6% YoY |
Tongfu Microelectronics |
Deeply bound to AMD, mass production of large-size FCBGA | Revenue reached RMB 13.038 billion, up 17.67% YoY; net profit reached RMB 412 million, up 27.72% YoY |
Huatian Technology |
Advanced packaging technologies such as SiP, FC, TSV, WLP | Advanced packaging accounts for over 50% of revenue |
JCET, with its independently developed “XDFOI Chiplet Platform” as the core, has built an advanced packaging system covering multi-dimensional heterogeneous integration scenarios such as 2.5D and 3D:
- Made breakthroughs in key technologies such as glass substrates, CPO photoelectric co-packaging, and large-size FCBGA in the first half of 2025
- The main construction of the automotive-grade chip packaging and testing factory in Lingang, Shanghai has been completed
Tongfu Microelectronics, relying on its in-depth cooperative relationship with AMD, has unique advantages in advanced packaging layout for AI and HPC fields:
- Large-size FCBGA has entered mass production, and extra-large-size FCBGA has passed engineering verification
- Solved the problems of product warpage and heat dissipation for extra-large-size products
Montage Technology is one of only three companies worldwide capable of mass-producing server memory interface chips. In 2024, it ranked as the world’s largest memory interconnect chip supplier by revenue, with a market share of
| Financial Indicator | 2024 | YoY Growth | 2025Q1 | YoY Growth |
|---|---|---|---|---|
| Operating Revenue | RMB 3.639 billion | +59.20% | - | - |
| Net Profit Attributable to Parent | RMB 1.412 billion | +213.1% | - | - |
| New Products such as PCIe Retimer | RMB 422 million | 8-fold growth | RMB 135 million | +155% |
- Increasing DDR5 Penetration: The penetration rate of DDR5 in server DRAM reached 40% in 2024, and is expected to exceed 60% in 2025
- Explosive Demand for AI Transport Capacity: Demand for PCIe Retimer chips has surged, with an 8-GPU AI server requiring 8-16 chips
- CXL Memory Expansion: Test data shows that GPU utilization can be increased by 72%, and inference time can be reduced by 35%
According to statistical data, at least
- Regional Distribution: Covers over 87 prefecture-level cities in 29 provinces and cities, with the densest projects in Beijing, Xinjiang, Shanghai, Anhui, Ningxia, and Sichuan
- Construction Model: Prefabricated and modular construction methods are accelerating in popularity, reducing the 1-2 year construction cycle to within six months
- Liquid Cooling Technology: Has shifted from an “optional” to a “mandatory” option, aligning with the trend of green and efficient computing power
In the first half of 2025, 23 listed companies in edge-side AI chips achieved a total revenue of
| Company | Jan-Sep 2025 Revenue | Jan-Sep 2025 Gross Profit | R&D Expense Ratio |
|---|---|---|---|
Amlogic |
RMB 5.071 billion | RMB 1.883 billion | - |
Ingenic Semiconductor |
RMB 3.437 billion | RMB 1.175 billion | - |
Rockchip Semiconductor |
RMB 3.141 billion | - | - |
Bluetrum Technologies |
- | RMB 1.130 billion | - |
In 2025, China achieved a series of major breakthroughs in the integrated circuit field [3]:
| Breakthrough Field | Institution | Significance of the Achievement |
|---|---|---|
| ISSCC 2026 | Tsinghua University (18 papers, world’s first) | Global leader in integrated circuit design |
| 2D Semiconductors | Fudan University (“Wuji” Chip) | World’s first 32-bit RISC-V architecture microprocessor |
| Analog Computing | Peking University | Precision increased to 24 bits, energy efficiency 100 times higher than GPUs |
| Quantum Computing | University of Science and Technology of China (USTC) | “Zu Chongzhi 3” set a new record |
| Optical Computing | Shanghai Jiao Tong University | All-optical large-scale semantic generation chip |
Core Targets: Hygon Information (dual-drive of CPU + DCU), Cambricon (leading AI chip provider that turned profitable), Loongson Technology (self-controllable CPU)
Core Targets: Zhongji Innolight (leading optical module provider, 1.6T products in mass shipment), NeoPhotonics (exclusive NVIDIA LPO certification for LPO technology), Foxconn Industrial Internet (leading AI server foundry)
Core Targets: Changjiang Electronics Technology (JCET) (leading advanced packaging provider), Tongfu Microelectronics (benefiting from AI via partnership with AMD), Huatian Technology (advanced packaging accounts for over 50% of revenue)
Core Targets: Montage Technology (world’s top memory interface chip provider), GigaDevice Semiconductor (multi-category layout of memory chips)
| Investment Style | Strategy Recommendations | Focus Targets |
|---|---|---|
Aggressive |
Focus on core targets benefiting from technological breakthroughs | Cambricon, Zhongji Innolight, NeoPhotonics |
Conservative |
Allocate to industry leaders and high-certainty opportunities | Hygon Information, Changjiang Electronics Technology (JCET), Montage Technology |
Long-Term |
Focus on targets with strong technical reserves and reasonable valuations | Loongson Technology, Huatian Technology, Accelink Technologies |
- Downward pressure on the global macroeconomy may affect capital expenditures
- Risk of cyclical fluctuations in the semiconductor industry
- Short-term stock price volatility caused by market sentiment fluctuations
- Competitive pressure from international giants (NVIDIA, AMD, etc.)
- Uncertainty in technological routes (e.g., evolution of CPO vs LPO technologies)
- Supply chain risks caused by restrictions on advanced manufacturing processes
- Restrictions on the procurement of advanced equipment and materials
- Production capacity expansion progress falls short of expectations
- Risk of high customer concentration
The breakthrough in post-Moore era computing architecture published by Peking University in Nature Electronics marks China’s important position in the global computing power revolution. This breakthrough will reshape the investment pattern of the A-shares semiconductor and AI computing power industry chain from the following dimensions:
-
Technology Spillover Effect: The multi-physics domain fusion computing architecture will accelerate the industrialization of new devices, benefiting upstream segments such as new materials, device preparation, and EDA tools
-
Application Scenario Expansion: Emerging scenarios such as embodied intelligence, edge sensing, and neuromorphic computing will bring incremental demand for computing power, benefiting niche tracks such as edge-side AI chips and sensor chips
-
Increased Value of Advanced Packaging: As a core path to break process limitations, technologies such as 2.5D/3D packaging and Chiplet will receive more capital attention
-
Upgrade of Computing Power Infrastructure: Supporting segments such as liquid cooling and modular data centers will see deterministic growth
From the perspective of industry trends, 2025 is ushering in a historic investment opportunity driven by the resonance of three cycles: demand, production capacity, and inventory. As the main investment line for the chip design sector in 2025, the cloud-side AI computing power chip market has broad space. Driven by the dual engines of independent innovation and AI, the A-shares semiconductor and AI computing power industry chain will usher in historic investment opportunities.
[1] Xinhua News Agency - Let New Devices “Run”: Chinese Scientists Create a New Computing Architecture to Improve Computing Power (https://www.news.cn/tech/20260110/3a3516486b154f01b1eb66308289b630/c.html)
[2] Xinhua News Agency - Breaking the Century-Old Problem of Analog Computing: High-Precision Scalable Analog Matrix Computing Chip Comes Out (http://www.news.cn/auto/20251017/843e5e268856424f98d88871b5d4ce46/c.html)
[3] JW Insights - Top 10 Breakthroughs in China’s Integrated Circuits in 2025 Open Up a Multi-Paradigm Era in the Post-Moore Era (https://jiweipreview.laoyaoba.com/html/news/newsdetail?news_id=971852)
[4] Eastmoney - Computing Power Chip Concept Rebounds Volatilely, Hygon Information Rises Over 7% (https://finance.eastmoney.com/a/202511063556836557.html)
[5] Securities Times - 2025 Outlook for the Semiconductor Industry: Dual Engines of AI and Independent Innovation Drive Industrial Recovery (https://www.stcn.com/article/detail/1486267.html)
[6] Sina Finance - In the Post-Moore Era, Advanced Packaging Moves to the “Center Stage” (https://cj.sina.cn/articles/view/5953190035/162d67893019017454)
[7] 21st Century Business Herald - Unigroup Guoxin Launches IPO Tutoring, AI Ignites IPO Wave of Memory Chips (https://cj.sina.cn/articles/view/1651428902/626ece)
Insights are generated using AI models and historical data for informational purposes only. They do not constitute investment advice or recommendations. Past performance is not indicative of future results.
About us: Ginlix AI is the AI Investment Copilot powered by real data, bridging advanced AI with professional financial databases to provide verifiable, truth-based answers. Please use the chat box below to ask any financial question.
