TSMC accelerates 3nm production as demand for AI and HPC chips surges

Taiwanese chipmaker TSMC ramps up 3-nanometre output amid soaring demand for artificial intelligence and high-performance computing hardware, signaling a strategic shift towards more advanced chip manufacturing nodes.

TSMC is accelerating output of its 3-nanometre manufacturing line as demand for advanced chips for artificial intelligence and high-performance computing continues to outpace supply, according to a report by Economic Daily News cited by Aju News. The Taiwanese foundry is now expected to generate more than T$400bn from 3nm products in the third quarter, a record level that would lift the node’s share of total revenue above 30%.

The business case for the shift is clear. TSMC chairman and chief executive C.C. Wei has said the company is expanding 3nm capacity more aggressively than originally planned, but that new wafer fabs take two to three years to build, leaving supply tight in the meantime. To bridge the gap, the company is also converting some existing 5nm lines to 3nm production, a move that is expected to push 3nm revenue ahead of 5nm for the first time in the quarter.

That transition matters because leading-edge nodes carry better pricing and should support gross margin. As TSMC moves more of its mix to 3nm, the company is likely to benefit from the stronger economics of its most advanced process technologies, even as it continues to face bottlenecks caused by surging demand from AI accelerators and HPC systems.

TSMC’s next steps are already visible. The company’s 2nm process, scheduled for mass production in the fourth quarter of 2025, is reportedly progressing faster than expected, while work on angstrom-class manufacturing below 2nm is also moving ahead. The A16, or 1.6nm-class, process is expected to be production-ready later this year and is likely to supply major US technology companies including Apple and Nvidia.