Author: Peng Cheng
The growth in demand for AI computing power and high-end storage has driven a large-scale expansion of the advanced packaging industry both domestically and internationally. The total investment across the entire industry, including mainland Chinese manufacturers such as JCET (Jiangsu Changjiang Electronics Technology), TFME (Tongfu Microelectronics), HT-Tech, Forehope Electronic, SJ Semiconductor, Shenzhen Kaifa Technology, and BIWIN Storage Technology, has exceeded CNY 40 billion. The expansion is concentrated on advanced technologies such as 2.5D/3D packaging, Chiplet, and HBM.
However, the investment amount is just the beginning. From factory construction and equipment installation to production line customer certification, yield stabilization, and the formation of profitable capacity, there are many steps that require time. Money has been invested, but capacity does not easily keep up. The domestic supply chain will benefit from this, but the timing and extent of benefits vary significantly across different segments. Orders do not equal revenue, sample submissions do not equal mass production, and OSAT expansion does not necessarily mean an immediate rise in domestic market share.
01、Ten-Billion-Level Capex Directed at 2.5D/3D and HBM
JCET has made the largest investment in this round. Recently, JCET's board of directors approved the "Proposal on Mid-Year Adjustment of the Company's 2026 Investment Plan," proposing to increase the2026 investment plan from CNY 11.01 billion to CNY 18.2 billion, with fixed asset investments rising from CNY 9.98 billion to CNY 17.31 billion. This additional investment is primarily directed toward building advanced packaging-related production lines. JCET plans to build a high-end advanced OSAT factory in Shanghai Lingang in two phases, with a total investment of CNY 7.8 billion. JCET also intends to raise no more than CNY 6.5 billion through a private placement for projects including the expansion of the high-performance computing high-end advanced packaging platform, the upgrade of high-end power module advanced packaging test capacity, the upgrading and expansion of the wafer-level packaging advanced process platform, and the enhancement of system-level packaging test capabilities for high-density large-capacity memory chips.
TFME's expansion mainly revolves around core customer AMD's high-end AI packaging and testing needs. In May 2026, the second phase of the Tongfu Superwei Suzhou project was launched, attended personally by AMD CEO Lisa Su; just months prior, the company announced its intention to raise no more than CNY 4.4 billion to invest in four major high-end packaging and testing fields: storage, automotive electronics, wafer-level, and high-performance computing. The pace of capacity expansion is clearly accelerating. In 2026, the company plans to invest CNY 9.1 billion in capacity construction and other areas, a significant increase from CNY 6.0 billion in 2025.
HT-Tech is based on memory packaging and testing. Its controlled subsidiary, HT Nanjing, is investing CNY 3.0 billion to build the second phase of the advanced packaging and testing industrial base. Upon reaching full capacity, it is expected to package and test approximately 430 million integrated circuits for memory annually. The project construction period is two years, from June 2026 to May 2028, with an estimated annual operating revenue of CNY 2.15 billion and net profit of CNY 126 million after reaching full capacity. HT-Tech stated that this project investment is a crucial measure for the company to further expand its layout in memory integrated circuit packaging and testing business and adapt to the rapid development of the industry. The products of the project are mainly applied in fields such as AI computing power, servers, consumer electronics, smart terminals, automotive electronics, and data centers.
Among other companies, SJ Semiconductor plans to invest CNY 10.0 billion to build the first phase of the three-dimensional integrated chip manufacturing project, aiming for large-scale mass production of 3DICs. Forehope Electronic announced two investments, domestic and international, totaling CNY 12.4 billion within half a year, including a 2.5D product line. A subsidiary of Shenzhen Kaifa Technology plans to invest CNY 1.85 billion to expand high-end memory chip packaging and testing capacity.
It is worth noting that the expansion is not limited to traditional OSAT firms. Memory manufacturer BIWIN Storage Technology's controlled subsidiary, Guangdong Xincheng Hanqi, plans to sign a cooperation agreement with the Dongguan Songshan Lake Management Committee for the third phase of the wafer-level advanced packaging and manufacturing project, with a total investment of approximately CNY 4.5 billion, including CNY 4.0 billion in fixed asset investments. The first and second phases, involving planned investments of CNY 3.09 billion, have been completed, corresponding to a monthly capacity of approximately 3,000 wafers.
02、Expansion Encounters Bottlenecks in Equipment, Materials, Talent, and Certification
But capital expenditure is only the first step in expansion. If any link—factory, equipment, materials, talent, or customer certification—is stuck, the investment cannot translate into capacity. The special nature of this wave of expansion lies in the fact that demand arrived too quickly, and every link in the supply chain was unprepared.
First is the issue of factory and equipment delivery. Advanced packaging requirements for cleanrooms and space far exceed those of traditional packaging. Building high-end cleanrooms from scratch often takes more than 18 months, making existing quality factories a scarce resource. More commonly, equipment lead times have stretched: due to global OSAT vendors expanding, the delivery time for some core advanced packaging equipment has extended to over one year, while general equipment lead times have increased from 3 to 6 months to 6 to 8 months. The increasing complexity of AI chips is also amplifying equipment demand. Morgan Stanley data shows that the test time per GPU has lengthened from about 350 seconds for the Hopper architecture to 700–1000 seconds for Blackwell, with the next generation expected to reach 1800–2000 seconds; the pin count of test sockets has surged from about 1500 for traditional mobile phone chips to 6000 for AI chips, with the next generation exceeding 10,000. More critically, the localization rate of key equipment such as hybrid bonding, high-end lithography, and high-precision inspection remains low, with the most critical categories still relying heavily on imports, meaning the supply rhythm is not fully controlled by domestic enterprises.
Second is the material issue, which extends beyond ABF substrates. For advanced packaging substrates, especially ABF substrates, Ajinomoto of Japan supplies more than 95% of the world's ABF build-up film, and some capacity has already been booked until 2027. In August this year, Ajinomoto notified mainland Chinese customers to cut ABF film supply by 30%. This reduction directly pushed the domestic IC substrate and AI chip packaging supply chain into a tense state. Meanwhile, the extreme requirements for signal transmission in AI servers and high-speed switches have triggered a "super cycle" for high-end copper-clad laminates: the supply of upstream HVLP low-profile copper foil and low-dielectric electronic glass cloth is tight, the global delivery cycle for Toyota Type JGT looms used to produce electronic glass cloth exceeds one year, and the delivery time for high-end copper-clad laminates has generally lengthened from the normal 7 days to 15–20 days, with some special items requiring waits of over 6 weeks. Furthermore, from equipment procurement and commissioning to customer certification, material expansion typically requires another 18 to 24 months. The pace of OSAT expansion is clearly constrained by upstream single-supplier and oligopoly structures on the material side.
In terms of talent, advanced packaging requires interdisciplinary professionals who understand architecture, packaging, and interconnects simultaneously. Currently, the number of architects nationwide meeting the criteria of "over 10 years of experience plus significant AI/GPU Chiplet project experience" is estimated to be only 200 to 300, most of whom have already been locked in by top-tier companies. Newly posted jobs related to advanced packaging grew by 134.6% year-on-year, and Chiplet-related positions grew by 211.2%, but such talent cannot be solved through short-term training. Talent shortages do not directly reduce capacity numbers but slow down the progress of production line commissioning, customer certification, and yield improvement.
Finally, there is customer certification and yield ramp-up. Moving from verification to large-scale mass production in advanced packaging requires multiple steps, including customer certification, yield ramp-up, and reliability verification. Conceptual hype does not equate to current performance. TrendForce predicts that the severe shortage of global 2.5D packaging capacity will only begin to ease slightly by 2027. Before then, NVIDIA's lock-in of large amounts of CoWoS capacity and key materials through supply chain control has already left other tech giants facing material shortages and limited product growth rates. The situation is similar domestically: Forehope Electronic's 2.5D packaging is currently still in the sample verification stage and has not entered mass production; BIWIN Storage Technology's wafer-level packaging line is still in process verification and small-batch trial production; TFME's realization of high-end capacity largely depends on the adoption progress of its core customer, AMD. These factors will not change the direction of continued domestic advanced packaging expansion but will affect the speed at which investment converts into effective capacity and the timing of profit release.
Equipment lead times cause delays of several months to over a year, potentially postponing nearly half of the current capacity for individual projects. ABF substrates cause actual capacity losses, and the gap ratio is expected to widen year by year. This loss cannot be resolved solely by efforts within the packaging segment. Talent and certification issues affect the speed of yield ramp-up, causing the formation of effective capacity to lag behind equipment availability by some time.
03、Domestic Supply Chain Benefits: Layered Rhythm, Differentiated Realization
On the flip side of the coin: the tighter the supply and the more urgent the expansion, the larger the window for domestic supply chain integration. While bottlenecks constrain the capacity rhythm of OSATs, they open opportunities for the volume ramp-up of domestic equipment and materials. However, the transmission of this opportunity varies in speed across different segments.
Order and revenue realization is fastest in the equipment segment. Hwatsing Technology's thinning and polishing machines have cumulatively shipped over 20 units, receiving repeat orders from multiple leading enterprises. New models, including the 12-inch advanced memory thinning equipment Versatile-GH300, compound semiconductor wafer thinning equipment Versatile-GN200, and 8-inch ring-retaining wafer thinnner Versatile-GR200, have all completed their first unit shipments, with the GR200 already delivered to customers. GL Tech's domestic mechanical dicing saws continue to operate at full capacity and have received batch repeat purchases from top domestic OSAT vendors. Xinqi Micro-Electronics' wafer-level direct-write lithography equipment has passed verification on a leading OSAT's CoWoS-L production line, achieving the first large-scale introduction of domestic direct-write equipment into high-end AI chip packaging lines, and has secured repeat batch orders. However, some segments remain in early stages. Piotech's hybrid bonding equipment generated revenue of CNY 136 million in 2025, accounting for only about 2% of the company's annual revenue. Those truly benefiting solidly are niche categories that have entered top customer production lines and secured repeat orders, rather than all companies involved in advanced packaging equipment.
The overall revenue growth in the materials sector is relatively fast, but profit transmission is slower than in equipment. In the first half of 2026, the combined operating revenue of 12 listed packaging material companies was approximately CNY 70.1 billion. The proportion of advanced packaging materials in the global packaging materials market rose from 37.93% in 2023 to 42.45% in 2025. Huahai Chengke's epoxy molding compounds (EMC) entered advanced computing packaging, with revenue growing by 161.77% year-on-year, and first-half revenue already exceeding the full year of 2025. Grinm Advanced Materials and Kangqiang Electronics maintained growth rates above 50%. Feikai Materials' EMC epoxy molding compounds and solder balls are stably mass-produced, and thick-film negative photoresist has obtained verification from mainstream packaging plants and formed small-batch orders. It is worth paying extra attention to the fact that the tight overseas supply of high-end copper-clad laminates has opened a critical window for domestic HVLP copper foil and low-dielectric electronic glass cloth: major laminate manufacturers have begun actively supporting domestic suppliers and accelerating test certifications, representing a rare "demand-driven" opportunity in localization. However, overall, material verification cycles are longer, and batch stability and long-term reliability certification are difficult to accelerate with capital alone; moving from small batches to scale procurement often takes several quarters or even longer. Current revenue growth for material companies may stem more from price-volume increases in mature categories during industry booms, while profit contributions from new advanced packaging categories will be slower.
Within the materials sector, ABF substrates are a special variety: the segment with the greatest long-term potential but the thinnest short-term revenue. Shennan Circuits' ABF substrates are already shipping in volume on flagship CPUs from AMD, Intel, and Qualcomm, making it the only domestic enterprise with high-end ABF substrate technology that has broken through to overseas major clients. Its Guangzhou factory has mass-produced FC-BGA packaging substrates with 22 layers or fewer, and R&D prototyping for24 layers or more is progressing on schedule. The mass production of 22-layer substrates signifies that domestic substrates have crossed the threshold into the AI chip level.
04、Summary
The drive for the domestic supply chain from this round of expansion is clear, but not all segments can benefit simultaneously. Order and revenue realization comes first in the equipment segment, where the truly solid beneficiaries are categories like thinning, dicing, and direct-write lithography that have entered repeat procurement, while hybrid bonding and advanced packaging metrology are still in the 0-to-1 stage. Revenue growth in the materials sector is relatively fast, but new advanced packaging categories must first pass verification, so profit transmission lags behind revenue. High-end copper-clad laminates and their upstream copper foil and electronic glass cloth have welcomed a demand-driven window for domestic integration due to tight overseas supply. ABF substrates have great long-term potential, but their short-term revenue volume remains small. Shennan Circuits' breakthrough with overseas clients and Kinsus' capacity ramp-up represent different paths, with the common point being that both require time.