TSMC to build 30% of its 2nm and more advanced chips in the U.S., to speed up Fab 21 build out

[fansink]

Production at Fab 21 module 2 is set to start at least two quarters earlier.

TSMC plans to produce 30% of its N2 (2nm-class) output in the U.S. and make its Fab 21 site near Phoenix, Arizona, an independent semiconductor manufacturing cluster, the company's management revealed during the company's earnings call on Thursday. The world's largest contract maker of chips also indicated intentions to speed up building new Fab 21 modules to produce chips on N3 (3nm-class), N2, and A16 (1.6nm-class) nodes.

"After completion, around 30% of our 2nm and more advanced capacity will be located in Arizona, creating an independent leading edge semiconductor manufacturing cluster in the U.S.," said C.C. Wei, chief executive and chairman of TSMC, in his prepared remarks. "It will also create greater economies of scale and help foster a more complete semiconductor supply chain ecosystem in the U.S."

To produce 30% of its N2 and A16 output in Arizona, TSMC will build two additional Fab 21 modules. So far, the company has confirmed plans to build at least three N2 and A16-capable fab modules in Taiwan's Hsinchu and Kaohsiung Science Parks with more modules incoming, so the island will still produce the lion's share of TSMC's advanced chips. Yet, 30% of TSMC's N2 and A16 chips produced in the U.S. is certainly a big deal.

TSMC's Fab 21 module 1 in Arizona is currently ramping up volume production of chips for its American customers using the company's N4 and N5 process technologies. Construction of the company's N3-capable Fab 21 module 2 (its second fabrication facility in Arizona) is complete, and the company is working to start installing equipment there ahead of time in a bid to pull in volume production of chips at the facility by at least a couple of quarters from its original vague schedule of 2028.

Construction of Fab 21 module 3 and module 4 — the company's facilities that will use N2 and A16 nodes — is expected to start later this year, assuming all required permits are obtained. TSMC did not reveal schedules for these Fab 21 modules, but it is reasonable to expect at least one of them to come online by early 2029 provided that TSMC acquires all the necessary equipment on time.

TSMC's Fab 21 module 5 and module 6 will use process technologies beyond A16 (think A14 and possibly more advanced), but their construction timelines and production ramp will depend on future customer demand.

TSMS's grand plan for Fab 21 is to evolve it into a GigaFab cluster with production capacity of at least 100,000 wafer starts per month, though when exactly this happens is something that remains to be seen.

"Our expansion plan will enable TSMC to scale up to a GigaFab cluster to support the needs of our leading-edge customers in smartphone, AI and HPC applications," added Wei.

TSMC to build 30% of its 2nm and more advanced chips in the U.S., to speed up Fab 21 build out

Production at Fab 21 module 2 is set to start at least two quarters earlier.

www.tomshardware.com

 

[Brady]

Construction of Fab 21 module 3 and module 4 — the company's facilities that will use N2 and A16 nodes — is expected to start later this year, assuming all required permits are obtained. TSMC did not reveal schedules for these Fab 21 modules, but it is reasonable to expect at least one of them to come online by early 2029 provided that TSMC acquires all the necessary equipment on time.

So for N2 that's at least N-3 years off the leading edge, perhaps less for A16 but still not leading edge.

"Our expansion plan will enable TSMC to scale up to a GigaFab cluster to support the needs of our leading-edge customers in smartphone, AI and HPC applications," added Wei.

Interesting that they keep saying "leading edge" in these comments when, ostensibly, it is not? Playing semantics? Or does the industry actually call N-2/N-3 years behind Taiwan "leading edge"?

 

[fansink]

So for N2 that's at least N-3 years off the leading edge, perhaps less for A16 but still not leading edge.


Interesting that they keep saying "leading edge" in these comments when, ostensibly, it is not? Playing semantics? Or does the industry actually call N-2/N-3 years behind Taiwan "leading edge"?

C.C. Wei is not misleading, nor is HE being obtuse. He clearly stated, “an independent leading edge semiconductor manufacturing cluster in the U.S.,” explicitly scoping the term to the United States, not the global industry.

Grammatically, “leading edge” is an adjective modifying the noun phrase “semiconductor manufacturing cluster.” The prepositional phrase “in the U.S.” directly follows and modifies “cluster,” specifying its geographical scope. In English, a prepositional phrase like “in the U.S.” placed immediately after a noun phrase limits the scope of that phrase. Here, it indicates that the “leading edge semiconductor manufacturing cluster” pertains to the U.S. context.

Given that the U.S. currently has limited domestic production of advanced nodes like 2nm, Wei’s statement reflects TSMC’s goal to establish cutting-edge capabilities in Arizona.

Is C.C. Wei forecasting that by 2028, roughly 2.5 to 3 years from now, 2nm will be considered “leading edge” in the U.S.? Absolutely.

This aligns with his strategic intent to strengthen the U.S. semiconductor ecosystem, not to claim global supremacy.

 

[jabo]

It is very clear to me TSMC has in effect tricked the current administration. Hats off to them...they've played on Trump's ego perfectly. They knew announcing a $100+ B deal to build "leading" edge fab in the USA would allow Trump to brag and boast. He brags about how just with a few comments and tariff threats, he got what he wanted. The only leading edge fab in the USA for the foreseeable future will be at Intel Foundry (18A, 18A-P, 14A, etc...). This will rival whatever TSMC is doing in Taiwan at even given time.

 

[fansink]

It is very clear to me TSMC has in effect tricked the current administration. Hats off to them...they've played on Trump's ego perfectly. They knew announcing a $100+ B deal to build "leading" edge fab in the USA would allow Trump to brag and boast. He brags about how just with a few comments and tariff threats, he got what he wanted. The only leading edge fab in the USA for the foreseeable future will be at Intel Foundry (18A, 18A-P, 14A, etc...). This will rival whatever TSMC is doing in Taiwan at even given time.

TSMC’s 2nm will absolutely be “leading edge” in the U.S., even if global nodes reach ~1.6nm. Intel’s 18A, targeted for early 2026, is promising but unproven in high-volume production and customer adoption, unlike TSMC’s mature and dominant ecosystem.

 

[Brady]

C.C. Wei is not misleading, nor is HE being obtuse. He clearly stated, “an independent leading edge semiconductor manufacturing cluster in the U.S.,” explicitly scoping the term to the United States, not the global industry.

Grammatically, “leading edge” is an adjective modifying the noun phrase “semiconductor manufacturing cluster.” The prepositional phrase “in the U.S.” directly follows and modifies “cluster,” specifying its geographical scope. In English, a prepositional phrase like “in the U.S.” placed immediately after a noun phrase limits the scope of that phrase. Here, it indicates that the “leading edge semiconductor manufacturing cluster” pertains to the U.S. context.

Not being sarcastic, but did ChatGPT write this?

I don’t think it’s clear to A Human Being, regardless of this paragraph trying to explain it in English grammatical and syntactical specifics, that he was speaking only “to the U.S. context”.

Anyways, we’re discussing semantics and regardless it’s great to see more investment in the US capabilities here and that TSMC Arizona looks to not be a flash in the pan but a large strategic investment. Perhaps that was never in doubt, but the proof is only in the nodes actually shipping wafers and that looks to be moving towards the leading edge with both Intel and TSMC.

Hope the US (and close strategic allies) can start to pick up more of the electronics supply chain because leading edge wafers without other components are still useless…

 

[fansink]

Not being sarcastic, but did ChatGPT write this?

I don’t think it’s clear to A Human Being, regardless of this paragraph trying to explain it in English grammatical and syntactical specifics, that he was speaking only “to the U.S. context”.

Anyways, we’re discussing semantics and regardless it’s great to see more investment in the US capabilities here and that TSMC Arizona looks to not be a flash in the pan but a large strategic investment. Perhaps that was never in doubt, but the proof is only in the nodes actually shipping wafers and that looks to be moving towards the leading edge with both Intel and TSMC.

Hope the US (and close strategic allies) can start to pick up more of the electronics supply chain because leading edge wafers without other components are still useless…

I'm glad you agree.

TSMC’s Fab 21 Phase 2 in Phoenix, producing 3nm chips, may be as early as 3Q26.