TSMC Pauses New 3nm Projects, Accelerates Shift to 2nm

TSMC’s 3nm process continues to face tight capacity, prompting the company to actively encourage customers to consider its next-generation 2nm technology. According to chip industry sources, TSMC has already raised pricing for 3nm wafers this year and has temporarily paused the kick-off of new 3nm projects. The reason is straightforward: existing 3nm capacity is fully booked by AI GPUs, cloud data-center ASICs, and high-end mobile processors, leaving little room for additional demand in the near term.

Industry observers note that this move reflects a broader strategic shift. Rather than expanding 3nm at a limited pace, TSMC is guiding customers still in the early stages of product planning to evaluate 2nm directly, helping optimize long-term production schedules and cost structures. The 2nm process has already entered volume manufacturing, with early adopters including major smartphone chip designers such as Apple, Qualcomm, and MediaTek. While 2nm wafers carry a higher unit price than 3nm, adjustments in die size and large-scale shipments have helped absorb the cost, meaning 2nm is not expected to be a major driver of rising smartphone bill-of-materials costs.

From a technology perspective, 2nm represents a key turning point in TSMC’s advanced node roadmap. It is the company’s first process to adopt nanosheet transistor architecture, delivering clear gains in performance, power efficiency, and transistor density over 3nm. At the same time, optimizations in critical process steps—such as atomic layer deposition (ALD)—help prevent a sharp increase in EUV lithography layers, keeping overall manufacturing costs more balanced. ALD plays a particularly important role at 2nm, as it must form uniform, defect-free high-k dielectrics and metal gates around suspended nanosheet structures, placing extremely high demands on deposition consistency.

In contrast, 3nm remains constrained by capacity. Although TSMC offers multiple variants within its 3nm family, all available output is currently reserved by AI, data-center, and premium mobile customers. Semiconductor analysts believe the temporary slowdown in launching new 3nm projects is not only due to capacity limits, but also a deliberate effort to free up resources and accelerate customer migration toward the 2nm node.