TSMC (TSMC) faces a two-front squeeze from Samsung (Samsung) and Intel (Intel), choosing to take the offensive rather than defend by launching what will be the largest wafer-fab expansion boom in global history. 18 12-inch wafer fabs are being built in parallel, driving capital expenditures to surge and ramping up advanced packaging capacity. Behind this semiconductor race, players in the facilities engineering sector include Hanton (2404) and Fanxuan (6196); in the equipment and materials sector include Hongsu (3131) and Zisheng (2467), which are expected to become the biggest beneficiaries.
Geopolitics and AI demand break TSMC’s monopoly pattern
For years, the global advanced wafer foundry market has been almost entirely dominated by TSMC. However, amid the explosive demand for artificial intelligence (AI), rising geopolitical risk, and ongoing pressure from the United States’s supply-chain security policies, the industry landscape is gradually loosening.
The Economic Daily News cited semiconductor supply-chain insiders in Taiwan as saying the key driving this shift is not only pricing and capacity, but also the high anxiety among global customers in the AI era over “supply-chain security” and “geopolitical risk”: if issues arise in Taiwan’s supply chain, how should companies respond? In response, major international companies are actively looking for a second source of supply.
Samsung ties up memory with foundry services, while Intel plays the “Made in the USA” card to win orders
In this wave of change, Samsung Electronics and Intel are the two most aggressive players in positioning themselves.
For Samsung, although its 2 to 4-nanometer processes still have a clear gap versus TSMC, it leverages its advantages in HBM and DRAM to roll out a “memory-bundled foundry” strategy. It grabs orders with low pricing, even pricing only based on good dies (Good Die), encouraging customers to take on part of the yield risk. AMD CEO Lisa Su previously traveled to Samsung’s Pyeongtaek plant in South Korea to deepen cooperation; meanwhile, Tesla signed an AI chip supply agreement with Samsung as early as 2025 for the period through 2033. Its new Texas plant will be dedicated to producing Tesla’s next-generation AI6 chips.
For Intel, it is pushing the policy advantage of “Made in the USA,” actively advancing 18A and 14A process technologies, and building differentiated competitiveness through advanced packaging technologies such as EMIB and Foveros. Currently, Tesla is planning to use Intel 14A for automotive and AI chips production. Google TPU has also adopted EMIB advanced packaging; Amazon, NVIDIA, and Apple are evaluating shifting some chips to Intel foundry services.
(Guo Ming-chi discusses the gap between TSMC CoWoS and Intel EMIB, revealing that Google previously asked about skipping MediaTek and producing chips by itself)
However, analysts point out that the orders from the two companies are currently mostly at the testing, verification, and evaluation stage, and are still unlikely to shake TSMC’s dominant position in scale. In the AI era, requirements for yield and supply stability are far higher than before. Any mistake could lead to large-scale customer loss, meaning that in the short term, orders for core AI chips will still rely heavily on TSMC.
TSMC takes the offensive to defend: 18 wafer fabs built in parallel worldwide
Facing the challenge, TSMC is responding head-on with an unprecedented scale of expansion.
TSMC Chairman Wei Zhejia broke the past practice of “not expanding after hitting mature node targets,” announcing continued expansion of 3-nanometer and advanced packaging capacity, and unveiling future process blueprints such as the 2-nanometer family, A14, and A13. He emphasized that TSMC’s global expansion pace over the past two years has reached 2 times that of the previous period. Currently, a total of 18 12-inch wafer fabs are simultaneously undergoing new construction and renovations, including 5 advanced packaging plants.
Overseas, the US state of Arizona has become TSMC’s most important base for advanced process manufacturing abroad. In the future, new plant areas are planned to build 11 wafer fabs and the first advanced packaging facility. In Japan, Kumamoto Plants 1 and 2 continue mass production while introducing advanced processes. In Germany, the new plant in Dresden will focus on the automotive and industrial control markets.
Taiwan’s biggest TSMC expansion boom in history: 12 plants advancing in parallel
The scale of expansion in Taiwan is also massive. A total of 12 wafer fabs and advanced packaging facilities are under simultaneous construction, located across Hsinchu Science Park, Zhunan, Central Taiwan Science Park, Chiayi, Tainan Science Park, and Kaohsiung Nanzih, creating a large-scale expansion scene rarely seen in recent years.
For Hsinchu Science Park, the F20 plant’s Phase 4 projects are the most important base for generations below 2 nanometers. P1 and P2 plants focus on 2 nanometers; P3 also plans 2 nanometers and A14 in parallel, with future space reserved for development of A10 generation processes. Zhunan’s AP6 advanced packaging plant’s A and B factories can reach nearly 10k wafers per month by the third quarter, continuing to expand CoWoS, SoIC, and InFO capacity.
Chiayi’s AP7 is the largest advanced packaging zone. In Phase 1, Plant P1 has been completed and supports Apple’s WMCM; Plant P2 is nearing completion, focusing on SoIC with monthly capacity of about 12k wafers, and its main customer is NVIDIA. Phases 2 includes Plants P3 through P7 under planning; in the future, they will also support SoIC and CoPoS. Tainan Science Park’s AP8 plant is rebuilt from Innolux’s former plant and is an important base for CoWoS. By end-2026, monthly capacity will exceed 40k wafers. F18 P9 is responsible for the back-end processes after 2 nanometers and the 3 to 5 nanometer TSV processes, with equipment planned to enter production in the first quarter of 2027 and monthly capacity reaching 25k wafers.
Kaohsiung Nanzih’s F22 plant is currently TSMC’s most important 2-nanometer stronghold, with plans covering six phases. P1 and P2 each target monthly capacity of 25k wafers; P3 has already begun equipment installation, focusing on 2 nanometers and A16; Plant P4 targets N2 and A16, and construction has begun recently.
Moving toward below 1 nanometer: TSMC’s next-generation advance deployment
TSMC’s outlook has also shifted from focusing on 2 nanometers to 1 nanometer. The third-phase expansion plan in Longtan is expected to secure land in the fourth quarter of 2029. The fourth phase of Tainan Science Park and the Shalun Park area have already initiated environmental impact assessment preparations. The Pingtung Haifeng Industrial Park plans to develop semiconductor back-end processes and AI-related industries, while the three plans in Zhudong Ke also continue to advance. This shows TSMC is deploying in advance for the below-1-nanometer generation.
(TSMC Longtan plant “revived”? An EME-level process moving into Longke Phase 3; Hanton and Fanxuan will take over orders worth over NT$10k)
Who is the biggest beneficiary of this expansion wave?
This global, largest-scale semiconductor construction wave is not just TSMC’s story—it is also expected to benefit the entire supply chain across the board.
In facilities engineering, major orders visibility for companies such as Hanton (2404), Fanxuan (6196), Yaxiang (6139), Sheng Hui (5536), and Youngkey Engineering (6691) has significantly improved; on the equipment and materials side, companies such as Jadeng (3680), High Sand (1560), XinYun (3583), Hongsu (3131), Zisheng (2467), Inno Energy (7734), and Santa Toshi (3595) are also benefiting in tandem from the massive construction and mass-production demand.
As TSMC simultaneously starts construction in Taiwan, the United States, Japan, and Germany, related facilities and equipment suppliers are expected to receive long-term orders.
Guo Ming-chi warns: TSMC becomes a common hedging target for the semiconductor industry
As Tienfeng International Securities analyst Guo Ming-chi warned earlier, as TSMC’s advanced process resources continue to tilt toward AI and high-performance computing customers, Apple is preparing to cultivate Intel as a second supplier for foundry services across its entire product line. It is expected to ramp up mass production in 2027 and expand shipments in 2028.
In response, Guo Ming-chi advises TSMC to accelerate internal capital accumulation and incorporate uncertainties such as geopolitical risk and changes in customer structure into advanced-process pricing, adopting a more flexible risk-based pricing strategy rather than only maintaining the current pricing model.
(Will Apple support Intel as a backup? Guo Ming-chi reveals TSMC’s crisis and Intel’s 18A-P turnaround opportunity)
TSMC’s position is solid in the short term, but competition is entering a critical stage
Industry insiders generally believe that Samsung and Intel’s aggressive moves are creating unprecedented competitive pressure for TSMC. However, in the short term, the global foundry landscape will still maintain TSMC’s dominance, while Samsung and Intel will only be able to capture a small share of non-core chip orders.
TSMC responds with the largest global expansion wave in history, continuing to widen the gap with rivals through technological leadership, capacity scale, and deep supply-chain integration. However, as Guo Ming-chi has warned, when all the most important global customers begin searching for alternatives, what TSMC faces is not only process technology competition, but a long-term tug-of-war over pricing and capacity strategy.
This article, “Samsung and Intel squeeze in on TSMC—18 plants push the largest expansion plan in history! Facility and materials stocks to benefit,” was first published on Lianxin News ABMedia.
