Foundry chipmaker TSMC held an expansion ceremony for 2nm at its second module at Fab 22 in Kaohsiung on Monday March 31, aiming for a manufacturing capacity of 30,000 wafer starts per month by year end.
www.eenewsanalog.com, Apr. 01, 2025 –
TSMC is accelerating the introduction of its next manufacturing node and leaving rivals Samsung and Intel behind. TSMC’s production was due to start before the end of 2025 but now looks set to begin in 3Q25, enabled by yields have already reportedly reach commercial levels of 60 to 70 percent.
Samsung has struggled to achieve commercial yields in its 3nm and 2nm process nodes and Intel’s 18A process has yet to ramp. It is essentially an internal process without foundry customers yet. With government backing Japanese startup Rapidus is attempting to bring up a 2nm manufacturing process licensed from IBM, but with no commercial track record could find it hard to attract customers.
As a result, TSMC has numerous customers lined up and eager to use the 2nm process including Apple, Broadcom, Marvell, Mediatek and Intel but as a result 2nm wafers could cost US$30,000 each, according to a Digitimes report. Intel has been using TSMC for some of its production but under incoming CEO Lip-Bu Tan, the plan of record is to transform Intel manufacturing into a foundry business with the 18A node as competitive with TSMC’s 2nm node.
One other major TSMC customer, Nvidia, is in no hurry to adopt 2nm manufacturing. Nvidia will use the 3nm process for its Rubin platform in 2H26, Digitimes asserted.
Qin Yongpei, TSMC’s co-COO, said that the number of tape-outs on 2nm in the first two years of production will exceed the number on 3nm in its first two years of production.
TSMC’s first 2nm mass production fab is Fab 20 in Hsinchu while module 1 at Fab 22 in Kaohsiung is ahead of schedule and ready to start 2nm mass production, Digitimes reported.