Sinolink: The demand for AI computing power in 2026 is expected to remain strong, optimistic about AI copper-clad boards/PCBs, core computing power hardware, semiconductor equipment, etc.

Sinolink research report stated that the capital expenditure of the four major cloud vendors in North America (Microsoft, Google, Meta, Amazon) continues to increase, with a positive outlook for future capital spending. The institution believes that the high capital expenditure of the four major CSPs in North America is sustainable, with room for further improvement. The NVIDIA AI server cabinet is expected to grow significantly by 2026, with a explosive growth in Token numbers. Sinolink predicts that the number of ASICs will see explosive growth, and continues to be optimistic about AI copper-clad PCBs, core computing hardware. In 2026, Sinolink expects growth in AI copper-clad PCBs, core computing hardware, semiconductor equipment, domestic computing power, storage price increases, and the industry chain benefiting from AI edge hardware. Sinolink's main points are as follows: AI copper-clad PCBs and core computing hardware continue to increase in quantity and price: NVIDIA GPUs are growing rapidly, while Google and Amazon ASICs are experiencing explosive growth, driving strong demand for copper-clad PCBs. In 2026, NVIDIA's Rubin PCBs will begin to use M9 material, and the ongoing introduction of orthogonal backplanes requires large amounts of M9 material. Google's TPU products are also expected to use M9 material, and other ASIC manufacturers are likely to adopt M9 material. Sinolink predicts a explosive growth in demand for M9 material in the next three years. AI PCBs have two major technology iteration trends, introducing orthogonal backplanes (transfer of copper value to PCB) and using CoWoP packaging technology (transfer of carrier plate value to PCB). Sinolink believes that the strong trend of AI copper-clad PCBs can continue, and other AI computing hardware such as AI servers, optical modules, liquid cooling, and power supplies are expected to continue to increase in quantity and price. Storage expansion cycle combined with independent controllable acceleration, optimistic about the semiconductor equipment industry chain: Semiconductor equipment is the cornerstone of the semiconductor industry chain, and the resonance of storage expansion and independent controllable technologies creates a vast space for domestic substitution. The storage chip architecture is undergoing a deep transformation from 2D to 3D. With the introduction of 3D DRAM technology and the evolution of NAND stacking layers to 5xx layers and above, the requirements for high aspect ratio etching and advanced thin film deposition in the manufacturing process are exponentially increasing, and related equipment manufacturers will benefit greatly from the complexity of the process. The domestic storage industry faces a more severe production capacity gap than the global market, with an urgent need for independent controllability. The expansion of Longcang and Longxin will directly increase the share of domestic equipment in key manufacturing processes. New opportunities for domestic computing power development: Domestic cloud vendors still have room for growth in capital expenditure, with Tencent, Alibaba, and Baidu's 25Q3 capital expenditures at 13 billion yuan, 31.5 billion yuan, and 3.4 billion yuan respectively, a year-on-year change of -24%, +80%, and +107%, respectively. Domestic cloud vendors are also increasing long-term AI investments, with Alibaba announcing in February this year an investment of 380 billion yuan over the next three years in cloud and AI hardware infrastructure. The future is promising for domestic AI chips, storage chips, advanced wafer manufacturing, exchange chips, and optical chips. Optimistic about Apple hardware and edge AI innovation in the AI/AR glasses industry chain: The global deployment of large AI models is experiencing rapid growth, and the industry has entered a period of explosive application scale. Apple's AI strategy is based on hardware, prioritizing the edge, and strong privacy protection. Core capabilities include language text, image/video, cross-app operations, and personal context understanding. Apple's AI is not a single function or large model, but a "personal intelligent system" deeply embedded in the operating system, chips, and application ecosystem, enabling cross-app operations based on personal context. AI/AR glasses are expected to continuously achieve technological breakthroughs and increase in sales, with prospects for SOC chips, optical core components, etc. Risk warning: Capital expenditures on AI are lower than expected, AI edge applications do not meet expectations, rising storage chip prices affect electronic hardware sales, and storage and wafer factory expansions are lower than expected.