TSMC plans 1nm roadmap, 12 fabs

Taiwan Semiconductor Manufacturing Co. is sketching a manufacturing plan that runs well past the current AI boom. TechNode reported on May 19 that TSMC is preparing a 1nm-era roadmap and building up to 12 new wafer fabs for technologies spanning 2nm to 1.4nm. (technode.com) That matters because the company is not talking only about one next node. TSMC’s own technology pages say its 2nm process entered volume production in the fourth quarter of 2025, while its N2P extension is scheduled for volume production in the second half of 2026. (technode.com) Its A14 process, the next major step after the 2nm family, is targeted for volume production in the second half of 2028. (tsmc.com) The result is a useful way to read what the company is doing now: this is less a single fab story than a multi-year capacity build across several generations of advanced logic. TSMC’s first-quarter results showed why it has room to do that. Revenue reached $35.9 billion in the first quarter of 2026, gross margin was 66.2%, and the company said advanced technologies of 7nm and below made up 74% of wafer revenue. (technode.com) ### Why does “up to 12 fabs” matter more than the 1nm headline? TechNode’s report said the planned fabs would serve as production bases for nodes ranging from 2nm to 1.4nm, not just a distant 1nm-class process. (tsmc.com) That distinction matters because advanced-node expansion is usually constrained by more than transistor design. Each new fab adds pressure on lithography tools, advanced packaging lines, power and water infrastructure, and trained labor. TechNode separately reported in April that TSMC was targeting a five-fab 2nm ramp in 2026, which it described as the most aggressive expansion in the company’s history. (investor.tsmc.com) ### How does AI demand show up in the numbers? TSMC’s April earnings release gave the clearest baseline. First-quarter 2026 revenue rose 40.6% year over year in U.S. dollar terms to $35.90 billion, and gross margin reached 66.2%. (technode.com) Investor-focused coverage has tied that performance directly to AI datacenter demand. Insider Monkey said TSMC sits at the center of next-generation datacenter demand because major chip designers still need its advanced manufacturing capacity before AI demand can turn into shipped silicon. (technode.com) ### Where does Apple fit into this? Apple is not named in the TechNode report as a participant in the new fab plan, but TSMC’s annual report lists systems companies alongside fabless designers among its customer base, and Apple has long depended on TSMC for leading-edge application processors. (investor.tsmc.com) For Apple, the practical read-through is that more fab construction does not automatically reduce dependence on one supplier. If TSMC is expanding 2nm, N2P and A14 capacity while also serving AI chip customers, the competition is still happening inside the same manufacturing network. That is an inference from TSMC’s roadmap and capacity plans, rather than a statement the company made directly. (insidermonkey.com) ### Is 1nm actually close? TSMC’s public materials do not say it is starting 1nm production soon. The company’s disclosed roadmap today runs through the 2nm family and A14, with A14 volume production scheduled for the second half of 2028. (investor.tsmc.com) So the near-term milestones to watch are more concrete: continued 2nm ramping in 2026, N2P in the second half of 2026, and A14 in 2028. Those dates, more than the 1nm shorthand, will show how fast TSMC can translate AI demand into new leading-edge capacity. (technode.com) (tsmc.com) (technode.com)