The power chain of the US stock market has erupted again, and AI infrastructure has entered a comprehensive expansion cycle.

On September 3 local time, the technology sector of the U.S. stock market made a strong rebound, with the AI computing power chain collectively gaining strength. Dell Technologies rose by 15.81%, NVIDIA increased by 3.21%, Micron went up by 2.43%, and the Nasdaq climbed by 1.40%. In this round of rebound, the leading gains were not solely from pure chip stocks, as server manufacturers showed greater resilience. Dell Technologies, as one of the core suppliers of AI servers globally, saw its stock price surge significantly, indicating strong industry relevance. Its performance reaffirms that AI capital expenditures are transitioning from the chip sector to broader infrastructure domains. The future development of the AI industry requires not only more powerful computing chips but also the computing infrastructure capable of supporting large-scale applications. From chip competition to competition in computing power systems, the AI industry has entered a period of infrastructure expansion. Over the past two years, market attention has been focused on the capabilities of large models and GPU supply, with companies like NVIDIA being the core beneficiaries of the AI wave. Now, the focus is shifting toward the backend: the importance of high-speed interconnections, AI networks, and data center infrastructure is continuously increasing, with companies like Broadcom benefiting from the growth in AI data center construction and high-speed interconnection demand. The industrial chain of AI infrastructure is becoming more complete, spanning from GPUs to accelerated chips, from servers and high-performance storage to network devices and data centers. The reasons behind this shift are changes in demand structure. If the core question of AI previously was Can large models be trained?, the current question has turned to How can large models be widely applied?. Training requires concentrated high-performance computing power, while inference necessitates continuous, stable, and efficient power supply; as AI applications enter physical industries such as manufacturing, finance, healthcare, and energy, the demand for computing power is shifting from phased investments to long-term infrastructure needs. The domestic computing power industry chain is accelerating its layout, with equipment and intelligent computing operations expanding simultaneously. Changes in overseas capital expenditures first reflect in the domestic chip and server sectors. Companies like Inspur Electronic Information Industry (000977.SZ), Foxconn Industrial Internet (601138.SH), and Dawning Information Industry (603019.SH) are continuously laying out AI servers and high-performance computing to provide foundational support for large model training and industry applications. Cambricon (688256.SH) is a domestic chip example: with limited supply of high-end GPUs, its cloud training and inference chips are continuously iterating, serving leading internet companies and intelligent computing center operators, with revenue rapidly climbing alongside domestic computing power procurementthe shift to domestic substitution has turned from a thematic discussion into actual orders. The acceleration of computing power infrastructure is being reflected in the mid-term performance of domestic companies. Range Intelligent Computing Technology Group (300442.SZ) focuses on AIDC business: in the first half of 2026, it achieved revenue of 3.746 billion yuan, with AIDC revenue of 1.995 billion yuan, a year-on-year increase of 126%, making up more than half for the first time, and a gross profit margin of 47.3%; net profit attributable to the parent company was 1.203 billion yuan, with an operating cash flow to revenue ratio of 79%. The company operates about 750 MW of data centers and has delivered around 100,000 high-density liquid-cooled cabinets, with a PUE as low as 1.08, serving clients including ByteDance, Alibaba Cloud, and Tencent Cloud, and achieving heavy asset cycle exits through REITs. Sharetronic Data Technology (300857.SZ) is shift from smart terminal manufacturing and trade to computing power services, with its main business including equipment sales and computing technology services: during the first half of the year, it achieved revenue of 12.523 billion yuan, a year-on-year increase of 153.3%, and net profit attributable to the parent company was 1.838 billion yuan, with an ROE of 34.3%. The company plans to raise 8 billion yuan to invest in intelligent computing centers, aiming for a computing power scale of 50,000 P (FP16 dense) by the end of 2026, and extend its offerings through platforms like FCloud and TokenShare towards the MaaS model. Jiangsu Lettall Electronic (603629.SH) is laying out its trade in server resources and computing power services: in the first half of the year, it achieved net profit attributable to the parent company of 702 million yuan, a year-on-year increase of 1275%, with a comprehensive gross profit margin of 46.6%, ranking at the top in the sector; revenue from computing-related services was 1.274 billion yuan, accounting for 61% of total revenue. The company holds the preferred-level qualifications for NVIDIA and operates 38,000 P of high-end computing power, maintaining full rental at its data center, and has signed a three-year contract worth 5 billion yuan with Tencent. The Hong Kong stock market's computing power services are seeing new opportunities, with business models shifting from resource construction to operational upgrades. As the AI industry enters a phase of large-scale application, related companies in the Hong Kong stock market are also adjusting their layout around cloud services, computing power operations, and AI infrastructure services. KINGSOFT CLOUD (03896) has continuously strengthened its AI cloud service capabilities in recent years, upgrading its traditional cloud computing business to AI infrastructure services. As one of the earliest independent cloud service providers in China, KINGSOFT CLOUD has fully embraced AI over the past two years: leveraging the natural scenarios of the Xiaomi and Kingsoft ecosystems, AI-related revenue has continued to grow, with intelligent cloud services becoming the main engine of growth. Meanwhile, it is continuously increasing its investment in AI infrastructure, enhancing its full-stack capabilities from IaaS to MaaS. As enterprises' demand for AI applications continues to grow, companies with cloud platform capabilities, customer resources, and ecosystem systems are likely to gain new growth space during the industrialization process of AI. In the computing power service field, GBA AI COMP (01396) is gradually realizing its upgrade from a large computing power technology service provider to a super TOKEN factory. The company is laying out AI computing power services and operations, integrating computing power resources, building computing power cloud platforms, and expanding application ecosystems, promoting the transformation of computing power from a basic resource to industry service capabilities. In the first half of the year, it achieved revenue of 2.56 billion yuan, ten times that of the same period last year; net profit was 280 million yuan, five times last year's total. Currently, the operating computing power scale has exceeded 50,000 P (FP16 dense), accounting for over 2% of the national total computing power, with plans to expand to 80,000 to 100,000 P by the end of the year; it is pushing for the integration of computing power resource scheduling and application services through its Quantum Pie computing power cloud platform. It has signed contracts with over 200 corporate clients and over 3,000 OPC and individual customers. The sustained strength of the computing power chain reflects the expansion of AI capital expenditures, with capital expenditures transitioning from chips to servers, data centers, and computing power operations. For companies, scale is just the entry ticketorder support, delivery capability, resource utilization efficiency, and profitability are the variables determining long-term value.