Fueled by the global surge in computing power demand driven by artificial intelligence, technology stocks in the A-share market have reported substantial earnings growth, with their share prices continuing to climb.

Amid the global increase in demand for computing power sparked by AI, Chinese A-share technology stocks, particularly those focused on AI computing hardware, have maintained an upward trajectory. On April 17, domestic optical chip leader Yuanjie Technology (688498.SH) saw its shares strengthen further, surpassing 1,410 yuan per share and overtaking Kweichow Moutai (600519.SH) during the session to become the new most expensive stock in the A-share market. By the close, Yuanjie Technology's stock had risen 10.05% to 1,445 yuan per share, while Kweichow Moutai fell 3.80% to 1,407.24 yuan per share. Since September 24, 2024, Kweichow Moutai has accumulated a gain of 20%, whereas Yuanjie Technology has surged more than 15-fold.

Concurrently, A-share optical communication concept stocks, such as those related to "optical modules," "CPO (co-packaged optics)," and "optical chips," have continued their ascent. The phrase "you need to stand in the light" has become a hot topic among A-share market investors recently. Meanwhile, the ChiNext Index, heavily weighted by optical communication concept stocks like the "Yi Zhong Tian" trio—Xinyisheng (300502.SZ), Zhongji Innolight (300308.SZ), and Tianfu Communication (300394.SZ)—has demonstrated strong performance, hitting a new 11-year high.

In reality, as AI-led technology stocks have rallied, their prices have continued to rise, while consumer stocks, represented by baijiu, have underperformed in recent years. In August 2025, domestic computing power leader Cambricon (688256.SH) also briefly surpassed Kweichow Moutai in share price. Subsequently, Cambricon's stock experienced some pullback. As of the close on April 17, Cambricon was trading at 1,334 yuan per share, 73.24 yuan lower than Kweichow Moutai, ranking third.

Statistics show that as of the close on April 17, apart from Kweichow Moutai, the top ten highest-priced stocks in the A-share market are almost entirely AI-related technology stocks. These include Yuanjie Technology, Cambricon, Zhongji Innolight (300308.SZ), MetaX MUXI (688802.SH), Moore Threads (688795.SH), QuantumCTek (688027.SH), Xinyisheng (300502.SZ), Robotech (300757.SZ), and Hengdong Guang (920045.BJ), all with share prices exceeding 500 yuan per share. They are primarily concentrated in AI computing hardware segments such as optical communication, optical modules, and domestic computing power.

The direct driving force behind this bull market in A-share technology stocks is the explosive growth in global demand for AI-related computing power. Since ChatGPT went viral worldwide at the end of 2022, global cloud service providers have significantly increased their capital expenditures, boosting investments in AI computing infrastructure. It is understood that the four major North American cloud providers (Microsoft, Amazon, Google, Meta) have seen sustained substantial growth in capital expenditures since 2024. Their combined capital expenditure in 2024 was approximately $228 billion, a year-on-year increase of about 55%. In 2025, the combined capital expenditure of these four cloud providers continued to grow, reaching around $406 billion, a year-on-year increase of approximately 78%. Entering 2026, based on provided guidance, the total capital expenditure of the four major cloud service providers is expected to exceed $600 billion, a year-on-year growth of over 60%. Specifically, Google plans a total capital expenditure of $175-185 billion for 2026, a year-on-year increase of 91%-102%, nearly doubling. Amazon has guided for a total capital expenditure of $200 billion in 2026, a year-on-year increase of about 59%. Meta's capital expenditure guidance for 2026 is $115-135 billion, a year-on-year increase of 60.84%-88.81%, while Microsoft's guidance is approximately $140-150 billion.

Simultaneously, major domestic internet companies are also ramping up their AI computing infrastructure investments. According to public information, ByteDance's capital expenditure for 2025 was 150 billion yuan, with a planned capital expenditure of 160 billion yuan for 2026. Alibaba's (9988.HK) capital expenditure for fiscal year 2025 was 92.5 billion yuan, with a planned capital expenditure of 380 billion yuan for 2025-2027. Tencent's (0700.HK) capital expenditure for 2025 was 79.2 billion yuan, with plans to at least double its investment in AI new products in 2026.

The substantial increase in capital expenditure by global cloud providers is primarily used for building AI data centers, directly driving demand for AI computing hardware. From the perspective of the computing hardware industry chain, it mainly includes AI computing chips dominated by GPUs, memory chips dominated by HBM, and communication/interconnect sectors like optical modules and PCBs, corresponding to computing, storage, and communication segments, respectively.

Within the global AI industry chain division of labor, A-share listed companies hold a significant advantage in communication/interconnect segments like optical modules and PCBs, while also catching up rapidly in the GPU computing chip segment. Consequently, benefiting from the surge in global AI computing demand, especially A-share optical module and PCB companies embedded in the North American computing power supply chain, have seen sustained and substantial earnings growth.

The Q1 2026 report disclosed by optical module leader Zhongji Innolight on the evening of April 16 showed the company achieved revenue of 19.496 billion yuan, a significant year-on-year increase of 192.12%. Net profit attributable to shareholders of the listed company was 5.735 billion yuan, a substantial year-on-year increase of 262.28%. The single-quarter profit exceeded half of the full-year 2025 net profit, demonstrating a strong growth trend. Zhongji Innolight's stock price has been rising continuously since 2023. As of the close on April 17, the stock had accumulated a gain of over 46 times, with a total market capitalization reaching 945.56 billion yuan, ranking 16th among all A-share listed companies by total market cap, behind BYD (002594.SZ).

The other two optical module companies, Xinyisheng and Tianfu Communication, also reported significant earnings growth. Xinyisheng's 2025 earnings forecast indicated a net profit attributable to the parent company of 9.4-9.9 billion yuan, a year-on-year increase of 231.24%-248.86%. Tianfu Communication's net profit attributable to the parent company for 2025 was 2.017 billion yuan, a year-on-year increase of 50.15%. As of the close on April 17, the combined market capitalization of the three "Yi Zhong Tian" optical module companies exceeded 1.8 trillion yuan. Notably, due to their significant weight in the ChiNext Index, their rising share prices have also contributed to the index's strong performance, pushing it to a new high since 2015.

Yuanjie Technology, which recently surpassed Kweichow Moutai in share price, operates in the same optical communication field as optical modules. Its main products are key upstream optical chips required for manufacturing optical modules. As optical module shipments have grown significantly, driving a substantial increase in demand for optical chips, Yuanjie Technology has benefited from the domestic substitution of optical chips, leading to robust earnings growth. According to its 2025 annual report, the company achieved operating revenue of 601.4 million yuan, a year-on-year increase of 138.5%, and a net profit attributable to the parent company of 190.9 million yuan, a significant year-on-year increase of 3212.62%. Yuanjie Technology's stock price has already doubled this year, accumulating a gain of over 125% as of the April 17 close, with a total market capitalization exceeding 120 billion yuan.

As share prices of A-share AI computing hardware-related companies have surged significantly, market focus is gradually shifting to whether the growth trend in global computing power expenditure driven by this round of AI technological revolution can be sustained and for how long. This will determine whether the earnings growth of related companies can be maintained.

Regarding global AI computing power demand, NVIDIA's founder and CEO Jensen Huang previously provided a ten-year demand outlook. At the APEC CEO Summit in October 2025, he stated that global trillion-dollar computing infrastructure construction had just begun, with the new round of global computing infrastructure build-out only completing its "first year." He suggested a full cycle would take at least ten years. He mentioned that the rise of AI marks the beginning of a brand-new computing era. We are at the start of a new ten-year construction cycle. With the advent of the AI era, every layer of the entire computing architecture is being radically transformed. He also predicted that AI would reshape global industries worth up to $100 trillion in the future.

According to Goldman Sachs estimates, based on the continued investment in AI infrastructure by global tech giants and the widespread application demand for AI technology across various industries, cumulative global AI infrastructure investment is expected to reach $3-4 trillion by 2030, approximately equivalent to 20.48-27.3 trillion yuan.

However, as large-scale AI computing infrastructure build-out enters its fourth year, the market is increasingly concerned about whether such massive computing power investments can yield corresponding returns. This year, following the earnings releases of US-listed cloud providers like Google and Amazon, although their performance was strong, the simultaneous announcement of significant capital expenditure increases caused stock price volatility, indicating the market's "capital expenditure punishment" for tech giants (a shift from anticipating "future growth" to anxiety over "current costs," leading to short-term stock price declines). Domestically, after internet giants like Tencent indicated substantial increases in AI investment, their stock prices also experienced significant fluctuations.

From the demand side, as the global number of tokens accelerates growth, each token requires computing power or cloud computing support. These capital expenditures can achieve considerable returns on investment in cloud revenue and cloud profits, with relatively high visibility. Furthermore, at the application level, recalling the development of mobile internet, the initial phase also saw rising users and usage rates without immediate revenue generation. Once scale was achieved, monetization was eventually realized through models like subscriptions, advertising, and transaction commissions.

Meanwhile, although A-share AI computing-related company stocks have performed well, many A-share market investors, having missed out on the earlier rally, remain cautious about buying at high levels. A Beijing-based private equity investor mentioned that while the industry trend for optical modules is indeed positive, chasing highs is not advisable, as the previous price increases have already priced in future earnings. He stated that if stock prices later correct to a reasonable range, he would consider making some allocations. Completely missing out would be regrettable, but if no suitable buying opportunity arises, he would forgo it, noting that the market is never short of opportunities, but also never short of risks. Another Shanghai-based subjective private equity firm with billions under management, which also missed the optical module rally, bluntly stated that trees don't grow to the sky.

However, based on industry expectations from relevant A-share companies, AI computing-related demand is likely to remain high for the next year or two. The minutes from Zhongji Innolight's investor communication meeting following its earnings disclosure mentioned that customers have provided demand guidance for 2026-2027, with some key customers already planning for 2028 demand. They are expected to further increase capital expenditures in 2028, requiring more diverse product types and larger quantities. This does not yet fully reflect the 2028 demand, as many customers have not provided guidance. These demand figures are expected to become clearer over time. The primary reason behind this is that AI computing demand is still growing rapidly, and some AI applications have even formed closed loops, giving key customers the confidence to further increase capital expenditure investments.

Goldman Sachs' latest in-depth report on AI capital expenditure points out that AI has moved from pilot projects to scaled implementation, with Agentic AI becoming the core evolution direction. The global technology supply chain is experiencing the largest capital expenditure cycle in history, benefiting computing power, chips, semiconductor equipment, and storage across the board, with clear long-term growth logic for the industry.

Morgan Stanley predicts that future computing power demand growth will be three times the compound annual supply growth forecast by NVIDIA. The shortage of computing power will persist and intensify. With the launch of new-generation chips, AI computing costs will decrease significantly, leading to a further explosion in demand. This implies that computing power suppliers, including chip manufacturers, optical communication companies, and data center equipment providers, will enjoy long-term structural benefits.