According to reports, TSMC is accelerating the advancement of its sub-2nm process technology, with wafer foundry prices reportedly reaching as high as $30,000 per wafer. Market sources indicate that following the start of volume production in the second half of 2025, TSMC's 2nm process monthly capacity is expected to reach 40,000 wafers, with a significant expansion to 200,000 wafers per month projected by 2028, setting a new industry benchmark.
Industry insiders highlight that, in addition to long-term clients in smartphone and high-performance computing (HPC) sectors such as Apple, Nvidia, and Qualcomm, there is a growing number of companies developing AI-related chips, including application-specific integrated circuits (ASICs), investing heavily in this advanced process node.
Notably, AI leaders including OpenAI, Elon Musk's xAI and Tesla, along with numerous AI startups, began research collaborations with TSMC as early as 2023 to secure production capacity.
Sources emphasize that the 2nm capacity scale will surpass that of the 3nm node. Volume production for the 2nm process is anticipated to commence in the second half of this year. The Baoshan F20 fab in the Hsinchu Science Park has already increased its monthly capacity to 30,000 wafers, while the Kaohsiung F22 fab contributes 6,000 wafers monthly.
By December 2025, combined monthly capacity across the Hsinchu and Kaohsiung facilities is expected to reach 40,000 wafers, increasing to 53,000 wafers by January 2026, 85,000 wafers mid-year, and ultimately 100,000 wafers by the end of 2026. Monthly capacity is projected to surge to 200,000 wafers by 2028.
Currently, Nvidia's Blackwell architecture remains at the 4nm node, with a transition to 3nm expected by mid-2026. The move to 2nm with the Feynman architecture is anticipated around 2028.
Despite broader market challenges, TSMC is expanding its 2nm production capacity, a move traditionally indicative of secured large-scale orders and careful capacity planning.