TSMC Breaks Records and Adds Another US$100 Billion in the U.S.

Index
TSMC closed the best quarter in its history and, in the same move, committed another US$100 billion to its production in the United States. The world's largest chipmaker earned nearly US$22 billion between April and June, up 77% from a year earlier, and the engine was the same one driving the entire industry: demand for silicon for artificial intelligence. When a company that does not sell a single consumer product earns that kind of money manufacturing everyone else's chips, its numbers stop being a financial footnote and become the most honest gauge of the AI boom.
A Record Quarter, Driven by AI
The results reported by TSMC break its own streak: a fifth consecutive quarter of record profit, with margins that no consumer tech company comes close to matching.
Behind those margins, there is one clear protagonist. The high-performance computing segment—the chips that power AI data centers—accounted for 66% of sales, compared with 22% from smartphones. AI is no longer a promise in TSMC's revenue: it is two out of every three dollars coming in.
Why TSMC Is the Best Gauge of AI
Nvidia, AMD, Apple, and almost everyone designing a leading-edge chip end up on the same production line. TSMC does not compete with its customers; it manufactures for them. That is why its results say more about real AI demand than any announcement: they do not measure expectations or future orders, but wafers that have already been paid for and shipped.
That position makes the company both optimistic and cautious. TSMC raised its 2026 growth outlook to more than 40%—the second upward revision this year—and increased its capital expenditure budget to between US$60 billion and US$64 billion. "Our conviction in the multi-year AI megatrend remains very high," CEO C.C. Wei summarized.
Another US$100 Billion for Arizona
The announcement that accompanied the results is just as forceful. TSMC will add another US$100 billion to its Arizona operations, bringing its total commitment in the United States to US$265 billion, up from the US$165 billion it had pledged in March 2025. With that money, it plans at least four new 2-nanometer fabs and advanced packaging plants, putting the complex on track for roughly ten fabs. It is the largest foreign direct investment in a new industrial project on U.S. soil.
The reason is not only commercial. The other side is tariffs: Washington has been conditioning exemptions from semiconductor duties on manufacturers like TSMC producing inside the country, and the Donald Trump administration made local manufacturing a flagship policy. For TSMC, spreading production beyond Taiwan also reduces the risk of concentrating almost all of the planet's advanced silicon on a single island.
What It Means for the Hardware You Buy
The 2-nanometer process coming to Arizona is not a lab statistic: it is the process that will be used to manufacture the next iPhone, Mac, and high-end GPUs. Producing that node in the United States as well, not only in Taiwan, begins to change where the chips in your devices come from.
There is a bottleneck less visible than manufacturing itself: advanced packaging, the stage that integrates multiple chips into a single module to accelerate AI. Wei acknowledged that this step limits how much supply can grow, and it is one of the reasons AI GPUs remain expensive and scarce. More packaging capacity translates into more available GPUs and less pressure on prices. That is why the investment jump matters even for anyone who will never set foot in a fab.
Conclusion
In the AI race, everyone watches Nvidia, but the company posting record revenue quarter after quarter is the one manufacturing its chips. TSMC has established itself as the quiet winner of the boom, with profitability that no consumer tech company can touch. The uncomfortable downside: the more AI grows, the more the entire world depends on a single company to produce the chips that sustain it. The US$100 billion for Arizona is, before being a business move, an attempt to ensure that this single point of failure is no longer all in the same place.
Frequently Asked Questions
Who makes the chips for Nvidia, Apple, and AMD?
All three design their chips, but they do not manufacture them. Production of their most advanced processors is handled by TSMC, the Taiwanese foundry that dominates leading-edge semiconductor manufacturing worldwide.
How much did TSMC earn in the second quarter of 2026?
TSMC reported net income of NT$706.56 billion, about US$22.35 billion, up 77.4% from the same period in 2025. It was its fifth consecutive record quarter, driven by demand for AI chips.
Why is TSMC investing US$100 billion in Arizona?
To bring its production closer to the United States, take advantage of tariff exemptions, and reduce its dependence on Taiwan. The investment adds at least four 2-nanometer fabs and raises its total commitment in the country to US$265 billion.
Report based on official announcements and verified public sources at the time of publishing.
Comments
Share article
You might also like

United States invests US$2 billion in quantum computing and takes stakes in nine companies
The United States is distributing US$2 billion among nine quantum computing companies and taking an equity stake in each one. IBM gets the largest share to build the country's first quantum foundry.
8 July 2026
Apple vs. ChatGPT, Claude, and Gemini: why Apple isn't playing the same game
Apple does not want to be ChatGPT, Claude, or Gemini. While OpenAI, Anthropic, and Google compete to build the best model, Apple is betting on being the platform where they all run. Why this is a different strategy.
16 June 2026Recent articles

Apple is preparing two new Apple Pencil models for 2027 with replaceable batteries
Apple is preparing two new Apple Pencil models for the first half of 2027 with replaceable batteries, pushed by European law. What is known and what is not.
16 July 2026
IBM Nighthawk: the 120-qubit chip IBM is using to target fault-tolerant quantum computing in 2029
IBM introduced Nighthawk, a 120-qubit quantum chip aimed at quantum advantage and anchoring its plan toward the first fault-tolerant computer in 2029. What it is, how it is made, and how it compares with Google, Microsoft, and Amazon.
7 July 2026


