Driven by demands such as AI inference, the need for AI computing power is continuously rising, and the supply-demand imbalance for related hardware is expanding from isolated points to broader areas. In the packaging and testing sector, the strong demand for AI computing power, coupled with pressure from rising raw material costs, has led some packaging and testing factories to recently initiate price increases for their services. In the memory sector, AI continues to fuel storage demand, with TrendForce forecasting the memory industry's output value to reach $5.516 trillion by 2026, a year-on-year increase of 134%. From the device-side perspective, AI holds vast potential and is expected to accelerate in 2026, deeply integrating into various hardware products and industrial scenarios, thereby creating investment opportunities across the industry chain.
Orient Securities believes that the current demand for AI computing power, driven by applications like AI inference, is climbing steadily, and the supply-demand imbalance for related hardware is spreading from specific points to wider fronts. In the upstream semiconductor foundry and packaging & testing sectors, foundries are pushing for price increases on some mature process nodes, spurred by growing AI-related power demands and production cuts by major manufacturers. In packaging and testing, the powerful demand for AI computing, resonating with cost pressures from raw materials, has prompted several packaging and testing plants to recently begin raising their service prices. For memory, AI persistently boosts storage needs; TrendForce estimates the memory industry's output value will hit $5.516 trillion by 2026, a 134% year-on-year increase, with the upward trend in DRAM and NAND Flash contract prices expected to continue into 2027.
Regarding CPUs, the demand for general computing power from AI agents is accelerating rapidly, and with general-purpose servers entering a refresh cycle, this is driving up prices for server CPUs. In the passive components sector, demand from areas like power management in AI servers is leading major manufacturers to continuously push for price increases on high-end passive components; for instance, Yageo raised prices for some tantalum capacitors used in AI servers and automotive electronics in November 2025. In server manufacturing, the U.S. easing restrictions on exports of NVIDIA's H200 chips to China is expected to boost demand for domestic server production. Furthermore, the demand for AI computing power on interconnect hardware like server PCBs and data center optical modules is expected to remain strong, driving demand for upstream materials such as high-end copper foil, electronic cloth, and optical chips, thereby maintaining a tight supply-demand balance for these materials.
Domestic computing-related hardware is continuously breaking through technical bottlenecks, promising to deepen import substitution. In the computing chip sector, domestic players like Cambricon, Hygon Information, and Moore Thread have emerged, while internet companies such as Alibaba and Baidu are advancing their own chip development, allowing domestic computing chip manufacturers to steadily narrow the gap with overseas giants. In the packaging field, domestic companies are making strides in advanced packaging breakthroughs; for example, JCET announced significant progress in integrated photonic packaging technology. A silicon photonics engine product based on the XDFOI® multi-dimensional heterogeneous advanced packaging platform has completed customer sample delivery and passed testing. In the memory sector, CXMT launched DDR5 products in November 2025, achieving mainstream technical parameters like peak data rates at international first-tier levels; Yangtze Memory's self-developed Xtacking architecture has enabled leapfrog development in 3D NAND technology.
Looking ahead, the space for on-device AI is vast, with attention on investment opportunities arising from AI upgrades and innovative hardware across the industry chain. Some investors believe current progress in on-device AI is relatively limited, with bottlenecks remaining in practical deployment. However, the implementation of AI on devices is expected to accelerate in 2026, deeply integrating into various hardware products and industrial scenarios and creating investment opportunities along the supply chain. For traditional consumer electronics terminals like PCs, TVs, and smartphones, AI is poised to accelerate empowerment, bringing opportunities for value increase in related hardware such as SoCs, on-device storage, thermal management, and sensing components. Regarding innovative AI hardware, new product categories are becoming increasingly functional, promising breakthroughs in human-computer interaction experiences.
Looking to this year, leading tech companies like OpenAI and Apple are expected to launch innovative devices such as AI headphones and AI Pins, while the smart glasses industry continues its rapid growth. These innovative hardware offerings are likely to usher in a new wave of growth opportunities for related enterprises in the industry chain. Risks include potential shortfalls in AI deployment, weaker-than-expected downstream demand, and fluctuations in upstream raw material prices.
