Following ChatGPT, DeepSeek’s emergence, the AI boom has been further fueled. Driven by AI, the demand for advanced process chips continues to surge. While 3nm chips have already entered mass production, multiple manufacturers are actively investing in 2nm fabs to meet future market needs, targeting mass production or trial production in 2025. As the target timeline approaches, the construction of 2nm fabs has entered an accelerated phase.
Rapidus Advances Smoothly, Trial Production Set for April 2025
Rapidus plans to begin trial production of its first 2nm fab in 2025, with mass production scheduled for 2027.
Recently, Rapidus President Atsuyoshi Koike stated that the construction of the 2nm fab is progressing smoothly, with more than 200 pieces of equipment to be installed, and trial production set to begin on April 1, 2025. Regarding the economic benefits of Rapidus’ production, Koike estimated that from the start of mass production in 2027 to 2036, the cumulative revenue could reach 18 trillion yen.
Rapidus plans to introduce 10 extreme ultraviolet (EUV) lithography machines in its 2nm fab. According to previous reports, the EUV lithography machines ordered by Rapidus arrived at New Chitose Airport at the end of 2024. The model introduced is ASML’s Twinscan NXE: 3800E, capable of processing up to 220 wafers per hour, marking Japan’s first-ever deployment of EUV lithography technology.
Following the April 2025 trial production, Rapidus plans to deliver 2nm chip samples to Broadcom by June 2025. Once Broadcom verifies the chip performance, Rapidus will be commissioned for production.
TSMC On Track for Mass Production as 2nm Demand Surges
TSMC continues to advance its leading-edge process production. In January 2025, TSMC CEO C.C. Wei announced in an earnings call that N2 (2nm process) is expected to enter production in the second half of 2025, while N2P (enhanced 2nm process) and A16 (1.6nm process) are set for 2026 H2 production.
Regarding fab progress, TSMC’s Kaohsiung 2nm fab, originally planned to introduce equipment in mid-2025 and start production by year-end, has accelerated its timeline by six months due to high demand. On November 26, 2024, TSMC held an equipment move-in ceremony for the facility.
Additionally, TSMC’s Baoshan P1, P2 fabs and Hsinchu F20 fab will also produce 2nm chips. Baoshan P1 received equipment in April 2024 and is expected to enter verification in Q4 2024, with mass production in 2025. The Hsinchu F20 fab has already established a 2nm trial production line, with capacity ramp-up planned for Q4 2025.
According to TSMC’s roadmap, by the end of 2025, including contributions from the Kaohsiung fab, its 2nm process total monthly capacity is expected to exceed 50,000 wafers. By the end of 2026, this figure is projected to rise further to 120,000–130,000 wafers per month.
Key customers for TSMC’s 2nm chips include Apple, NVIDIA, AMD, and Qualcomm.
Samsung Focuses on Yield Improvement, Pushing 2nm Forward
Samsung has begun installing 2nm production equipment at its S3 fab in Hwaseong, South Korea, aiming to start 2nm chip trial production in Q1 2025. By the end of 2025, Samsung plans to convert its remaining 3nm production lines at S3 to 2nm lines to further expand capacity.
Industry sources indicate that transitioning from 3nm to 2nm involves fundamental changes in semiconductor transistor structures at the microscopic level. This requires more complex bonding, higher wafer flatness standards, and upgrades to certain key equipment.
Amid these challenges, reports suggest that Samsung has faced yield rate issues in 2nm production. However, Samsung remains committed to overcoming these difficulties, actively improving yields, and innovating in technology and equipment to meet market demands.
Regarding customers, Japan’s PFN will use Samsung’s 2nm process for AI chip production, and U.S.-based Ambarella has also commissioned Samsung to manufacture 2nm chips. Additionally, Samsung has received 2nm orders from South Korean design companies.
(Photo credit: Rapidus)