Overview
- Industry reports Monday said TSMC’s most advanced capacity, including 2nm and 3nm, is effectively sold out through 2028, with even its future Arizona plant largely preallocated.
- TSMC is channeling scarce 3nm output to a small set of top clients such as Apple and Nvidia, which is forcing other chip designers to look for alternative production slots.
- Samsung, the only other foundry capable of 2nm production, has reportedly lifted some 2nm yields to about 60% for specific chips, narrowing but not closing the gap with TSMC’s roughly 60%–70%.
- To catch spillover demand, Samsung plans to start its Taylor, Texas fab in the second half of this year and is building a flexible P5 complex in Pyeongtaek that can switch between logic chips and next‑generation memory.
- The race is already shifting to sub‑2nm nodes, with TSMC guiding to a 1.4nm launch in 2028 and Samsung outlining 1.4nm around 2029 and 1nm by 2030 using forksheet transistors that pack more devices into the same area.