AI is pushing advanced packaging into a new round of expansion cycle, but what is truly holding the industry back is not just factories, but equipment, materials, and even software.
On September 1st, at the 3D IC Advanced Packaging Manufacturing Alliance event, He Jun, Vice General Manager of TSMC Advanced Packaging Technology and Services, and Hong Songjing, Executive Vice General Manager of Riyueguang, simultaneously released a signal: as the iteration cycle of AI products is compressed to nearly one year, the entire advanced packaging industry chain is entering a new stage of "simultaneous research and development, simultaneous expansion, and simultaneous verification", and equipment delivery time and software capabilities are becoming key bottlenecks determining production capacity and technology landing speed.

Hong Songjing revealed that Riyueguang is currently simultaneously building 7 new factories in Kaohsiung, including 4 in Nanzi Park, 2 in Renwu Park, and 1 in Luzhu Park. Such a large-scale expansion of production is not common in the company's development process. However, even with the comprehensive promotion of factory construction, the real limitation on capacity release is the speed of equipment supply.

He stated that the demand for advanced packaging equipment is currently extremely strong, and the delivery cycle of some key equipment has become the biggest limiting factor for expanding production. To confirm the market heat, he also shared a detail: during lunch that day, he asked three suppliers about their business performance this year, and the answers he received were year-on-year growth of 40% to 50%, 70% to 80%, and over 100%. This also reflects indirectly that the investment boom in advanced packaging brought by AI is rapidly spreading to the equipment industry.
If devices are the visible bottleneck, then software is a more easily overlooked aspect.
He Jun admitted that what really made him unable to sleep was not the hardware, but the lack of software ecosystem. He pointed out that in the past, semiconductor technology usually had an upgrade cycle of two to three years, and research and development, equipment introduction, and material verification could be carried out step by step; Nowadays, AI products are upgraded almost every year, and the industry no longer has time to gradually build production capacity after research and development is completed. Processes, equipment, materials, and design tools must be developed and verified simultaneously.
In his opinion, hardware issues can still be solved through collaboration between system vendors and supply chains, but software capabilities are still the weakest part of the entire ecosystem at present. Especially in the era of 3D ICs, EDA tools, manufacturing software, system collaboration platforms, and design verification capabilities all need to keep up with the development speed of advanced packaging, otherwise even the most advanced processes cannot truly form complete mass production capabilities.
In addition to software, He Jun also mentioned another persistent pressure - the supply of carrier boards and PCBs. With the continuous integration of more HBM, high-performance logic chips, and Chiplets into AI chips, the complexity of packaging is rapidly increasing. Carrier boards and high-order PCBs have become indispensable basic resources for advanced packaging, and supply shortages will continue for some time.
More importantly, the competition for advanced packaging is no longer a competition for single processes, but a competition for system level collaborative capabilities.
He Jun believes that in the future, the industry must comprehensively promote the STCO (System Technology Co Optimization) concept, allowing system design, chip design, packaging manufacturing, materials, and equipment to participate in development together. Because in AI systems, every addition of a chip or introduction of a new material affects thermal management, power supply, signal integrity, and manufacturing yield. Without the participation of any party, technical validation cannot be considered truly completed.
The 3D IC Advanced Packaging Manufacturing Alliance, established in September last year, was born based on this background. The alliance is jointly chaired by TSMC and Sunlight, with members covering multiple fields such as equipment, materials, measurement and testing, automation, etc. Its goal is not only to promote the localization and standardization of key equipment, but also to establish a practical platform for the industry to jointly develop, verify, and certify.
He Jun emphasized that this platform cannot just be limited to technical exchanges and conference briefings, but truly enables system companies, manufacturing enterprises, and supply chain partners to complete joint development of processes, equipment, and materials on the same platform, accelerating the introduction of new technologies into mass production.

For the equipment industry, this means that new opportunities have just begun. Hong Songjing stated that in the future, advanced packaging will continue to challenge larger mask sizes and further introduce CPO (co packaged optics) into the manufacturing process. Whether it is bonding, testing, automation, or optoelectronic integration, there are still many links waiting for equipment companies to jointly break through.
AI is not only driving a revolution in chip computing power, but also reshaping the entire advanced packaging manufacturing system. As the expansion speed becomes faster and faster, the true determinant of industrial competitiveness has shifted from "having factories" to "having equipment" and "having a mature software ecosystem".