TSMC hurrying to bring advanced chip tech to Arizona fab
Story by Tobias Mann • 10h
CEO C.C. Wei cites strong demand for AI products. Intel may also be a factor
TSMC is accelerating the rollout of advanced process nodes at its Arizona fabs to meet growing demand for American-made AI products.…
The Taiwanese foundry giant currently produces chips based on its older N4 (4-nanometer) process technology at its Fab 21 site in Arizona, which began volume production earlier this year, making chips for Apple and Nvidia.
TSMC's next fab, the second of six planned for its Arizona manufacturing campus, will bring its N3 node to the US in 2028. Across the Pacific, TSMC is preparing to ramp production of its N2 node, but it wasn't expected to find its way into the foundry operator's US fabs until the end of the decade.
Now, its plans have apparently changed. During TSMC's Q3 earnings call this week, CEO C.C. Wei revealed the company would accelerate this timeline.
"We are preparing to upgrade our technologies faster ... to more advanced process technologies in Arizona, given the strong AI-related demand from our customers," he said.
Whether that means TSMC's second fab will skip N3 and jump straight to N2 or its yet unreleased A16 process, the exec didn't say. El Reg reached out to TSMC for clarification and we'll let you know if we hear anything back.
Since first announcing its US fab expansion in 2020, TSMC has reserved its most sophisticated process technology for its Taiwanese plants. However, since announcing a $100 billion expansion of its fab footprint in an appearance alongside US President Trump early this year, the company has seen a wave of interest from American chipmakers eager to build their parts domestically.
Apple, Nvidia, and AMD are all producing or plan to produce parts at TSMC's first Arizona chip plant. This surge in demand has no doubt been influenced by the persistent threat of tariffs on foreign-made electronics and semiconductors.
Another possible factor is the imminent availability of Intel's 2-nanometer class 18A process technology, which will power the x86 giant's Panther Lake and Clearwater Forest Xeons. The success of these products could herald a shift in demand toward the rival foundry.
Intel's Foundry division has been searching for a hero customer for some time, initially targeting 18A as its breakout moment. The company has since refocused on its upcoming 14A node as its first mainstream foundry offering.
While unproven, 18A features several elements not yet available on TSMC nodes, including the use of backside power delivery, which promises to boost transistor densities by a significant margin. Perhaps more important, Intel offers a 2nm class process that's made right here in the US-of-A today. ® |
