The 3nm Battle Between Taiwan Semiconductor and Samsung: The Winning Formula for Process Breakthroughs

Opening the financial reports of major global foundries for Q2 2025, TSMC stands out: revenue growth year-on-year exceeds 40%, far ahead of the industry average, with market share further expanding to 70%.

Looking back at history, TSMC has made the right choices at several significant technological bifurcation points in advanced logic processes recently, without any obvious mistakes.

The 7nm node, which began risk production around 2018, saw Intel choose DUV lithography machines, facing difficulties in mass production due to yield issues. TSMC opted to switch to EUV lithography machines, which had lower costs and higher yields. Intel took three years to resolve its yield problems but missed the market window. TSMC surpassed Intel in one fell swoop, achieving a leading position in advanced processes.

The 3nm node, which began risk production around 2022, saw Samsung choose to switch to the more advanced GAA transistors, which underperformed expectations and faced mass production difficulties due to yield issues. TSMC chose to continue using the FinFET architecture and leveraged material and process innovations to improve performance and density, entering mass production ahead of others. Although Samsung announced the start of mass production of its GAA process in Q2 2022, it was exposed in Q1 2024 that yields were below 20%, and it is still struggling in the yield ramp-up phase.

With the 3nm process, TSMC is expected to rake in $16.2 billion in 2024 (with its 3nm process accounting for 18% of total annual revenue), and this proportion is expected to rise to 22% in Q1 2025. Meanwhile, Samsung has yet to win the favor of any major customers, with a market share close to zero. In this battle, TSMC's FinFET architecture has achieved a resounding victory, while Samsung's GAA architecture is struggling. It is estimated that both TSMC and Samsung have invested over $10 billion each in R&D for the 3nm process node.

The balance of this multi-billion dollar business opportunity had already tilted in favor of TSMC back in 2018. In May 2018, Samsung announced it would adopt GAA transistors for the 3nm node. At that time, the 7nm process had not yet been mass-produced, and the technology for 3nm was still in the early research phase. Foundries like IBM had just made significant breakthroughs in GAA transistor research in 2017, demonstrating substantial performance improvement potential. Meanwhile, key technologies for further enhancing FinFET performance were still in the academic discussion stage. TSMC, Samsung, and Intel made their technology choices for the 3nm process without data support, making decisions worth hundreds of billions of dollars five years in advance.

01 Why does TSMC always manage to choose the right path at technological crossroads?

TSMC's process R&D department has built a digital twin system that explores a vast array of material and process combinations in a simulation environment, evaluating performance gains and yield risks one by one. Supported by this quantitative assessment system, TSMC is able to "calculate without fighting," avoiding reckless gambles and continuously winning in technological and commercial competition The core of this digital twin is a unique branch of EDA software—TCAD simulation software. TCAD (Technology Computer Aided Design) is a semiconductor process and device simulation software that comprehensively describes the physical aspects of processes and devices. It summarizes the steps in semiconductor manufacturing processes, such as thin films, etching, photolithography, ion implantation, diffusion, oxidation, and chemical planarization, as well as the higher-order physical effects that need to be considered behind the physical characteristics of transistors, such as quantum constraints and ballistic transport, into a system of partial differential equations, which are solved using numerical methods. With TCAD products, wafer fabs can replace expensive and time-consuming experiments with numerical simulations, shortening the process development cycle by more than 30% and reducing wafer costs by over 50% (according to the International Technology Roadmap for Semiconductors (ITRS) data). Wafer fabs also utilize TCAD to simulate and optimize different device structures, simulating circuit performance and electrical defects to improve the performance of devices and circuits. In the development of advanced process nodes such as FinFET and GAA, TCAD contributes over 70% to device structure optimization.

It can be said that the application level of TCAD determines the technological advancement and yield of wafer fabs. It is the core software for device and process development in wafer fabs and an essential tool for formulating chip process specifications. For many years, the global TCAD simulation tools have been mainly monopolized by two American companies: Synopsys and Silvaco. Synopsys, as the global leader in TCAD software, focuses on the most advanced process nodes (such as 5nm, 3nm, 2nm), FinFET, GAA, and other complex three-dimensional device simulations, and is recognized as the gold standard in the industry; Silvaco's TCAD has a significant advantage in power devices (Power Devices) and compound semiconductors (GaN, SiC). After 2011, two emerging players entered this field: the Austrian company Global TCAD Solutions and the Chinese company Suzhou Peifeng Tunan Semiconductor Co., Ltd. The former is known for commercializing cutting-edge device simulation academic achievements from the Vienna University of Technology (TU Wien), while the latter is renowned for being able to fully benchmark Synopsys and having an outstanding virtual wafer fab tool, Emulator.

Source: Semiconductor Industry Overview, Original title: "The Battle Between TSMC and Samsung for 3nm: The Winning Formula for Process Breakthroughs" (Article has been edited) Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment based on this is at one's own risk