Node yields: Samsung vs TSMC
A recent social post compared 2nm yields, showing Samsung at roughly 55% while TSMC is reported in a 60–70% range, highlighting yield gaps at the most advanced nodes. (x.com)
A chip factory’s “yield” is the share of good chips on a silicon wafer, and at 2 nanometers that number still separates Samsung from Taiwan Semiconductor Manufacturing Co. Reports cited by TrendForce put Samsung near 55% and Taiwan Semiconductor Manufacturing Co. in a 60% to 70% range during trial production. (trendforce.com) At this stage, a gap of 5 to 15 percentage points can decide who wins the first big orders, because each wafer at the leading edge costs tens of thousands of dollars and bad dies cannot be sold. TrendForce, citing Taiwan reports, said Taiwan Semiconductor Manufacturing Co.’s 2 nanometer trial yield had already cleared 60% to 70% by late 2024. (trendforce.com) The two companies are not at exactly the same point in rollout. Taiwan Semiconductor Manufacturing Co. says its N2 process started volume production in the fourth quarter of 2025, while Samsung has said its SF2 process would begin with mobile chips in 2025, then expand to high-performance computing in 2026. (tsmc.com) (semiconductor.samsung.com) That timing matters because 2 nanometer is the next manufacturing step after 3 nanometer for the phones, artificial intelligence accelerators, and server chips due in 2026 and 2027. The foundry with the steadier line usually gets the earliest, highest-volume products from customers such as Apple, Qualcomm, Nvidia, and Advanced Micro Devices. (tsmc.com) (trendforce.com) Both companies are also changing transistor design at this node. Taiwan Semiconductor Manufacturing Co. says N2 uses nanosheet transistors, a gate-all-around structure that wraps the switch more completely to cut leakage and improve control, and Samsung has been pushing the same gate-all-around approach across its advanced roadmap. (tsmc.com) (semiconductor.samsung.com) Samsung has framed SF2 as a step up from its 3 nanometer process, with a 12% performance gain, 25% better power efficiency, and 5% area reduction. Taiwan Semiconductor Manufacturing Co. has made a similar pitch on N2, describing it as a full-node advance aimed at better performance and energy efficiency. (semiconductor.samsung.com) (tsmc.com) Yield figures themselves need caution because companies rarely publish a single official number, and reports can refer to different test vehicles such as memory blocks or logic chips. Taiwan Semiconductor Manufacturing Co. researchers have publicly said N2 passed qualification with more than 90% yield on a 256 megabit static random-access memory block and a logic test chip with more than 3 billion gates, which is not the same thing as full commercial product yield. (research.tsmc.com) Outside reports suggest Taiwan Semiconductor Manufacturing Co.’s line kept improving after early trial runs. A June 2025 report in Commercial Times said initial 2 nanometer yield started near 60% and later pushed past 90% on 256 megabit static random-access memory, while Samsung was still reported below that level on its own 2 nanometer process. (ctee.com.tw) Samsung’s case is that process performance is improving and its roadmap is intact. At Samsung Foundry Forum events, the company has kept its 2025 start for mobile SF2, 2026 expansion for high-performance computing, and 2027 move into automotive and 1.4 nanometer. (semiconductor.samsung.com) (news.samsung.com) The short version is that 2 nanometer is no longer a lab milestone but a production contest, and the yield gap is the scoreboard customers watch first. Taiwan Semiconductor Manufacturing Co. has already moved N2 into volume production, while Samsung is still trying to prove SF2 can deliver enough good chips, cheaply enough, to win the same class of orders. (tsmc.com) (trendforce.com)