Explosive Demand for AI Computing Power: A Comprehensive Investment Landscape of the Semiconductor Industry

The semiconductor industry is entering a new growth cycle driven by AI as the core force, with cloud computing focusing on substitution and terminals focusing on incremental growth.

Since the beginning of this year, with breakthroughs in the performance of large models like DeepSeek, the demand for AI computing power has shown explosive growth. The global semiconductor industry has entered a new upward cycle catalyzed by the AI wave. Data shows that in the first half of 2025, the electronic sector’s operating income is expected to grow by 19.2% year-on-year, and net profit attributable to the parent company is expected to grow by 29.0% year-on-year.

The semiconductor sector has performed strongly, with the Shenwan Electronics Index showing a cumulative increase of 32.0% from early January to September 5, 2025. AI not only drives the demand for cloud computing power chips but also accelerates the implementation of edge AI applications, creating new investment opportunities.

01 AI Drives a New Semiconductor Cycle, Industry Experiences Dual Prosperity Resonance

AI is becoming the largest driving force in the semiconductor industry. From 2023 to 2024, AI demand is mainly concentrated in the cloud, with the iterative evolution of large models driving rapid growth in computing power infrastructure demand. Currently, AI demand is extending from the cloud to the edge, leading the entire semiconductor industry chain into a new growth cycle.

The semiconductor industry is in a strong recovery phase in 2025. According to data from the Semiconductor Industry Association of America, global semiconductor market quarterly sales are expected to grow by 23.2% year-on-year in the third quarter of 2024, marking the largest increase since 2016.

The strong demand for AI servers is a key factor driving growth in the semiconductor industry. TrendForce estimates that global AI server shipments will reach nearly 1.2 million units in 2023, accounting for nearly 9% of total server shipments, and it is expected that the proportion of AI server shipments will increase to about 12.1% in 2024, with an average annual compound growth rate of 25% over the next three years.

Guojin Securities pointed out in a research report: “If 2023 is the year of AI training, and 2024 is the year of AI inference, then 2025 will be the year of explosive growth in AI terminal applications.” This trend is expected to drive a resonance of the three cycles of “demand – capacity – inventory,” propelling the semiconductor industry into a significant upward cycle.

02 Cloud Computing Power: Domestic Substitution is the Key Logic

In the field of cloud computing power, domestic substitution has become the core investment logic. The overseas AI industry continues to exceed expectations, while the gap between domestic foundational large models and overseas counterparts is narrowing in a spiral manner, driving continuous growth in domestic computing power demand.

Advanced processes are the foundation of domestic computing power. SMIC, as a scarce supplier of advanced processes in China, has strengthened its industrial position. CITIC Securities’ research report indicates that the successful expansion of SMIC’s advanced processes will significantly improve the company’s overall profitability.

Computing power chips are a direct beneficiary of autonomous control in cloud computing power. As overseas restrictions on high-end chip exports to China intensify, domestic computing power chip manufacturers are seizing development opportunities. Domestic computing power chip manufacturers such as Cambricon and Haiguang Information are expected to accelerate their growth in client-side deployment.

Advanced storage, as a supporting component for computing power, is also accelerating its domestic production process. In terms of HBM (High Bandwidth Memory), it is expected that by 2025, HBM will account for over 30% of DRAM output value, currently dominated by overseas giants, while domestic manufacturers are gradually breaking through.

The domestic production rate of enterprise-level SSDs is also accelerating. According to IDC data, the domestic production rate of enterprise-level SSDs in China is about 35% in 2023, driven by multiple forces.

Semiconductor equipment is a more fundamental investment direction. With the expansion of domestic wafer fabs, domestic equipment manufacturers are seizing development opportunities. Key breakthrough areas include photolithography equipment, high-end thin-film equipment, high aspect ratio etching equipment, and measurement equipment, which currently have low domestic production rates.

03 Edge AI: Three Major Hardware Upgrade Directions

The acceleration of edge AI applications brings considerable incremental demand. The upgrade of large model capabilities drives the implementation of AI applications, and the edge, as the hardware carrier for AI applications, has significant growth potential in the future.AI Phones/AIoT/Smart Driving and other smart hardware will bring new semiconductor demand.

Main Chip Upgrades are the first hardware upgrade direction for edge AI. Model deployment places higher demands on main chips, such as integrating NPUs, leading to increased value. AI glasses SoC chips are worth paying close attention to, with global shipments of AI glasses expected to reach 8 million units in 2025, and over 20 million units in 2026, presenting significant long-term potential.

Storage Upgrades are the second direction. 3D DRAM is expected to become an important solution for inference, with initial implementations in AIoT and AI phone scenarios. Domestic storage manufacturers are expected to accelerate their entry into the mid-to-high-end storage market through customized solutions.

Sensor Upgrades are the third direction, especially automotive-grade CIS (CMOS Image Sensors). High-level intelligent driving requires a greater number of onboard cameras, as well as higher pixel configurations. It is expected that by 2025, the demand for onboard cameras will reach 130-140 million units, with a year-on-year growth of nearly 60%, and domestic manufacturers are expected to further increase their market share.

04 Analysis of Investment Opportunities in Niche Tracks

Semiconductor Equipment and Materials

The domestic self-sufficiency rate of semiconductor equipment is at a low level, and domestic equipment is mainly applied in mature processes, indicating a strong demand for domestic production. Companies like Huawei are playing a positive role in accelerating the maturation and upgrading of the domestic semiconductor supply chain.

Leading quality equipment companies will undertake more challenging tasks, and breakthroughs in “0-1” in blank areas and improvements in domestic production rates in key areas will present significant investment opportunities.

Automotive Electronics and CIS Chips

Automotive intelligence has become the main theme of current automotive development, with BYD leading the way in equalizing intelligent driving. High-level intelligent driving requires a greater number of onboard cameras, as well as higher pixel configurations, leading to an increase in both quantity and price of automotive CIS.

PCB Industry Chain

In AI servers, due to the relatively complex architecture and higher performance requirements of AI servers, the value of PCBs in a single server will significantly increase compared to traditional servers.

The evolution of the PCIe bus towards high speed will drive continuous upgrades in PCB specifications, with PCBs made from copper-clad materials of M6 and above, with more than 16 layers, gradually becoming standard for servers.

05 Investment Strategies and Risk Warnings

For investments in the semiconductor industry, Guotai Junan Asset Management Fund Manager Chen Sijing proposed that the “Tick-Tock” strategy is worth referencing. “Tick represents when stock prices are at a low point, which is the time to layout offensively; Tock corresponds to when stock prices are at a high point, which is the time to take profits.”

This strategy has helped him accurately layout the Apple supply chain, semiconductors, and AI applications during the market rotation from 2024 to 2025, successfully avoiding risks and capturing rebounds.

Currently, attention should be paid to the risk of crowded trading. Chen Sijing warns: “Although the performance of companies related to the computing power supply chain is expected to maintain high prosperity, the sector currently faces issues of excessive short-term gains and crowded trading, which may lead to increased stock price volatility in the future.”

Investment directions should focus on three major standards: first, innovative markets from “0 to 1,” such as AI, cloud computing, AR/VR, etc.; second, growth tracks with continuous innovation capabilities, such as smart vehicles and optical innovations; third, opportunities for independent innovation brought about by China’s industrial upgrading.

In the next five years, as AI application scenarios continue to expand, the semiconductor industry will usher in a new round of growth opportunities. Investors can layout along the two main lines of “cloud substitution” and “terminal increment,” with a focus on core areas such as semiconductor equipment, advanced processes, and edge AI chips.

Those who can grasp the rhythm of the industry and find high-margin targets in niche directions are more likely to achieve excess returns in this wave of technological innovation.

Leave a Comment