Editor's Note
The highly sophisticated semiconductor industry actually conceals many industrial secrets unknown to the general public. TOTO, a bathroom fixture manufacturer, monopolizes the semiconductor electrostatic chuck market; Ajinomoto, a MSG producer, controls the key materials for high-end packaging substrates. Traditional daily industries and core semiconductor tracks have long been deeply intertwined. Beneath the boom, the real landscape of the semiconductor industry is far more complex than it appears. In this episode of "Chip Unveiled: Experts Talk Chips," we have invited Mr. Guan Jian, a senior expert in the semiconductor industry. He entered the industry in 1997 and has been deeply engaged in frontline semiconductor technology. With decades of industry accumulation, he has focused on industrial research and in-depth industry judgment over the past ten years, and is the founder of Semiconductor Comprehensive Research. Facing numerous current industry questions: Who will ultimately reap the dividends of the AI boom? Is domestic substitution a breakthrough or a plunge into a red sea? Can advanced packaging reshape the industrial landscape? Has the current industry cycle already peaked? This in-depth dialogue will lead you through the industrial fog to understand the true status and trends of the semiconductor industry.
Welcome to the Dialogue Podcast
*The following is a text transcript of the podcast, confirmed and authorized by the guest.
AI Dividends Flow Upstream: Who Is Making a Fortune Quietly?
Huanshi: Mr. Guan himself has a frontline technical background. Compared with practitioners engaged in pure theoretical research, his industry perspective is more grounded. Today, I would like to ask Mr. Guan a few questions. Many people discuss the development of the AI industry with you. Could you combine the data from the supply chain orders and corporate financial reports you have been tracking for a long time to talk about: In this round of the AI boom, who exactly are the dividends flowing to? Which segments have full orders and truly reap the dividends, and which segments are merely runners-up in the industry boom? Are there any hidden companies with a low public profile but very strong performance growth?
Guan Jian: From the current situation, the AI industry itself has not yet entered a real period of profitability, and overall capital expenditures are massive. In this wave, the true beneficiaries are the upstream "shovel sellers." The most visible and well-known is certainly NVIDIA, followed by several memory manufacturers, who are now practically counting money until their hands cramp. Then there is TSMC (Taiwan Semiconductor Manufacturing Company), which, with its supply capability for advanced processes of 3nm and below, is also experiencing extremely booming business as a scarce supplier.
However, the truly unexpected dividends that most people cannot see are largely hidden in upstream equipment, component, and material companies deep within the supply chain. In recent years, as the optical communication track has gained popularity, the growth of related companies has also been very impressive. Many upstream component manufacturers in Japan, Europe, South Korea, and the United States have achieved performance growth by riding this AI boom, although the public profile of such companies is far lower than that of NVIDIA and TSMC.
A relatively representative one is the Japanese equipment company Advantest. You can check its financial reports; over the past two or three years, its financial growth data has been almost synchronized with NVIDIA. Why? Because they are NVIDIA's dedicated supplier. Almost all of NVIDIA's GPU chips use Advantest's testing equipment, so as NVIDIA's production capacity and sales surge, it directly drives a massive increase in Advantest's shipments.
Huanshi: Among numerous equipment manufacturers, it is indeed surprising that a testing equipment company can achieve such outstanding performance.
Guan Jian: Under the same production capacity, the consumption of testing equipment by GPUs is far greater than that of other equipment. Because GPUs have more ports, higher speed requirements, longer testing times, and stricter testing coverage standards, the demand for testing machines far exceeds that for other equipment. Of course, other equipment also benefits from the industry boom, but the dividends for testing equipment are the most prominent.
In addition, there are some component and material manufacturers, such as Nitto Boseki's Q-cloth, and TOTO's ceramic products. Many people only know TOTO for making ceramic toilets, but in fact, they also do exceptionally well in electrostatic chucks for semiconductor equipment. With the explosion of industry orders, TOTO's revenue in the semiconductor field now exceeds its revenue from civil products, which is also a very interesting phenomenon.
Q-cloth: Quartz fiber electronic cloth, which is the third-generation low dielectric constant electronic glass fiber cloth, mainly used in ultra-high frequency and high-speed scenarios such as AI servers and high-speed optical modules.
Huanshi: So essentially, TOTO is now a semiconductor material company, right? Bathroom fixtures are no longer its main business.
Guan Jian: You could say that. The secondary market used to consider it a civil product company, so its valuation was not high. But recently, the market discovered that TOTO is actually a supplier of core semiconductor materials, with high monopoly power and profits, which is why the valuation logic has changed.
There are many similar companies, such as Ajinomoto. Many people think Ajinomoto only produces MSG, but in fact, its ABF material is an indispensable raw material for high-end packaging substrates, holding an almost monopolistic market share in the global market. After the explosion in demand for GPU chips, Ajinomoto's related business has also ushered in rapid growth.
ABF material: A high-performance insulating film material developed by Japan's Ajinomoto. ABF film is a key insulating material in high-end chip packaging, widely used in the manufacturing of FC-BGA packaging substrates for CPU, GPU, and AI chips.
Overall, over the past year, most leading enterprises in the upstream semiconductor supply chain have seen improving business, and the industry prosperity is very high. The above-mentioned ones are just a few typical cases.
Structural Contradictions of Domestic Substitution: "Involution at the Low End, Choke Points at the High End"
Huanshi: In recent years, the industry has always been under the broader background of domestic substitution. Most of the companies you just mentioned are overseas manufacturers. I would like to ask Mr. Guan, according to your observation, how has the localization rate of domestic upstream equipment, components, and materials progressed in recent years? Has a group of companies quietly emerged by taking advantage of this trend?
Guan Jian: A few years ago, the localization base for core components and consumables was very low; precisely because the base was small, the growth in recent years has been particularly eye-catching. But overall, the localization rate in the high-end field remains low, especially for categories like photoresist, grinding slurry, and polishing slurry. There is still a lot of room for improvement in the localization level. Once there are problems in the supply chain for polishing slurry and photoresist, it will face significant risks.
Huanshi: But at the same time, we can also see that some tracks have achieved technological breakthroughs, domestic enterprises are very enthusiastic about entering the market, and there has also been a phenomenon of rushing in and overcapacity. How do you view this issue?
Guan Jian: In tracks like silicon carbide (SiC), as well as the mid-to-low-end consumables field, the degree of involution is already very high. In the early stage, market investment enthusiasm was high, and a large amount of capacity was implemented, but the actual downstream demand failed to keep pace, resulting in an inverted supply and demand situation where supply exceeds demand. However, this overcapacity is mainly concentrated in the mid-to-low-end market; the high-end segment is just the opposite, with supply falling short of demand and a very high dependence on overseas supply chains. For example, the photoresist and grinding slurry mentioned earlier, as well as equipment components like electrostatic chucks, probe cards, RF power supplies, and light sources, we still rely heavily on American and Japanese manufacturers.
Of course, there are also domestic companies that have grown quietly with outstanding performance. A typical example is Hangzhou Cobetter, which focuses on high-end filter cartridges in the semiconductor and biopharmaceutical fields, similar to water purification filter cartridges, but with strict application scenario requirements and very high product profit margins. The company's annual revenue is in the billions of CNY, and its net profit can exceed CNY1 billion.
Huanshi: In my impression, the leader in this track is Entegris. Has Cobetter already seized a part of its market share?
Guan Jian: Yes, Cobetter has already captured a portion of the share. The largest global supplier in this field is Entegris, followed by Purolite in the United States. European and Japanese manufacturers have a very low share, and the global market is monopolized by these two American companies.
Huanshi: So is this track now considered a red sea market?
Guan Jian: No. Even though Cobetter has not yet fully achieved domestic substitution in some high-end categories, the gross profit margin of the track can still reach 60%-70% or even higher, which is far from being a red sea.
Huanshi: In other words, the number of domestic companies that can currently produce qualified products is limited, and fierce competition has not yet formed among domestic manufacturers?
Guan Jian: There are quite a few companies entering the market, but there is still a big gap in product quality compared to the international top level.
Huanshi: In your view, which segments in the wet electronic chemicals track have already entered the red sea stage?
Guan Jian: For bulk categories with low purity requirements, the profit margin here is already very thin, basically equivalent to ordinary industrial products. However, formulated wet electronic chemicals still highly rely on imports. Many high-end chemicals related to photoresist supporting agents, electroplating, and grinding still need to be procured from Japan, which is the core direction for domestic substitution.
Huanshi: Whether it is investors or local governments, they often encounter a dilemma when supporting the industry: in the early cultivation stage, the market scale is not large, there are few practitioners, and corporate profits are acceptable. But when capital is invested and fixed assets are implemented, two or three years later, a large number of players suddenly flood into the market, corporate overcapacity occurs, and they fall into continuous losses. From your perspective, in the early research and judgment stage, what judgment methods can be used to avoid such risks as much as possible?
Guan Jian: First, find the right people. Products in the high-end semiconductor field cannot be made by just finding anyone to do it. Core personnel must be deep cultivators in the industry, with years of practical accumulation in top large factories, and truly understand technology and processes. Second, do not pin your hopes on flood-style investment. The semiconductor industry is a track that requires a steady stream of continuous effort and sufficient patience. In the past, when we developed industries like photovoltaics, LEDs, and panels, we could rely on large amounts of capital to quickly promote the formation of the industry. But semiconductors are different; simply throwing money makes it difficult to produce results quickly.
A large amount of capital has been invested in the semiconductor field domestically, but there is generally a lack of long-term patience. Coupled with the fact that many local governments want to layout the semiconductor supply chain locally, it has caused a lot of redundant construction, further exacerbating overcapacity in the mid-to-low-end. Technological iteration itself requires a long cycle. The path that overseas companies took 20 years to complete might be compressed to 12 to 15 years for us, but we cannot expect to achieve a breakthrough in three to five years.
Will the Popularization of Advanced Packaging Reshape the Landscape, or Will the Strong Remain Strong?
Huanshi: Mr. Guan himself has a professional background in the packaging field. I would like to invite you to share from an industry perspective whether the rapid popularization of advanced packaging has redefined the standards for upstream materials, components, and equipment? What brand-new changes has it brought to the entire supply chain? Now that the industry is rushing to launch advanced packaging projects, will it ultimately reshape the supply chain landscape, or bring brand-new opportunities for new entrants?
Guan Jian: The reconstruction of the upstream supply chain by advanced packaging is an inevitable trend. Simply put, the core logic of advanced packaging is to apply the precision processes of front-end wafer manufacturing downstream to the packaging stage, which is also the most essential difference between it and traditional packaging. The landscape of core players in the track has also changed completely as a result: in the past, the iterative R&D of packaging technology was basically led by traditional OSAT (Outsourced Semiconductor Assembly and Test) companies; whereas today, the core technological iteration of advanced packaging is mainly driven by leading wafer foundries.
Mainstream advanced packaging technologies we are familiar with, such as CoWoS and InFO, are all done by TSMC, not traditional OSAT companies. TSMC's 2.5D and 3D advanced packaging technologies are refined and polished by its front-end wafer manufacturing technical team. The equipment and material requirements corresponding to the entire process system are completely different from traditional packaging. For example, the Micro bump process and Hybrid Bonding technology have never appeared in traditional packaging.
Along with the changes in processes and demands, the technical standards and performance requirements for core packaging materials such as die attach adhesive, special gas chemicals, epoxy resin, and underfill have undergone subversive changes. It is clear that the currently hot 2.5D and 3D advanced packaging have been separated from traditional packaging into two completely independent tracks; they are no longer the same thing.
Development History of Advanced Packaging Technology (Image source: Semiconductor Comprehensive Research)
Huanshi: Will the explosion of advanced packaging drive a reshuffle of the upstream supply chain? Which companies will benefit in the long term, and which companies will face industry shocks?
Guan Jian: This round of technological iteration has indeed given a group of companies that have mastered new technologies the opportunity to break through, but traditional industry leaders are also completing track switching by updating products and iterating technologies. Many advanced packaging equipment are now being made by traditional packaging companies. For example, old-line packaging equipment manufacturers such as ASMPT, K&S, and Hanmi have all entered the advanced packaging equipment track. Of course, advanced packaging has also brought opportunities to some unknown small companies. Companies like EVG in Austria and BESI in the Netherlands are emerging suppliers that have rapidly risen in recent years relying on the advanced packaging track.
Geographical Distribution of Semiconductor Business Revenue of the Top Three Packaging Equipment Companies (Image source: Semiconductor Comprehensive Research)
Huanshi: So traditional large companies might just be a bit slow to react, but in reality, they will still be the mainstream force in the advanced packaging track?
Guan Jian: Actually, their reaction is not slow. When TSMC is developing a new generation of advanced packaging technology, it will connect with the core supply chain in advance, clarify the development roadmap for the next few years, and link upstream suppliers for synchronous R&D and supporting upgrades. Those who can deeply bind with TSMC and participate in forward-looking R&D are basically leading suppliers in the industry. The illusion of "lagging reaction from large companies" arises because the public only sees the results when new technologies and equipment are officially commercialized, not because they are slow. Of course, technological changes will certainly bring new opportunities to some new small companies, but the iteration of traditional leaders is continuous, so it is not so conspicuous.
Huanshi: In other words, a single breakthrough in new technology or new process is difficult to completely rewrite the original industry landscape.
Guan Jian: From the product level, there will be earth-shaking changes; but from the perspective of the company level, market entities are still dominated by traditional industry leaders. The semiconductor belongs to the traditional manufacturing industry, unlike the Internet, there are not so many disruptive things. Generally speaking, the strong remain strong for the leaders.
Overseas Giants Target the High-End Market, with Orders Exceeding Supply
Huanshi: A group of domestic start-ups and gradually growing local manufacturers all hope to seize the development opportunities brought by the new domestic capacity. But overseas giants will not easily give up the Chinese market and will adopt various strategies to block domestic manufacturers. Combining your data models and industry data observed through long-term tracking, in which links is the competition in this game between domestic and foreign manufacturers the most fierce? Is there any relevant data that can reflect the situation of this game?
Guan Jian: It is difficult to directly see the full picture of this game solely relying on financial data. However, at present, domestic highly values supply chain security and continues to support the local supply chain, which has intensified industry competition. In the mid-to-low-end track, domestic manufacturers are developing well, and their market share has reached a relatively high level; but in the high-end field, it is still dominated by overseas equipment and material manufacturers. Domestic enterprises still lack competitiveness and have a high dependence on overseas supply chains.
The strategy of overseas leading suppliers is also very clear. They focus their resources on high value-added high-end products. Under the premise of not being affected by export controls, they actively participate in the Chinese market and strategically give up some low-end markets for domestic suppliers to take. It is enough for them to hold on to their most profitable high-end business.
Cumulative Semiconductor Business Revenue Data for the First Half of 2026 (Image source: Semiconductor Comprehensive Research)
Huanshi: So it means that both sides may have both offensive and defensive strategies.
Guan Jian: For example, Advantest will not invest resources to compete for the low-end market. The orders in the high-end market are already enough to support its business, and its operating conditions are very good. Another example is ASML; their capacity for 2028 has already been fully booked, so they will not and do not need to engage in a price war.
Huanshi: Procuring such overseas equipment is different from procuring domestic equipment. For domestic equipment, you can still negotiate free trials and long-term testing; for overseas large companies, you basically have to pay first before there is a possibility of getting a production scheduling slot.
Guan Jian: It is already good enough if they can accept your order. The delivery time for Advantest's 93000 testing machine was as long as 20 months at the most tense time, and now the delivery time is still over half a year. It is already very rare for manufacturers to be willing to accept orders and supply goods. There are too many orders, and the capacity simply cannot keep up.
AI Breaks Cyclical Rules, Industry Will Continue to Grow
Huanshi: The semiconductor cycle is driven by the AI industry, and the industry cycle in the past two years has changed greatly compared to the past. I would like to ask you, when tracking the industry cycle, which core indicators do you usually refer to? Combining these indicators, where is the current semiconductor in the cycle? Has the industry upward trend brought by this round of AI already peaked, or is it still in the rebound stage?
Guan Jian: It is now very difficult to make accurate predictions. Since last year, the development of AI has broken the past operating rules of the semiconductor industry supply chain. The experience accumulated over the past decade or so is no longer applicable today. I can only make a qualitative guess. The long-term upward trend of the AI industry is certain. Bubbles will definitely appear in the process, and bubbles will also burst, but the industry will continue to develop after the bubbles are cleared. Moreover, AI's demand for chips far exceeds previous empirical cognition, so the demand for future capacity will surge. In Jensen Huang's words, global wafer capacity needs to be increased by 5 times in the future to match the demand.
Global Semiconductor Device Market Prosperity Data Table (Image source: Semiconductor Comprehensive Research)
In the long run, I am very optimistic about the entire semiconductor market in the next 10 years. The industry is currently in a step-like growth process. I judge that this round of growth will last at least until 2028. After 2028, it may peak periodically and see a slight pullback. However, from a macro perspective over a ten-year cycle, the industry will still maintain rapid growth, with an annualized growth rate of around 10%.
Huanshi: What is the core logic supporting you to make this judgment?
Guan Jian: AI. AI's demand for computing power far exceeds everyone's imagination. Rather than meaningless guessing, if you seriously deduce with logic, you will find—the Internet once crashed in 2000, but after 2000, the domestic and foreign Internet instead ushered in large-scale development. We cannot predict when a crash will occur, but the development of AI is unquestionable. As long as it continues to develop, the demand for chips will continue to rise. We can refer to the Internet back then. Is everyone using AI now, and is the usage growing larger and larger, and are the tokens burned every day increasing? If so, we have no reason to doubt the development of AI.
Huanshi: Now many work scenarios are already inseparable from AI.
Guan Jian: In the past, watching movies only required transmitting ready-made data, but now every time you ask AI a question, the backend has to compute in real time. AI services have turned into a heavy-asset model, and marginal costs are no longer low. In the future, the use of AI will make the consumption of chips reach an unimaginable scale. The existing capacity is far from enough; currently, it can only meet the needs of the mobile Internet era, while AI's demand for chips far exceeds that of the mobile Internet.
Future Upstream Industry: Will "Industry Common Sense" Be Subverted?
Huanshi: Looking ahead to 2026-2028, what "industry common sense" that everyone currently considers in the upstream semiconductor supply chain do you think might be subverted in the next three years?
Guan Jian: In the next three years, there will be no subversive breakthroughs in the upstream semiconductor industry, and the true core common sense of the industry will still all hold. The so-called common sense for many people is actually a cognitive illusion caused by a lack of understanding of the industry. In the future, only these illusions will be shattered, and the underlying logic of the industry will not change.
Huanshi: Based on your judgment that the semiconductor industry will remain positive in the long term over the next ten years, what impact will it have on your personal work, career, and business choices?
Guan Jian: This judgment will strengthen my confidence in the semiconductor industry—deeply cultivating the semiconductor is the right thing to do. Our generation can no longer possibly build large models, but for us, the semiconductor is one of the few tracks in traditional high-end manufacturing that still has huge growth potential in the next ten years.
Forecast Trend of the Global Semiconductor Device Market for the Full Year 2027 (Image source: Semiconductor Comprehensive Research)
Huanshi: So can we also call on young friends to join this industry?
Guan Jian: The industry will not provide absolute guarantees. From a macro perspective, the future development of the industry will definitely be concentrated in high-tech tracks. Of course, the high-tech field is not limited to semiconductors. For young people, there will be relatively more opportunities in the high-tech industry, but the most important thing is still personal choice and development.
Huanshi: Thank you, Mr. Guan, for bringing such a substantial and informative share. Finally, you can briefly introduce on our "Chip Unveiled" column how everyone can access your upstream semiconductor database, so that listeners in need can contact you.
Guan Jian: Thank you, Huanshi, for giving me this opportunity to share. My WeChat official account is called Semiconductor Comprehensive Research. I will organize industry data into articles and publish them on the official account every week. These articles are free. If you need the original data drafts and in-depth consulting services, you can join my Knowledge Planet. The joining method is at the end of each official account article. There is a member group in the Planet, and all database drafts are placed in the cloud drive, which members can check and use at any time. In addition, I also provide one-on-one free phone and WeChat consulting services to members.
This is my service model: providing data and consulting for practitioners with in-depth professional needs. Every week at 19:30 in the evening, I will also live stream on Tencent Channels to interpret the latest industry dynamics and conduct industrial science popularization. You can continue to pay attention.
Huanshi: Thank you, Mr. Guan. Many fans of our column are investors and industry researchers. If they need this set of systematic industry materials, they can contact Mr. Guan's platform. Thank you again for Mr. Guan's wonderful share, and we look forward to you visiting the "Chip Unveiled: Experts Talk Chips" column often.
Guan Jian: Okay, looking forward to more exchanges in the future.
END
Interview | Huanshi Editor | Xiaoqie Review | Huanshi