China's Semiconductor Equipment Component Industry Gains Momentum Through Import Substitution

China's semiconductor equipment component industry is reaching a critical stage of development. This growth is driven by increased demand stemming from the expansion of artificial intelligence infrastructure, the ramp-up of domestic wafer production, and geopolitical pressure on supply chains.

Achievements in Component Localization

ACM Research (Shengmei Shanghai) demonstrated the first systemic level of verification for a fully localized semiconductor equipment supply chain, assembling two semiconductor equipment processes using exclusively domestic components.

The drive to use internal components is dictated by several factors. The demand for AI computing extends from HBM memory and GPUs to inspection, power management, and advanced packaging equipment. SEMI forecasts that investments in 300mm memory equipment will exceed $50 billion by 2026.

Growth of China's Manufacturing Capacity

China's wafer manufacturing capacity has grown from 2% of the global volume in 2000 to 17% in 2020. SEMI expects China's share to reach one-third of the global volume by 2030. In the 22-40 nm node, China's share is expected to be 37% in 2026 and 42% by 2028.

Concurrently, tightening US export restrictions on critical equipment and components make securing supply chain safety a vital task.

Acceleration of Localization Among Manufacturers

At ACM Research's Lingang facility, capabilities for verifying all process categories have been implemented, including cleaning, electroplating, photoresist coating, tube furnaces, and PECVD processes. Critical standard components such as magnetic rotary pumps, heaters, and filters have achieved significant success in full localization.

ACM Research is not the only player: major Chinese equipment manufacturers, including New Kailai, AMEC, Piotech, and Hwatsing Technology, are actively accelerating the localization of components.

Target Categories and Partners

Key component categories include quartz parts (localization rate over 10%), RF generators, vacuum pumps (5-10%), valves, and rings for edge processing. Foxconn Precision completed validation with customers and began mass production of gas distribution wafers and special coatings, planning to focus on heating wafers and valves in 2026.

Siconix, one of the few domestic manufacturers with the full capability from 'crystal growth to finished silicon electrode,' supplies etching equipment to market leaders such as AMEC and NAURA. Feilihua, a supplier of quartz materials certified for Applied Materials, Lam Research, and Tokyo Electron, meets the highest certification standards, which typically require a qualification cycle of 2–3 years.

Global Expansion of Chinese Manufacturers

Chinese component manufacturers are beginning to expand their presence in the global market. Foxconn Precision gained access to international clients through investments in Compart Systems. Siconix supplies major Japanese and Korean silicon component manufacturers who work with Samsung, SK Hynix, Intel, and TSMC.

This dual strategy, combining 'following equipment' and 'independent expansion,' allows Chinese component manufacturers to transition from simple import substitution to global export.

pandaily.comJuly 1, 05:00

Chinese AI Chip Manufacturers Shift to 3D Stacking to Overcome Technological Limitations

Chinese companies producing AI chips are increasingly utilizing 3D stacking technology as a strategic alternative to upgrading advanced process nodes. This is happening amid the industry's confrontation with the physical limits of scaling flat chips and restricted access to cutting-edge manufacturing due to US export controls.

Overcoming Memory Constraints

The so-called 'memory wall,' which arises when memory bandwidth and capacity fail to keep pace with the exponential growth of AI model parameters, has forced the industry beyond the capabilities of standard 2.5D packaging technologies like TSMC's CoWoS. These flat solutions face fundamental limitations in resource routing, integration density, and reducing chip area for high-performance AI workloads.

Application of 3D Stacking in China

Several Chinese chipmakers are actively implementing 3D stacking solutions to gain a competitive edge. The company Tsingway is developing next-generation AI chips using heterogeneous 3.5D stacking. This architecture combines reconfigurable compute chiplets with vertically stacked DRAM in a 'four computing lanes plus four storage layers' spatial structure, significantly boosting data throughput and computational density.

Suanmiao Technology introduced its A4E TokenPU chip, which features eight layers of storage die vertically stacked above the compute logic. Thanks to TSV and micro-bump interconnects, this chip achieves a memory bandwidth of 16 TB/s. Lingchuan Technology, spun off from Kuaishou, developed a chip with fully domestic 3D stacking technology and an innovative 3D memory architecture adjacent to the processor. This chip is based on the SL200 model, which has already been sold to about 100,000 units to companies such as Alibaba Cloud, Baidu Cloud, and Bilibili.

Other Developments and Challenges

Unisplendour's Zixuan architecture focuses on 3D-DRAM with a target memory bandwidth of 30 TB/s and a Processing-in-Memory (PNM) mode. Meanwhile, Intellifusion is developing inference chips with 3D-stacked memory to ensure higher throughput and reduced access latency. Despite serious challenges remaining—including thermal management for stacked dies exceeding 350W, which requires liquid cooling, improving hybrid bonding yield, and the limitation of domestic EDA tools for 3D designs—the adoption of 3D stacking marks a strategic shift for Chinese semiconductor firms. This paves the way for increased computational performance regardless of the progress of the process node under conditions of limited access to advanced manufacturing technologies.