Introduction
Supply of ABF packaging substrates is concentrated in a small group of manufacturers headquartered in Japan, Taiwan, South Korea and Austria, and the tier that can build large-body, high-layer-count units for server processors, AI accelerators and switch ASICs is narrower still. Ibiden, Shinko Electric Industries and TOPPAN form the Japanese base; Unimicron Technology, Nan Ya PCB and Kinsus Interconnect Technology make up the Taiwanese cluster; Samsung Electro-Mechanics, LG Innotek and Daeduck Electronics represent South Korea; and AT&S is the only European-headquartered supplier of scale, with substrate plants in China, Malaysia and Austria. Mainland Chinese companies, led by Shennan Circuits, are building FC-BGA capability with support from domestic chip designers that want local supply, while Taiwan-listed Zhen Ding Technology is adding ABF substrate capacity at its sites in mainland China and Taiwan. Ownership has also changed: a consortium led by a fund managed by JIC Capital, with Dai Nippon Printing and Mitsui Chemicals, completed its tender offer for Shinko in March 2025, and TOPPAN opened its first FC-BGA plant outside Japan, in Singapore with Broadcom, in September 2026.
Competition is decided first by qualification rather than price. A new substrate source for a server CPU or GPU passes through design-in, reliability testing and OSAT assembly trials before volume orders, and the cycle restarts with each processor generation. Incumbency on a product family is therefore valuable, and entry is slow even for well-funded firms; Shennan Circuits told investors in January 2025 that customer certification for its packaging substrates takes longer than for its other PCB and substrate products.
Technical capability is the second axis. AI accelerators and chiplet processors push substrates toward larger body sizes, more build-up layers, finer SAP line/space, denser bump pitch and tighter warpage control, and LG Innotek exhibited an AI-accelerator FC-BGA with a side length exceeding 100 mm in September 2026. Yield on these units sets cost, because one defect in any build-up layer scraps the whole substrate. Suppliers with mature control of laser via formation, copper plating and lamination on thick, low-CTE cores gain both margin and allocation advantages from the same tooling.
The third axis is commercial: capacity reservations and customer co-development. High-end capacity is increasingly underwritten by buyers. AT&S linked its June 2026 Kulim expansion to long-term commitments from AMD and another leading technology company; Samsung Electro-Mechanics cited customer funding support and mid- to long-term volume commitments for its September 2026 FC-BGA investment; and NVIDIA was reported in October 2026 to be securing substrate capacity in Taiwan, Japan and South Korea through reservation agreements and upfront payments. Suppliers also work with chip designers, TSMC and OSATs at the package-definition stage, matching core thickness, stack-up and power delivery to the die layout of CoWoS and multi-die designs, and they compete on roadmaps for glass core, embedded-die and coreless structures. Secure access to ABF film from Ajinomoto Fine-Techno and to T-glass cloth, which remained tight in 2026, gives an added edge to suppliers with long upstream relationships.
Key Player Positioning
The positioning assigns each supplier to one of three groups, Leaders, Challengers and Niche Players, according to its presence in high-layer, large-body production, its depth of qualification on leading processor and accelerator programs, the breadth of its manufacturing footprint and its ability to fund capacity ahead of demand.
Ibiden, Unimicron Technology, Shinko Electric Industries and AT&S are the Leaders. Ibiden is the reference supplier for server and AI GPU substrates, a long-standing source for Intel processors and a member of the TSMC 3DFabric Alliance. It opened its Ono plant in Gifu Prefecture in October 2025 with staged mass production for AI servers, and in February 2026 approved electronics capital spending of about ¥500 billion over fiscal 2026 to 2028, centered on further capacity at Ono after a first phase at Gama. Unimicron is the largest Taiwanese ABF producer and supplies NVIDIA, Intel and Apple. Its Guangfu plant entered mass production in late 2024 for high-end AI GPU work, part of its Yangmei capacity was redirected to CoWoS customers in 2025, and in July 2026 it raised its 2026 capital budget with most of the spending assigned to ABF substrates. Much of its ABF output ships under long-term contracts, which gives volume visibility in exchange for less pricing flexibility. Shinko brings decades of high-layer server CPU experience and, since its 2025 privatization, a shareholder base that includes the materials and printing groups Mitsui Chemicals and Dai Nippon Printing; it showed a 22-layer glass-core build-up substrate in July 2026.
AT&S qualifies as a Leader through its position on AMD data center processors, served from Kulim, where high-volume manufacturing began in May 2025, alongside its long-established substrate operations in Chongqing. It extended the Kulim build-out in June 2026, added Marvell as a long-term customer for advanced IC substrates in September 2026, and opened an IC substrate plant and competence center in Leoben, Austria, in June 2025, the first such site in Europe.
The Challengers are Samsung Electro-Mechanics, Nan Ya PCB, Kinsus Interconnect Technology and TOPPAN. Each has qualified high-end products and meaningful capacity but trails the Leaders in either breadth of AI-class programs or scale of large-body output. Samsung Electro-Mechanics described itself in September 2025 as the only Korean company mass-producing server FC-BGA, runs FC-BGA lines in Korea and Vietnam, and is developing 2.1D substrates that connect chips without a silicon interposer. Its September 2026 investment plan for Sejong and Vietnam, whose Sejong portion is the largest single-product investment in its history and is due to reach mass production in September 2028, would move it closer to the Leaders if customer volumes follow. Its glass substrate program, run partly through a joint venture with Dongwoo Fine-Chem, targets glass core material in the second half of 2027 and finished substrates in 2028. Nan Ya PCB, part of the Formosa Plastics group, has heavy exposure to high-end ABF products and keeps part of its output outside long-term contracts, which analysts expected in 2026 to let it gain more than peers from rising prices; in 2026 it also outlined a smart factory dedicated to large-size, high-layer substrates.
Kinsus, affiliated with the Pegatron group, was reported in January 2026 to have reserved most of its advanced lines for AI customers, and a March 2026 report said it aims to match rivals such as Ibiden in ABF substrates by 2028. TOPPAN has a concentrated but valuable position in large, high-layer substrates for networking and AI ASICs, with Broadcom as its largest FC-BGA customer. Its Niigata base is now complemented by the Singapore plant of Advanced Substrate Technologies, its joint venture with Broadcom, which opened in September 2026 and targets high-value manufacturing by late 2026.
The Niche Players are LG Innotek, Daeduck Electronics, Zhen Ding Technology and Shennan Circuits, whose ABF activity is either recent, narrow in application or focused on a home market. LG Innotek entered FC-BGA by acquiring a Gumi plant from LG Electronics in 2022 and converted it into a highly automated "Dream Factory" that began mass production in February 2024. It ships PC FC-BGA substrates to North American technology customers, plans to move into server and AI products, and in 2026 announced plans for a substrate plant in Haiphong, Vietnam, whose product range includes FC-BGA. Its exhibit of a 100 mm-class AI-accelerator substrate in 2026 marks the boundary it intends to cross toward the Challenger group. Daeduck Electronics competes mainly in memory and mobile package substrates, with a smaller FC-BGA presence, so its ABF positioning rests on selected customers rather than flagship processor programs. Zhen Ding Technology is the largest PCB group in the reference set and brings scale in manufacturing and procurement; its June 2026 entry into the NVIDIA MGX ecosystem as a board supplier shows access to AI server customers that it can extend into substrates, though its ABF portfolio is still being built.
Shennan Circuits describes itself as the largest Chinese-owned packaging substrate supplier. Its Guangzhou project brought phase-one capacity online in the fourth quarter of 2023, and by May 2026 it reported mass production of FC-BGA at 22 layers or fewer, with products of 24 layers or more in development and sampling. Its prospects depend on domestic CPU, GPU and AI accelerator demand and on export-control conditions rather than on qualification at U.S. processor vendors.
Movement between the groups is most likely where capacity and qualification arrive together. Samsung Electro-Mechanics and Kinsus have the clearest paths to the Leader tier if their new lines are filled by AI accelerator programs, while LG Innotek and Shennan Circuits could rise within the Challenger group if their large-body products pass customer certification.
Competitor Dashboard
The dashboard compares the twelve suppliers on four qualitative dimensions, with ratings that follow the Leader, Challenger and Niche Player grouping. Product Portfolio reflects layer-count and body-size range, Regional Presence reflects manufacturing footprint, Strategic Alliances reflects customer and partner ties, and Industry Experience reflects years in high-end FC-BGA production.
Competitive forces in the ABF packaging substrate industry are shaped by an unusual combination of concentrated upstream materials, concentrated downstream buyers and a manufacturing tier that requires heavy capital and long customer qualification. The balance of power between these groups shifts with the semiconductor cycle: in the 2023 inventory correction, buyers held the advantage and substrate utilization fell, whereas the AI-driven demand for large, high-layer-count substrates in 2025 and 2026 has moved bargaining strength toward material suppliers and the fabricators able to deliver qualified capacity. The assessment that follows reflects conditions at the end of 2026 and their likely direction over the forecast period. Ratings apply to the industry as a whole; where conditions differ sharply between mainstream products such as 4–8 Layers substrates and high-end products such as Above 16 Layers and 2.5D interposer-based substrates, the difference is explained within each force.
Threat of New Entrants
The threat of new entrants is Low. Entry requires a dedicated fab with clean-room build-up lines, laser drilling, laser direct imaging, SAP plating and inspection capacity, and the entry ticket rises with each technology node because high-layer-count products need tighter registration, finer line/space and larger panel handling. Incumbent expansion programs now run to hundreds of billions of yen or tens of billions of New Taiwan dollars for a single business line, which sets the scale an entrant would have to match. Capital alone is not sufficient. A new fabricator must also secure ABF film and T-glass allocation from suppliers that are already rationing output among existing customers, recruit process engineers with SAP experience, and achieve yields that make large-body substrates economic.
The most durable barrier is customer qualification. A CPU, GPU or AI accelerator vendor qualifies a substrate source for each product, a process that includes reliability testing to JEDEC conditions, warpage characterization and pilot production, and that typically runs across multiple quarters before volume orders are placed. Once qualified, a source tends to supply the product for its life, so entrants must win new design slots rather than displace incumbents on running products. The entrants that do appear tend to be adjacent players: Chinese PCB and substrate groups expanding from BT substrates into ABF with domestic customer support, and existing substrate makers moving into the highest-layer-count tier. Glass-core substrate developers, including the SK group's Absolics in the United States, represent a different kind of entry through a new core technology, but they remain at low-volume or pilot scale. Their eventual success would add competitors in the largest package formats rather than in mainstream products.
Government support lowers the financial barrier in some regions without removing the technical ones. Industrial policy in China, India, the United States and Europe offers capital grants, tax credits and land for packaging projects, and new packaging investments in these regions are being planned around such incentives. Even with subsidized capital, a new site must still pass the same qualification and yield thresholds as an incumbent, and the most advanced chip customers have tended to place new programs with established suppliers that are adding capacity rather than with first-time producers. Entry is therefore more plausible in the mainstream tier, where the technology is mature and customers are price-sensitive, than in the highest-layer-count products, where an entrant would compete against suppliers that have co-developed substrates with the chip designer over several generations. Chinese entrants face an additional constraint: their access to build-up film and to advanced Japanese and Western process tools depends on supplier allocation and on the possibility of tighter export rules, while limits on foundry access for China's most advanced AI chips restrict domestic demand for the highest-layer-count substrates.
Bargaining Power of Suppliers
The bargaining power of suppliers is High. Build-up film for CPU, GPU and accelerator substrates comes almost entirely from one company, Ajinomoto Fine-Techno, and production-grade low-CTE T-glass is similarly concentrated at Nittobo. Both materials are qualified at the level of the specific grade and the specific chip program, so a substrate maker cannot switch supplier without the chip customer's agreement and requalification. This concentration became visible in 2026, when film and laminate producers raised prices and allocated output among substrate makers rather than simply expanding shipments to meet orders.
Supplier power also differs by input. Copper foil, standard laminates, plating chemicals and dry films are available from several qualified sources, and their producers have less pricing power. The most acute constraint arises where a material is both effectively single-sourced and consumed in proportion to substrate area and layer count, because demand for such inputs grows faster than substrate unit volumes as AI packages become larger. Build-up film and low-CTE glass cloth meet both conditions, which explains why their suppliers have been able to raise prices and allocate output while suppliers of more standardized inputs have not.
Equipment suppliers add to this power. Laser drilling systems, laser direct imaging tools, plating lines and automated optical inspection systems for fine-line SAP come from a limited group of makers, and the simultaneous expansion programs at Ibiden, Unimicron, Samsung Electro-Mechanics, Kinsus and others have lengthened lead times for long-lead tools. Some substrate makers have placed equipment orders well ahead of building completion to secure delivery slots, which ties capital to supplier schedules.
Two factors moderate supplier power over the forecast horizon. Ajinomoto and Nittobo are both expanding capacity, and both depend on the substrate industry as their main outlet for these products, which gives them a commercial interest in avoiding supply disruption to their largest customers. Alternative suppliers of build-up dielectrics and low-CTE glass cloth are also working toward qualification, and chip designers have an incentive to sponsor second sources for risk reasons. Upstream producers also hold an information advantage: because they receive orders from every substrate maker, they can judge aggregate demand more accurately than any single fabricator and set allocation and price terms accordingly, while their own expansion schedules set the outer limit on how fast the industry can grow. Even so, the time needed to qualify alternatives means supplier power is likely to remain high through most of the forecast period, particularly for materials used in Above 16 Layers substrates and 2.5D interposer-based packages.
Bargaining Power of Buyers
The bargaining power of buyers is Medium. The buyer base is highly concentrated: a small number of CPU, GPU and accelerator vendors, a group of hyperscalers with custom silicon programs, and the foundry and OSAT assembly houses that procure on their behalf account for most high-end substrate demand. These buyers set specifications, control design wins and can shift future programs between qualified suppliers, and during downturns such as 2023 they used this position to secure price reductions and defer volume.
Conditions in 2025 and 2026 have reduced that bargaining strength in the high-end segments. Large AI and server substrates are capacity-constrained, and buyers have moved from negotiating price to securing allocation through deposits, long-term agreements and co-investment in dedicated lines, commitments that tie the buyer to a particular supplier's capacity for several years. Qualification lock-in also limits a buyer's ability to switch suppliers within a product generation. In mainstream segments, by contrast, buyer power remains strong. Substrates for notebook processors in the 4–8 Layers segment are produced by many qualified suppliers, including Chinese entrants, and PC demand is sensitive to consumer spending, so buyers retain pricing power there.
The foundry and OSAT channel adds a further dimension. TSMC, ASE, SPIL and Amkor influence substrate selection through their packaging design rules, assembly capacity and qualification data, and when they procure on a turnkey basis they act as large, informed buyers able to compare suppliers across many customers. For CoWoS-class packages the chip designer generally keeps control of substrate selection, so the assembly house's purchasing influence is greatest in mainstream flip-chip products. Hyperscalers are emerging as an additional buyer group with substantial volume and a willingness to commit capital, which strengthens their hand in allocation talks but also deepens their dependence on specific suppliers. Chip designers fund capacity through deposits and co-investment but have not moved into volume substrate fabrication themselves, so the threat of backward integration remains limited and does not add materially to buyer power.
The overall rating of Medium reflects this split: strong positions for buyers of standard substrates, limited influence for buyers of the most advanced substrates, and a shift back toward buyers if capacity additions outpace AI demand toward the end of the decade.
Threat of Substitutes
The threat of substitutes is Low. Within the FC-BGA package format, there is no commercially established alternative to an organic build-up substrate for carrying large, high-pin-count logic die onto a printed circuit board. BT-based substrates serve memory, mobile application processors and flip-chip chip-scale packages, but they do not offer the fine-line SAP capability, low-loss build-up layers and large-body flatness needed for server CPUs or AI accelerators. Silicon interposers and embedded silicon bridges increase interconnect density between die, but they sit on top of an ABF substrate rather than replacing it, so their adoption raises substrate complexity rather than displacing demand. Switching costs reinforce this position: a chip designer that moved a product from an organic build-up substrate to a different carrier technology would have to redesign the package, requalify reliability and often change the assembly flow, which is practical only at a new product generation.
The substitutes most relevant over the forecast period are technologies that move routing or structural functions out of the organic build-up stack. Fan-out redistribution layers and panel-level packaging can absorb some fine-pitch routing, which may reduce build-up layer count for certain products. Glass-core substrates, which replace the glass-fibre-reinforced organic core with a glass panel, are the more significant long-term substitute because they target the very large body sizes where organic cores face warpage and dimensional limits. Intel, Samsung Electro-Mechanics and Absolics are among the companies developing glass-core technology, and Ibiden has placed a glass core on its roadmap around 2030. These substrates still require build-up dielectric layers, so even their adoption may retain demand for ABF-type films while changing the core. Panel-based variants of interposer packaging under development at TSMC change the format of the interposer rather than the carrier beneath it, and they continue to rely on a build-up substrate to connect the module to the board. Because glass-core volume production is not expected at scale until near the end of the decade, the threat remains low for most of the forecast period but rises toward 2032 for the largest AI packages.
Competitive Rivalry
Competitive rivalry is Medium. The industry is concentrated, with a top tier of Japanese, Taiwanese, Korean and Austrian fabricators supplying the most demanding CPU, GPU and accelerator programs, and a second tier, including the major Chinese makers, competing in mainstream products and expanding upward. Rivalry in the high-end tier is driven by technology rather than price: suppliers compete on line/space, layer count, body size, warpage control and yield on products such as large 2.5D packages, and the reward is a multi-year design win. Under current capacity constraints, these suppliers are largely sold out on qualified high-end lines, which reduces price competition for those products.
Rivalry is far more intense in the standard segments. The 2023 correction left substantial idle capacity for 4–8 Layers substrates, and new capacity from Chinese makers supported by domestic chip and PC customers has added pressure on pricing for notebook and consumer products. Fixed costs are high, so utilization drives profitability, and suppliers compete to keep standard lines loaded while converting what they can to higher-layer-count output. The large expansion programs announced in 2026 by Ibiden, the Taiwanese makers and Samsung Electro-Mechanics could raise rivalry once that capacity comes online, particularly if AI demand growth slows relative to the added supply.
Ownership structures add to the competitive dynamics. Several substrate makers sit within larger groups, such as Samsung Electro-Mechanics in the Samsung group, LG Innotek in the LG group, Kinsus in the Pegatron group and Nan Ya PCB in the Formosa Plastics group, which gives them access to group capital and in some cases to related demand. Exit barriers are high because substrate lines have little alternative use, so suppliers tend to remain in the market through downturns, which sharpens price competition whenever demand weakens. Chinese makers add a further layer of rivalry because domestic chip designers increasingly prefer local sources for strategic reasons, which gives them a protected base from which to expand into higher-layer-count products. The Medium rating balances strong rivalry in mainstream products against capacity-limited, technology-led competition at the high end.
Threat of New Entrants: Low; Heavy capital needs, scarce ABF film and T-glass allocation, SAP yield know-how and multi-quarter customer qualification keep entry limited to adjacent players.
Bargaining Power of Suppliers: High; Build-up film and low-CTE T-glass are effectively single-sourced and program-qualified, letting suppliers raise prices and allocate output during the AI-driven shortage.
Bargaining Power of Buyers: Medium; Concentrated chip designers and hyperscalers hold strong positions in standard substrates, but capacity tightness and qualification lock-in limit them at the high end.
Threat of Substitutes: Low; No established alternative carries large FC-BGA logic packages; interposers sit on ABF substrates, and glass cores reach volume only near 2030.
