Profit differentiation in the semiconductor industry by 2025

As the 2025 annual report disclosure season heads into full swing, many companies in the semiconductor industry have turned in impressive “report cards.” According to data from THS, as of March 30, the date of the reporter’s submission, among 173 semiconductor companies categorized under the Shenwan industry classification, 42 have officially released their 2025 annual reports. Of these, 29 companies saw year-over-year growth in attributable net profit; 117 companies disclosed performance briefings, with 76 companies recording year-over-year growth in attributable net profit. Overall, driven by the surge in demand for artificial intelligence (AI), industry-wide sentiment is moving upward. However, behind the “spotlight,” profit-earning capability across different links in the industrial chain is showing a clear structural divergence. Industry experts say this divergence is not simply a cyclical fluctuation, but a distinct reflection of how the AI-driven semiconductor industry is moving from “rising or falling across the board” toward “structurally favorable conditions.”

Opportunities for “chips” in sub-sectors such as memory

Benefiting from the accelerating rollout of AI large-model commercialization, compute demand has grown exponentially, boosting “multiple points blooming” performance across sub-tracks. In this context, memory chips in particular are seeing both volume and price rise, with an encouraging “uptrend.”

Pavilion Memory’s performance in 2025 was especially strong. The company achieved operating revenue of RMB 11.3B, up 68.82% year over year; attributable net profit of RMB 853M, up 429.07% year over year. Among them, revenue from AI emerging on-device storage products was about RMB 1.75B, up sharply year over year. This strong growth momentum continued into early 2026. The company expects attributable net profit of RMB 1.5 billion to RMB 1.8 billion from January to February 2026, representing an increase of 921.77% to 1086.13% year over year. This figure already exceeds the level of net profit for the full year 2025.

In an interview with reporters from Economic Reference News, Zhu Keli, founder and president of the Research Institute of New Economy, said that the simultaneous rise in volume and price of memory chips is fundamentally driven by the explosive growth of emerging demand such as AI compute power and data centers. The demand for storage from AI servers is 8 to 12 times that of traditional servers. Coupled with leading companies shifting production capacity toward higher-end products, this has led to structural supply-demand imbalance, pushing prices to historic highs. At the same time, the memory industry itself is in an up-cycle phase, further amplifying earnings resilience.

This market demand sparked by compute power also flows upstream to the semiconductor equipment segment. Based on disclosed performance briefing data, in 2025, 15 listed semiconductor equipment companies together achieved revenue of RMB 42.81B, up 30.97% year over year, and combined attributable net profit of RMB 6.31B. Among all sub-sectors, their growth rate led the pack. Among them, Peking Micro’s revenue was RMB 12.39B and attributable net profit was RMB 2.11B, firmly keeping it as an industry leader. AP Systems’ revenue growth rate was as high as 58.9%, and demand for thin film deposition equipment was fully validated. Huaweiicheng Technology’s revenue grew 36.5% year over year, and the company has continued to maintain competitive advantages in CMP equipment.

In the wafer foundry stage, leading companies delivered steady performance. According to the annual report, Semiconductor Manufacturing International Corporation (SMIC) achieved operating revenue of RMB 67.32B in 2025, up 16.5% year over year; attributable net profit was RMB 5.04B, up 36.3% year over year. The company said that the change in operating revenue was mainly due to an increase in wafer shipments during the year. GUC Semiconductor achieved operating revenue of RMB 10.89B in 2025, up 17.69% from the same period last year; attributable net profit was RMB 704M, up 32.16% from the same period last year.

However, industry momentum is not distributed evenly across the value chain. For example, based on performance briefings, many companies in the semiconductor materials segment saw “cooling” in their results. Seventeen companies combined for revenue of RMB 23.5B, up 16.95% year over year, but combined attributable net profit was -RMB 472M. Among them, in 2025, Huzsilicon Materials suffered a loss of RMB 1.48B, Xi’an Yicai suffered a loss of RMB 738M, and Tianyue Advanced turned from profit to loss with a loss of RMB 208M, all facing downside pressure to varying degrees.

AI is reshaping the industry and value logic

If we distill the main line of the 2025 semiconductor industry, it is undoubtedly “AI cashing in.” Judging from the overall data, the company group most directly related to AI compute power and storage contributed the vast majority of the industry’s incremental profit.

As a benchmark enterprise in China’s domestic AI chip sector, Cambricon has entered a “spotlight moment” since listing on the STAR Market. The annual report shows that the company achieved full-year operating revenue of RMB 6.5B, up 453.21% year over year; attributable net profit was RMB 2.06B. This marks the first time the company has achieved annual profitability since it listed in 2020.

Meanwhile, benefiting from the development of the AI industry and strong demand in the high-performance GPU market, China’s domestic AI chip lineup overall has shown a favorable earnings trend. The annual report shows that, in 2025, Huxi Semiconductor achieved operating revenue of RMB 1.64B, up 121.26% from the same period last year; attributable net profit was -RMB 789M, and the year-over-year loss narrowed by 43.97%. According to performance briefings, Moore Threads’ 2025 revenue was RMB 1.51B, up 243.37%; attributable net profit was -RMB 1.02B, and the year-over-year loss narrowed by 36.70%.

Overall, driven by continuous technological iteration, many companies in the domestic AI chip sector have already successfully crossed the “technology validation” stage and moved into the commercialization turning point of “mass shipments.” The deeper logic behind this leap lies in the large-scale rollout of AI application scenarios.

Regarding AI’s deep impact on the semiconductor industry chain, Tian Lihui, dean of the Institute for Financial Development at Nankai University, pointed out that AI’s reshaping of the semiconductor industry is evolving from “single-point demand pull” to “bottom-layer restructuring.” On the demand side, as AI moves from compute training toward the inference stage, it drives semiconductors from “general-purpose computing” to “heterogeneous computing,” leading to exponential growth in demand for high-bandwidth memory, advanced packaging, and dedicated chips. The essence of AI reshaping is to shift semiconductors from being priced “based on the number of transistors” to being priced “based on system performance.” This is both a challenge and a necessary path for industrial upgrading. The industry logic is shifting from “process node shrinkage” to “system-level innovation.” Companies need to build new “moats” through ecosystem integration and deep development in scenarios to move forward steadily in the compute revolution.

“AI’s penetration across the entire semiconductor industry chain is reshaping the industry’s competitive barriers and value allocation logic at the bottom layer. It is turning competition from a contest of a single technology or capacity ratio into an all-around duel of intelligent capabilities across the whole chain and ecosystem coordination abilities.” Zhu Keli added further: overall, AI is making competitive barriers in the semiconductor industry more diverse and shifting value allocation toward segments and companies whose “AI-oriented” capabilities are more prominent and whose ecosystem synergy is stronger.

The industry is moving toward “structurally favorable conditions”

Although the industry’s overall performance is improving, beyond the AI “spotlight,” significant divergence is emerging within the semiconductor industry.

A recent performance tracking special report on the semiconductor industry released by Dongguan Securities shows that, judging from companies at home and abroad that have disclosed results or earnings forecasts, industry-wide momentum is improving but sub-sectors are diverging. The demand pull from AI in sub-sectors such as compute power chips, memory chips, and wafer foundry is clearly evident. By contrast, sub-sectors not related to AI are showing a mild recovery trend. Some sub-sectors, such as consumer electronics, see overall sentiment under pressure because AI demand crowds out resources like memory, leading to rising costs.

On the earnings divergence phenomenon, Zhu Keli said this is an inevitable outcome of the overlap between the industry’s cyclical recovery and the restructuring of structural demand. Fundamentally, it comes down to fundamental changes in the supply-demand landscape and value logic in different links of the industrial chain as old and new growth drivers are being swapped. This divergence is not a simple cyclical fluctuation. It is a clear manifestation of how the AI-driven semiconductor industry is moving from “rising or falling across the board” to “structurally favorable conditions.” The links led by emerging demand are seeing development dividends, while links relying on traditional demand are stuck in the throes of transformation.

From Tian Lihui’s perspective, earnings divergence is an inevitable result of the semiconductor industry experiencing both cyclical recovery and structural change. The core reasons are concentrated in three areas: divergence in demand structure, mismatch in the inventory cycle, and differences in capacity and pricing power. He emphasized that divergence is not a short-term phenomenon but a new normal in industrial evolution. This is the process of the industry shifting from “general growth” to “precise prosperity.” High value-added links seize resources, while traditional links accelerate reshuffling. This will also become an unavoidable period of pain for China’s semiconductor industry to move toward high-quality development.

Zhu Keli believes that, based on the 2025 annual reports and 2026 Q1 performance guidance, the earnings divergence pattern in the semiconductor industry will further intensify in the short term, and in the medium to long term it will gradually converge as some links complete their transitions and as supply-demand reshaping proceeds. Its core reason is that the AI-driven structural demand dividend is still concentrated and being released, while the recovery of traditional demand is slow and cost pressures remain. As a result, development paces across different links cannot synchronize.