Cinda Securities released a research report stating that global infrastructure investment is in a phase of rapid growth, benefiting all segments of the core industrial chain. The cyclical recovery of the industry, coupled with surging demand for artificial intelligence (AI), is propelling the sector into a new "super growth cycle." Global AI computing infrastructure is entering another expansion phase, while AI technology is reshaping the hardware ecosystem of smart devices, driving breakthrough transformations in the industry.
**AI Computing: Global Infrastructure Boom Drives Full-Chain Benefits** Google's release of Gemini 3, a next-generation large model benchmark, has set new standards in multiple AI benchmarks. Gemini 3 Pro redefines front-end development by integrating Agent and UI, reinforcing the Scaling Law as a guiding principle toward AGI. The global race for advanced AI models is a key driver behind the explosive demand for upstream computing infrastructure.
Fueled by this demand, global cloud service providers (CSPs) are entering a new capital expenditure expansion cycle. TrendForce predicts that CSPs' combined capital expenditures will exceed $600 billion by 2026, growing 40% annually, reflecting the long-term growth potential of AI infrastructure. This surge in spending will boost demand for AI servers, driving expansion across upstream supply chains—including GPUs/ASICs, memory, and packaging materials—as well as downstream systems like liquid cooling modules, power supplies, and ODM assembly.
**Overseas Supply Chain: GPU and ASIC Synergy Reshapes Server and PCB Value** NVIDIA expects Blackwell and Rubin GPU shipments to reach 20 million units by late 2026, generating $500 billion in GPU sales. Rising inference demand is also accelerating ASIC adoption, with CSPs actively developing in-house chips for cost and energy efficiency. AI servers are evolving from single-GPU designs to rack-level integration, driving rapid growth in cabinet demand. ODM and PCB segments stand to benefit from this shift.
**Domestic Supply Chain: Decoupling Software and Hardware Accelerates AI Deployment** Amid supply chain security concerns, domestic computing chips are narrowing the gap with global leaders. Huawei’s Ascend, Cambricon, and Hygon are enhancing performance and gaining market share as yields improve. Meanwhile, breakthroughs in advanced manufacturing continue, with foundries like SMIC expanding production to support domestic AI chip development.
**AI Storage: Recovery Cycle Meets AI Demand, Creating a "Super Cycle"** Memory suppliers’ production cuts have rebalanced the DRAM and NAND Flash markets, pushing prices upward. With limited capacity expansion focused on high-end products like HBM, conventional memory is expected to remain tight, supporting further price increases.
- **DRAM:** HBM capacity constraints may tighten general DRAM supply. DDR5 is becoming standard in new data centers, while AI servers drive demand for high-capacity RDIMM modules (64GB/128GB). - **NAND Flash:** AI training’s data throughput needs are accelerating QLC SSD adoption, replacing nearline HDDs. TrendForce forecasts a surge in high-capacity QLC SSD shipments by 2026.
**Edge AI: Reshaping Smart Devices** - **AI Phones:** Upgraded NPU performance and model compression are boosting penetration. Canalys and Omdia project AI phone shipments will rise from 18% in 2024 to 45% by 2026, nearing 60% by 2029. - **AI Glasses:** Ray-Ban Meta’s success validates the market for AI-integrated wearables. Multimodal AI enables real-time translation and imaging, with Wellsenn XR predicting explosive growth by 2026. - **Robotics:** Tesla Optimus’s progress signals humanoid robots’ shift from labs to factories. Traditional electronics suppliers like Lens Technology are entering the robotics supply chain, leveraging expertise in precision components.
**Key Recommendations:** - **AI Computing:** Overseas (Foxconn Industrial Internet, Unimicron, etc.); Domestic (Cambricon, Hygon, SMIC, etc.). - **AI Storage:** Modules (Longsys, BIWIN, etc.); Niche (GigaDevice, Ingenic, etc.). - **Edge AI:** SoCs (Rockchip, Espressif, etc.); Consumer Electronics (Lens Technology, Lingyi, etc.).
**Risks:** Slower-than-expected macro recovery; delayed tech innovation; intensifying competition.
Comments