AI丨TSMC's 2nm Process in High Demand, Order Backlog Extends Beyond 2028

Benefiting from strong demand for artificial intelligence (AI), TSMC’s 2-nanometer process capacity is severely in short supply. According to Taiwanese media reports on Monday, Nvidia and Meta have joined the queue for capacity, with TSMC’s 2-nanometer orders already scheduled until after 2028. TSMC’s advanced processes will also see price increases for four consecutive years.

Industry sources indicate that the supply shortage of TSMC’s 2-nanometer process is more pronounced than before. In addition to existing customers like Nvidia, Broadcom, and MediaTek placing orders for the next two years, Intel is also in line to reserve capacity, with the possibility of launching new products as early as 2027. Furthermore, Meta has also joined the competition for capacity.

Nvidia plans to use TSMC’s A16 process for the next-generation Feynman platform. Due to A16 capacity constraints, Nvidia reportedly cannot secure enough A16 capacity, only using the most critical bare chips (dies) with A16, while some die designs are being switched to N3P process.

Additionally, TSMC has communicated pricing with customers based on their scale and procurement status since September last year. Prices are expected to increase by single-digit percentages in 2026, marking the fourth consecutive year of price hikes. Industry rumors suggest that prices may continue to rise for four years starting in 2026. Research firms previously forecast that advanced process prices could increase by 3% to 10% in 2026.

TSMC declined to comment on individual customer information and emphasized that it does not comment on pricing issues.