TSMC commits another $100 billion to Arizona for at least four more 2nm fabs — 2026 capex could hit $64 billion following another record quarterly earnings
The extra spending takes TSMC's total U.S. commitment to $265 billion.
TSMC will invest an additional $100 billion in the U.S. to build at least four more chipmaking plants and advanced packaging facilities in Arizona, CEO C.C. Wei announced at the company's second-quarter earnings conference in Taipei on Thursday. The new fabs will produce chips at the 2nm node and below, and the commitment lifts TSMC's total announced U.S. investment to $265 billion, confirming plans that appeared as a market rumor in February. TSMC didn’t attach a timeline, saying the pace of construction will be set by market demand.
Wei said the money will fund several more logic wafer fabs "for 2-nanometer and below technologies" alongside advanced packaging plants to serve multi-year demand from TSMC's leading U.S. customers. According to a Bloomberg report citing a U.S. official, the expansion takes TSMC's planned U.S. footprint to 10 fabs and two advanced packaging facilities.
A leading-edge 2nm-class fab module with a capacity of roughly 20,000 wafer starts per month costs roughly $25 billion to $35 billion to build and equip, so $100 billion funds approximately four modules, matching the "at least four" plants Wei referenced. It also puts the Arizona site around the halfway mark of the 12-fab, four-packaging-facility endgame reported in April as part of the broader U.S.-Taiwan agreement.
The packaging component is arguably the more consequential half of the plan, as CoWoS capacity, not wafer output, remains the binding constraint on AI accelerator production, and on-site packaging in Arizona would give TSMC's U.S. customers a complete domestic supply chain from wafer start to packaged accelerator for the first time.
The announcement comes alongside another record quarter for TSMC, which posted net income of NT$706.56 billion ($22.35 billion) for April-June, up 77.4% year over year and its fifth consecutive quarterly record. Revenue rose 36% to NT$1.27 trillion, and gross margin hit a record 67.7%. High-performance computing accounted for 66% of revenue by platform.
TSMC now expects third-quarter revenue of $44.6 billion to $45.8 billion, up 12% sequentially at the midpoint, and raised its full-year revenue growth forecast to slightly above 40% in U.S. dollar terms, from the roughly 30% it guided in June. CFO Wendell Huang said 2026 capital expenditure will rise to between $60 billion and $64 billion, up from a prior budget of $52 billion to $56 billion, with 70% to 80% of the total going to advanced process technologies. TSMC's suppliers describe the same demand picture, with ASML raising its 2026 outlook on Wednesday, and Applied Materials CEO Gary Dickerson telling Nikkei Asia the industry is set for years of capacity expansion.
TSMC declined to say when the additional $100 billion will be spent, but the commitment follows the U.S.-Taiwan trade agreement that cut tariffs on Taiwanese goods to 15% in exchange for $250 billion in planned Taiwanese investment in the U.S., roughly two weeks after President Donald Trump said TSMC was doubling the size of its Arizona operations. A demand-contingent pledge satisfies that arrangement without committing capital to a schedule. Wei himself bracketed the demand outlook, saying he expects demand to remain very strong through 2029 or 2030, but added: "Whether there is a dip in between or not, I'm not very sure."
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Wei also said TSMC is planning 13 leading-edge and advanced packaging fabs in Taiwan over the next several years, heading off concerns that the U.S. buildout comes at Taiwan's expense. Execution in Arizona still faces labor, water, and visa constraints, which will do as much as demand to determine how quickly four more fabs are built in Phoenix.
Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Luke James is a freelance writer and journalist. Although his background is in legal, he has a personal interest in all things tech, especially hardware and microelectronics, and anything regulatory.
-
Zaranthos The amount of global advanced fab production in wafers per month is growing at a mind boggling pace. The world is likely to double production in just a few years time. Even more shocking is that isn't enough to satisfy global demand. The amount of advanced world compute power available to the human race is growing like never before.Reply
How will it all be used? I sure hope it's for the good of mankind. The amount of problems that could potentially be solved is hard to quantify. -
usertests Reply
I hope we're in a slump and that we get another burst of computational power from a 3D transition. 3D DRAM will almost certainly happen, which will be good, but monolithic 3D SoCs will be a tougher nut to crack.Zaranthos said:The amount of global advanced fab production in wafers per month is growing at a mind boggling pace. The world is likely to double production in just a few years time. Even more shocking is that isn't enough to satisfy global demand. The amount of advanced world compute power available to the human race is growing like never before.
How will it all be used? I sure hope it's for the good of mankind. The amount of problems that could potentially be solved is hard to quantify. -
hotaru251 ahh yes the place that will be unlivable in few decades even before they arrived due to drought...lets build more water hungry factories...Reply
edit: yes I know most of it can be reused but the place already struggles for water to begin with so taking the amount needed even once or twice is a big issue -
SkyBill40 Replyhotaru251 said:ahh yes the place that will be unlivable in few decades even before they arrived due to drought...lets build more water hungry factories...
edit: yes I know most of it can be reused but the place already struggles for water to begin with so taking the amount needed even once or twice is a big issue
In a few decades? Pfft. It's approaching that already. The fact they're spending money and expanding is great, but at what greater cost? -
TechieTwo Replyhotaru251 said:ahh yes the place that will be unlivable in few decades even before they arrived due to drought...lets build more water hungry factories...
edit: yes I know most of it can be reused but the place already struggles for water to begin with so taking the amount needed even once or twice is a big issue
Maybe if there were not perpetual forest fires in CA and Canada there would be water for other uses? :mad: -
Zaranthos Reply
The earth doesn't lack the water. I think we should pressure AI big spenders to supply their own water along with a surplus for the communities where they build.hotaru251 said:ahh yes the place that will be unlivable in few decades even before they arrived due to drought...lets build more water hungry factories...
edit: yes I know most of it can be reused but the place already struggles for water to begin with so taking the amount needed even once or twice is a big issue
https://10across.org/experts-weigh-in-on-a-proposed-arizona-mexico-desalination-plant-to-solve-arizonas-water-problems/ -
Zaranthos Replyusertests said:I hope we're in a slump and that we get another burst of computational power from a 3D transition. 3D DRAM will almost certainly happen, which will be good, but monolithic 3D SoCs will be a tougher nut to crack.
I think technological advancement will explode right along with the AI boom. -
CrazyCarrot911 If Ai won't generate solid profit soon, then who is going to buy and pay for all those chips ?Reply
I still think this will end in a big implosion for one or the other reasons mentioned before.
Water is not negotiable. Some will find that out very soon. -
SkyBill40 ReplyZaranthos said:The earth doesn't lack the water. I think we should pressure AI big spenders to supply their own water along with a surplus for the communities where they build.
https://10across.org/experts-weigh-in-on-a-proposed-arizona-mexico-desalination-plant-to-solve-arizonas-water-problems/
This is, quite honestly, a dead in the water (pardon the pun) solution. The cost alone makes this relatively untenable, not to mention the infrastructure required to get whatever water back to the valley for use. -
usertests Reply
If we go by IMEC's latest roadmap, we get CFETs and incremental improvements on it through 2041. The end results of this 15 year period will be great, but not necessarily revolutionary. Think 10x performance per watt instead of 100-1000x.Zaranthos said:I think technological advancement will explode right along with the AI boom.
https://spectrum.ieee.org/semiconductor-technology-roadmap
Unless they're wrong, of course.