As the demand for AI computing power explodes and overseas controls continue to tighten, an intriguing new trend has emerged in China’s semiconductor industry: those who make chips are starting to make equipment.
Packaging and testing giants such as Huatian Technology, Changdian Technology, and Tongfu Microelectronics, as well as SMIC founder Zhang Rujing, have personally laid out upstream equipment.
This trend, combined with the dividends of domestic substitution and AI production expansion, is bringing several key incremental breakthroughs to the independent controllability of the entire chain.
From the perspective of supply chain security, the most direct effect of this model is to quickly fill the gap in back-end equipment that was “dare not to touch or untouchable” in the past.
In the past, the “pain point” of domestic equipment manufacturers was the lack of real production line verification scenarios – the cost of equipment manufacturers’ own verification was too high, and the downstream manufacturing end was afraid of affecting production capacity and was afraid to use it, forming an endless cycle of “equipment manufacturers are slow to iterate – downstream companies are afraid to use it”.
Now that the manufacturing side has taken action personally, this cycle of death has been directly broken.
The 12-inch wafer-level permanent bonding equipment launched by Haichuang Intelligent is the first domestic machine that can support HBM
The price of 3D stacked mass-produced domestic bonding equipment is only 1/3 of similar overseas products; the alignment accuracy of Jiaxin Semiconductor’s FC flip-chip soldering equipment reaches ±1.5μm, which has covered 30% of the production capacity of Tongfu Microelectronics’ AMD packaging orders; the localization rate of the core hardware of the 3nm wafer probe station tested by Statetan exceeds 90%, breaking the monopoly of Teradyne and Advantest.
These devices are not laboratory samples, but have completed large-scale mass production verification on their own production lines. The yield rates are stable at above 99.5%, which is the same as imported equipment.
In 2025, the localization rate of back-end packaging core segmentation equipment will increase from less than 8% in 2023 to 22%, of which manufacturing background equipment manufacturers will contribute 60% of the new replacement share.
From the perspective of efficiency improvement, the increment brought about by the manufacturing end is to compress the “process-equipment” adaptation cycle from 2 years to less than half a year.
Semiconductor equipment is not designed on drawings, but iterated in actual production lines.
For a piece of equipment to run through the complete process flow, it requires high-frequency and in-depth collaborative polishing with the manufacturing plant.
Traditional domestic equipment manufacturers often face the dilemma of “the process adaptation cycle is as long as 1-2 years after equipment delivery.” However, equipment manufacturers with a manufacturing background rely on the massive process data of their own production lines to directly compress the adaptation cycle to 3-6 months.
This means that the cycle from order signing to full-process yield verification for domestic equipment has been shortened from 18-24 months for traditional overseas equipment to 8-12 months, and the delivery efficiency has increased by more than 50%.
The procurement cost of domestic AI-related customized equipment is 40% – 60% lower than overseas, and the overall equipment investment threshold for HBM production lines has been reduced by 35%, supporting a year-on-year growth of more than 120% in new domestic HBM production capacity in 2026.
But switching the perspective to the risk dimension, this model is not all “dimensionality reduction strikes”, there are also obvious hidden dangers.
Process capabilities are not equal to equipment research and development capabilities.
Chip manufacturing focuses on process integration and yield control, while manufacturing equipment involves multiple interdisciplinary fields such as precision machinery, motion control, vacuum technology, and materials science.
There are many precedents where self-developed front-end core equipment projects at the manufacturing end have been unable to complete production line verification for a long time.
In addition, the compliance boundaries of related-party transactions are blurred – the actual controller controls both the listed company and the equipment company, and there is a natural regulatory risk of interest transfer. Many listed companies have received inquiry letters from the exchange because of this.
What needs to be more vigilant is that capital market speculation has preceded the implementation of the industry.
Suqian Liansheng’s cross-border indium phosphide substrate equipment business was issued a regulatory warning by the Shanghai Stock Exchange because it failed to disclose that the joint venture partner was a shell company and seriously exaggerated the project cycle.
Andre plans to cross-border mergers and acquisitions of integrated circuit equipment targets. On the day the news was disclosed, the stock price continued to rise by the limit, and the Shanghai Stock Exchange quickly issued a regulatory work letter.
Taken together, the model of extending the manufacturing end to equipment is an “emergency fill-in” path for China’s semiconductors under external extreme controls.
Its core value is not to build its own closed-loop supply chain, but to provide zero-threshold verification scenarios for third-party domestic equipment in the entire industry, breaking the long-standing “iteration dead cycle.”
Unlike TSMC’s model of maintaining its status as party A in purely joint research and development, and Samsung’s model of only deploying non-core links through subsidiaries, this domestic model is more like “learning to fight on the battlefield.”
Industry predictions indicate that by 2028, the overall localization rate of domestic semiconductor equipment will exceed 40%, and AI-related segments are expected to reach 60%.
Whether this increase can be realized as scheduled depends on whether the manufacturing side can balance the resource allocation of its own production line operations and equipment research and development, as well as the supervisory authorities’ ability to restrict the risks of related party transactions.
But one definite change is that the “equipment-process” adaptation gap is being effectively filled by this new trend.
