ABF is being frantically snapped up
Recently, NVIDIA announced cooperation with supply chain partners including Ibiden, Samsung Electro-Mechanics, Unimicron, Amkor, KYEC (King Yuan Electronics), and SPIL (Siliconware Precision Industries) to expand relevant manufacturing capacity.
Among them, the product substrate capacity is planned to be increased to 2.5 times the current level in the next three years. Amkor plans to increase its interposer wafer capacity by more than 300% by the end of 2028, while SPIL plans to increase the capacity of optical engines for Co-Packaged Optics (CPO) to 15 times the current level by the middle of 2027. With the release of this news, the continuous expansion of AI computing power infrastructure has ignited supply chain competition, which is further extending from core chips such as GPUs to key links including advanced packaging, testing and high-speed optical interconnection.
In this context, ABF has even become a must-contested ground for major chip manufacturers.
ABF Delivery Lead Time Keeps Extending
In fact, as we can see, the supply pressure of ABF substrates is further intensifying.
According to the data on delivery lead time of key components for AI infrastructure compiled by TrendForce, the lead time of ABF substrates has been extended to 48 to 56 weeks, which is 4 to 4.7 times the 12-week benchmark under the supply-demand balance state. In other words, a new order may take up to about 13 months from placement to delivery. In comparison, the lead time of HDD is 50 weeks, DRAM is 20 weeks, and MLCC is 30 weeks, which clearly shows the tension of the AI infrastructure supply chain.
ABF substrate is an important connection carrier between high-performance chips and motherboards, widely used in products such as CPUs, GPUs and AI accelerators. As AI chips develop toward higher performance, larger packaging size and more complex interconnection, substrates need to support higher-density circuits and more layers of wiring, which increases manufacturing difficulty. At the same time, the demand of AI servers for components such as storage, network and power management is growing simultaneously, making the supply chain pressure no longer limited to a single link.
TrendForce's data also shows that among the six types of key components for AI infrastructure it tracks, except for GPUs, the remaining five types are in shortage. Among them, ABF substrates, HDD and DRAM are listed as severely in short supply, and MLCC and enterprise-level SSD are also facing insufficient supply. Even if the GPU delivery lead time is in the supply-demand balance range of 20 to 30 weeks, the shortage of other key components may still slow down the production and delivery of the entire set of AI servers.
The problem is that new capacity cannot fill the gap immediately. The expansion of high-end substrate production lines involves plant construction, equipment procurement, process debugging and yield improvement. The investment of capital does not mean that the capacity can be released quickly. Currently, the demand for AI is growing rapidly, and the previous capacity expansion plans of suppliers may not be able to keep up in time, so the supply-demand mismatch continues to exist.
What is more noteworthy is that the factors restricting supply also extend to semiconductor equipment components. Reports point out that the maximum delivery lead time of some key equipment components manufactured in Japan reaches 40 months; some deposition equipment parts that could originally be delivered in about 4 months now need to wait for about 10 months. Even if some components of packaging equipment have been transferred to local supply in South Korea, the delivery lead time is still about 50% longer than in the past.
This has formed a warning cycle: the growth of AI chip demand drives the capacity expansion of substrates, storage and other components; but capacity expansion itself requires semiconductor manufacturing equipment, and equipment manufacturers are also affected by the shortage of key parts. The stronger the downstream demand, the more the supply chain needs to expand capacity, but the more likely it is to be restricted by the delivery cycle of upstream equipment.
In addition, the report also points out that factors such as increased capital expenditure, slowing production efficiency improvement, and more wafer input occupied by HBM production also put pressure on the supply expansion of the storage industry. The current supply shortage cannot be solved immediately by simply increasing investment, but involves the capacity, equipment and manufacturing efficiency of multiple links.
The continuously extending delivery lead time is also changing the transaction mode of the supply chain. The report points out that the price of high-specification FC-BGA substrates is facing upward pressure, and the importance of long-term supply agreements is also increasing. Customers hope to reduce the risk of future supply shortages by locking in quantities in advance and signing long-term contracts; suppliers with stable capacity and the ability to deliver on schedule will have the opportunity to gain stronger bargaining power.
Taking Samsung Electro-Mechanics as an example, the report points out that the operating rates of its FC-BGA substrate and MLCC production facilities are almost full. In the case that capacity is difficult to expand rapidly, manufacturers with stable delivery capabilities will more easily obtain long-term orders and increase the revenue proportion of high value-added products.
For the ABF industry, what really deserves attention is not only how many weeks the delivery lead time reaches, but also when the new capacity will be put into use and whether it can match the continuously upgraded specification requirements of AI chips. If the demand continues to grow while equipment, materials and manufacturing capabilities cannot be improved simultaneously, the supply-demand tension of ABF substrates may last for a longer time. However, the delivery lead time data reflects the current state of the supply chain, which does not mean that the delivery lead time of all specifications and all customers is exactly the same.
In this demand scenario, how to ensure the supply of ABF has become the top priority of major chip companies. This has also triggered a capacity battle.
Three Chip Giants Lock in Capacity
According to the Economic Daily, the three major chip manufacturers Intel, NVIDIA and AMD have reportedly paid deposits totaling more than NT$100 billion to pre-order high-end IC substrate capacity. At the same time, the three major Taiwanese substrate manufacturers Unimicron, Kinsus and Nan Ya PCB are accelerating capacity expansion, with the total investment in plant construction announced this year reaching NT$161.2 billion, hitting a new high.
Behind this capacity battle around ABF substrates is the reality of continuously growing demand for AI chips and the difficulty of rapidly expanding high-end substrate capacity in the short term. For chip manufacturers, the performance of GPUs and AI accelerators is certainly important, but whether products can be smoothly packaged, tested and finally delivered also depends on whether the upstream supply chain can provide sufficient capacity. When the supply of high-end substrates tightens, locking in capacity in advance has become an important means to ensure product delivery.
According to industry estimates cited in the report, Intel's long-term pre-purchased substrate capacity is about 3 billion US dollars. AMD previously announced an investment of 10 billion US dollars in Taiwan, and the market speculates that about 1 billion US dollars of it may involve the substrate supply chain. NVIDIA is also speculated by the market to have invested about 1 billion US dollars in Taiwanese substrate manufacturers, and further increase is not ruled out in the future. According to these estimates, the total amount of funds involved in the substrate industry by the three manufacturers is about 5 billion US dollars. However, these figures are industry rumors or market estimates, not the officially confirmed order and advance payment amounts of the enterprises.
Customers paying advance payments and locking in capacity in advance also provide motivation for substrate manufacturers to expand production. The construction of high-end ABF substrate production lines requires a large amount of capital investment, and equipment procurement, plant construction and subsequent process debugging all take time. For chip manufacturers, advance layout helps to reduce the uncertainty of future supply; for substrate manufacturers, the growth of customer demand provides a basis for capacity expansion.
The three major Taiwanese substrate manufacturers have begun to increase capital investment. Unimicron has added NT$19.7 billion in capital expenditure, bringing its total capital expenditure this year to NT$53.7 billion, and at the same time added NT$17.6 billion in long-lead-time equipment procurement orders. The company also acquired the Hukou plant and office of Bridgestone Taiwan for about NT$10.08 billion to pre-layout for medium and long-term operations and high-end substrate capacity expansion.
Kinsus continues to transform into high-end substrates and gradually fade out of traditional PCB production. The company uses NT$23.5 billion in capital expenditure to meet the demand for ABF substrate equipment expansion in the next three years, of which NT$8 billion is planned to be spent in 2026, and about NT$6 billion is used for ABF substrate investment. In addition, the company has added a new project capital expenditure budget of NT$19.6 billion. Nan Ya PCB also plans to invest NT$46.8 billion to build a smart factory that produces large-size, high-layer-count high-end substrates.
The capacity expansion direction of the three manufacturers reflects that the demand for AI chips is pushing the substrate industry into a new round of investment cycle. However, capacity expansion does not mean that the supply shortage can be alleviated immediately. Plant construction, equipment delivery, process debugging and yield improvement all take time, and whether the new capacity can meet the demand in time also depends on the customer's product specifications and mass production progress.
From chip giants locking in capacity in advance to substrate manufacturers sharply increasing capital expenditure, the impact of AI infrastructure construction is gradually transmitted to the upstream of the supply chain. As high-end chips continue to raise requirements for packaging capabilities, ABF substrates are no longer just supporting materials in chip packaging, and their supply capacity is becoming an important factor affecting the production and delivery of AI chips.
Samsung Electro-Mechanics Is the Target of Competition
While AI chip giants are scrambling to lock in ABF substrate capacity, Samsung Electro-Mechanics is becoming an important participant in supply chain competition. On October 21, Samsung Electro-Mechanics announced that it has established a strategic partnership with NVIDIA. The two sides will jointly invest to expand special production facilities in Sejong and Busan, South Korea, to supply ABF packaging substrates for NVIDIA's next-generation AI processors and data center high-speed network switches. Samsung Electro-Mechanics plans to start delivery in 2026, with the annual delivery volume expected to grow at a double-digit rate for many years in the future. By 2028, the capacity allocated to NVIDIA will account for a considerable part of the company's total annual substrate capacity.
This means that the cooperation between NVIDIA and Samsung Electro-Mechanics is not an ordinary supplier procurement, but a combination of capacity expansion and long-term supply demand. For NVIDIA, the construction of AI infrastructure requires continuous expansion of the supply scale of chips and data center network equipment. Locking in high-end substrate capacity in advance helps reduce the impact of supply chain bottlenecks on product delivery. Debora Shoquist, NVIDIA's Executive Vice President of Operations, also said that the speed of AI infrastructure construction requires the company to expand manufacturing capacity globally, and cooperation with Samsung Electro-Mechanics will further expand its substrate supplier network.
For Samsung Electro-Mechanics, this means an important opportunity to enter the high value-added AI semiconductor substrate market. The company will set up special production lines at its main production bases in Sejong and Busan, South Korea. The supply scope covers not only packaging substrates required for NVIDIA computing chips, but also network switch substrates that realize high-speed chip interconnection inside data centers. With the simultaneous development of AI computing and high-speed networks, substrate manufacturers need to meet the requirements of high-density interconnection for both computing chips and network equipment.
The real challenge lies in technology. Samsung Electro-Mechanics pointed out in its announcement that as the size of packaging substrates expands and embedded devices increase, the manufacturing complexity also rises. In order to reduce signal loss, substrates need to achieve ultra-fine circuits; in order to ensure mass production yield and long-term reliability, they must also have advanced warpage control and flatness management capabilities.
In addition, the high-speed operation of next-generation AI chips puts forward higher requirements for power supply stability. Samsung Electro-Mechanics plans to give full play to its technical accumulation in the field of multilayer ceramic capacitors (MLCC) and silicon capacitors, embed passive components into the substrate structure to suppress electrical noise generated by high-speed operation, and improve power management efficiency. This route that combines packaging substrate manufacturing capabilities with capacitor technology is also one of the competitive advantages that distinguish Samsung Electro-Mechanics from pure substrate suppliers.
From this cooperation, we can see that the competition in the AI chip industry is extending to the packaging and interconnection links. While GPU performance is important, whether chips can be mass-produced stably also depends on whether substrates can meet the requirements of larger size, higher-density circuits and stricter reliability. NVIDIA locks in capacity in advance, and Samsung Electro-Mechanics expands its high-end substrate business with the help of long-term supply cooperation, so both sides get what they need.
With the continuous expansion of AI infrastructure, ABF substrates have become a non-negligible link in the chip supply chain. Whether Samsung Electro-Mechanics can turn this cooperation into a long-term competitive advantage still depends on the construction progress of special production lines, mass production yield, and whether it can continuously meet NVIDIA's ever-improving product requirements.
Final Note
If you shift your sight from ABF substrates themselves, you will find that this capacity competition reflects a change in the competition logic of the AI industry: in the past, the industry paid more attention to who can design chips with stronger performance and who can take the lead in advancing advanced processes; now, with the continuous improvement of the scale and complexity of AI chips, the factors that determine the competition results are expanding from single chip performance to the collaborative capability of the entire manufacturing system.
This also means that the supply chain can no longer simply regard capacity expansion as adding plants and procuring equipment. For key links such as ABF substrates, real competitiveness comes from the long-term running-in among materials, equipment, processes, yield and customer demand. Even if capital investment continues to increase, if the delivery of key equipment, technology accumulation and talent reserve cannot keep up simultaneously, the new capacity may not be converted into effective supply in time.
For chip manufacturers, locking in capacity in advance can ease the current supply pressure, but cannot replace the capability building of the supply chain itself; for substrate manufacturers, the high-growth opportunities brought by AI are worth seizing, but they also need to be alert to equating short-term demand directly with long-term growth. AI infrastructure investment will eventually experience cyclical changes, and how to strike a balance between expansion speed and return on investment will test the judgment of every supplier.
From a longer-term perspective, what the ABF industry really needs is not just more capacity, but a more resilient supply system that can better adapt to technological iteration. As the connection between chip design, advanced packaging and substrate manufacturing gets closer and closer, suppliers also need to move from passively receiving orders to participating in customers' technical planning and product development at an earlier stage.
In the AI era, what is scarce is never just a certain material or a certain technology, but the ability to truly connect technology, manufacturing and large-scale delivery. The rush for ABF is just a visible section of this industrial competition.
*Disclaimer: This article is original by the author. The content of the article is the author's personal opinion. The reprint of Semiconductor Industry Observation is only to convey a different point of view, and does not mean that Semiconductor Industry Observation agrees with or supports this point of view. If you have any objections, please contact Semiconductor Industry Observation.
This article is from the WeChat official account "Semiconductor Industry Observation" (ID: icbank), written by the editorial department, and authorized for release by 36Kr.