In the summer of 2026, the delivery pace of NVIDIA's Blackwell architecture GPUs has become one of the most closely watched variables in the global tech industry. The metrics tracked by the market have long shifted from the yield rate of the most advanced 3nm process node to the capacity ramp-up progress of TSMC's CoWoS packaging. In the bill of materials for a high-end AI chip, high bandwidth memory (HBM) and packaging and testing combined account for over 60%, with packaging costs even approaching 91% of the manufacturing costs of advanced process nodes.
Behind this figure lies a shift in the balance of value distribution in the semiconductor industry. Over the past three decades, value has continuously tilted toward wafer fabrication; every step forward in Moore's Law has increased the bargaining power of the manufacturing end. Today, packaging is beginning to claim more pricing power. It is no longer just the 'back-end process' at the end of the supply chain with repeatedly compressed gross margins, but has become the capacity bottleneck in the computing power era.
The pace of this change is faster than the market anticipated.
01. Equipment First: The Transmission Chain of Hundred-Billion-Yuan Capital Expenditure
When observing the prosperity of the advanced packaging industry, equipment serves as the most sensitive thermometer.
Data from Yole outlines a clear growth trajectory: global capital expenditure in the advanced packaging industry reached USD 16.9 billion in 2025, and is projected to grow by another 40% to USD 23.8 billion in 2026. The primary drivers of this growth are capacity expansions by leading enterprises such as TSMC, ASE, Samsung, and SK Hynix. The funding for this round of expansion is almost entirely concentrated in high-end directions such as 2.5D/3D packaging, HBM stacking, and Chiplet heterogeneous integration.
The capacity expansion pace domestically is even more aggressive. In the first half of 2026, leading domestic OSAT companies rolled out hundred-billion-level expansion plans in succession: JCET's annual capital expenditure budget is close to CNY 10 billion, with CNY 7.8 billion allocated to the high-end advanced packaging and testing factory in Lingang; TFME's CNY 4.22 billion private placement was finalized, with 18 institutions competing for shares; Yongsi Electronics invested CNY 10.3 billion in the third phase in Ningbo; and SJ Semiconductor invested CNY 10 billion in a 3D integrated chip manufacturing project. This does not even include SMIC's investments in establishing the Advanced Packaging Research Institute and XinSanWei.
The transmission of capital expenditure to the equipment end is highly rapid. The project investment structure disclosed in SJ Semiconductor's prospectus is representative: in the CNY 8.4 billion 3D multi-chip integration packaging project, CNY 7.515 billion is allocated for equipment procurement, accounting for nearly 90%. In other words, for every CNY 1 spent on packaging and testing expansion, CNY 0.9 flows to equipment manufacturers.
The financial reports of equipment manufacturers corroborate this. In the first half of the year, Xingji Micro-Equipment's revenue reached CNY 1.106 billion, a year-on-year increase of 68.95%, with net profit growing by 98.13%. Revenue from the pan-semiconductor business accounted for over 40%. Its WLP 2000 direct-write lithography equipment secured repeated batch orders on the CoWoS-L production line, while its PLP equipment secured a position in the next-generation panel-level packaging route. In the first half of the year, Piotech's net profit surged by 1324.10%. Its hybrid bonding equipment is progressing smoothly in customer validation. This track, regarded by Nomura Securities as key to 3D stacking, is moving from 'having it' to 'usable' and then to 'highly effective'.
The financial data of the two leading test equipment manufacturers, Changchuan Technology and AccoTEST, are more indicative. Test equipment is typically a lagging indicator in the semiconductor capacity expansion cycle, as procurement for the testing phase only occurs after equipment is installed in the fab. In the first half of the year, Changchuan's net profit grew by 125.67%, and AccoTEST's gross margin reached as high as 74.70%, with installed capacity approaching 10,000 units. The high prosperity of test equipment indicates that the construction of advanced packaging production lines has transitioned from the equipment installation phase to the capacity validation and ramp-up phase.
The advanced packaging layouts of NAURA and AMEC represent another trend: leading front-end equipment manufacturers are collectively entering the intersection of advanced packaging. NAURA released a 12-inch hybrid bonding equipment and secured batch orders. AMEC's TSV etching machine won orders from multiple customers. Their goals are very clear—to extend from their main front-end positions in etching and thin films to the new, faster-growing market of advanced packaging. AMEC has set a goal for itself to cover over 70% of the advanced packaging equipment market in the next five years.
The high prosperity on the equipment end is not a short-term pulse. The capacity expansion cycle for advanced packaging has two characteristics: first, rapid technology iteration, with equipment update frequencies higher than traditional packaging; second, significant room for domestic substitution, with the localization rate of critical equipment such as hybrid bonding, high-end lithography, and high-precision inspection still at low levels. The breakthrough from 0 to 1 by domestic enterprises brings dual elasticity of market share expansion coupled with industry growth. The superposition of these two factors makes the order visibility of equipment manufacturers much further out than the market generally expects.
02. Valuation Switch: Repricing of OSAT Leaders
If the prosperity on the equipment end is an open card, then the valuation changes of OSAT leaders represent the area of greatest market divergence.
The OSAT industry has long been classified by the capital market as 'asset-heavy cyclical manufacturing,' with its valuation center hovering around a dozen times PE. The reasons are sufficient: asset-heavy, strongly cyclical, lower technological barriers compared to wafer fabrication, and relatively weak bargaining power. This logic basically held true in the era of traditional packaging, but the popularization of advanced packaging is loosening this framework.
JCET's financial data for the first half of 2026 is worth a closer look. Revenue grew by less than 5%, but net profit increased by 79.41%, and gross margin improved by 1.68 percentage points. Volume did not increase much, yet profits and gross margins improved significantly—behind this is a shift in product mix. The computing electronics business grew by 40.4% year-on-year, leapfrogging to become the largest application field. The increased proportion of high-end packaging related to AI directly boosted overall profitability. A more critical signal is that JCET's advanced packaging revenue now accounts for nearly 70%, and the CNY 7.8 billion investment in Lingang is entirely directed toward high-end businesses such as 2.5D/3D, HBM, and computing chips.
The changes at TFME are even more dramatic. In the first half of the year, net profit reached CNY 1.717 billion, a year-on-year increase of 316.77%, already surpassing the full year of 2025. Its deep binding with AMD is its core advantage. The VISIONS platform integrates silicon interposer, vertical stacking, and multi-bridge technologies, and CPO technology has also passed reliability validation. The subscription lineup for the CNY 4.2 billion private placement speaks volumes: Xingzheng Global, Huaxin Dingxin, CETC Investment, J.P. Morgan, and Morgan Stanley—public and private funds, industrial capital, and foreign capital entering the market simultaneously. The vote of capital is more honest than research reports.
Hua Tian Technology, on the other hand, has forged a differentiated path. In the first half of the year, revenue grew by 35.09%, net profit increased by 259.15%, and the gross margin for the single quarter of Q2 surged to 16.88%. The 2.5D platform has achieved large-scale mass production, memory packaging and testing orders continue to surge, and the CNY 3 billion Phase II, Stage II project at the Nanjing base focuses specifically on memory IC packaging. Beyond AI, memory chip packaging and testing is another high-prosperity direction, and Hua Tian Technology has caught this rhythm.
The common characteristics of these companies are: a rapid increase in the proportion of advanced packaging revenue, continuous improvement in gross margins, and an upgrade in product mix from the mid-to-low end to the high end. When over 70% of a company's revenue comes from businesses with faster growth, higher technological barriers, and stronger profitability, its valuation framework should no longer use the traditional OSAT standards.
The market is already pricing in this change. From the end of 2025 to the middle of 2026, TFME's valuation center has significantly moved upward. The high profit growth is just the surface; the more core driving factor is the re-evaluation of the weight of the packaging and testing segment in the AI computing power industry chain. The valuation recovery of JCET and Hua Tian Technology is slower, as their business structure transition and profitability validation are still in progress.
The switch in the valuation system is a gradual process. OSAT leaders need to continuously prove with their performance that the profitability and technological barriers of high-end packaging are sufficient to support a higher valuation premium. Every increase in gross margin, every onboarding of a major client, and every mass production of a high-end production line is consolidating this foundation.
From the perspective of industrial development, the story of advanced packaging is far from over. Technology routes are still evolving rapidly: CoWoS is moving from S to L and then to R, with packaging sizes getting larger and larger; HBM stacking is moving from 8 layers to 12 layers and then to 16 layers, stacking higher and higher; 3D stacking and hybrid bonding are still in the early stage of volume ramp-up; panel-level packaging (CoPoS) and glass substrates are targeting the next-generation cost-reduction route. Every round of technology iteration means an increase in equipment value, and also means that the technological barriers and profitability of OSAT leaders are stepping onto a new level.
The value shift in the semiconductor industry often first reveals itself in the data; by the time the market reaches a consensus, the market trend has already gone a long way. Advanced packaging is experiencing such a stage. When packaging costs approach manufacturing costs, when equipment order visibility extends two to three years into the future, and when the profit growth rate of OSAT leaders far exceeds the revenue growth rate, this valuation repricing is already on its way. - End -