Record earnings. Record margins. A massive capital commitment to the United States. And then the stock dropped 5%.
That is where TSMC stands today, and it may be the most telling moment in the entire semiconductor cycle right now.
On July 16, TSMC reported Q2 2026 revenue of $40.2 billion, up 33.7% year-over-year. Net income came in at roughly $22 billion, a 77.4% jump from the same period a year ago, and a record for the fifth consecutive quarter. Gross margin hit 67.7%. Operating margin reached 60.3%. Both were all-time highs. The company then raised its full-year 2026 revenue growth guidance from over 30% to slightly more than 40% in U.S. dollar terms.
By any reasonable standard, that is a flawless quarter.
Then It Announced the $265 Billion Commitment
Alongside the earnings report, TSMC CEO C.C. Wei announced an additional $100 billion investment in Arizona semiconductor manufacturing facilities, bringing TSMC’s total committed U.S. investment to $265 billion. The expansion covers four additional advanced chip fabrication plants for 2nm and below technologies, plus advanced packaging. Government statements around the announcement described an eventual footprint of 12 leading-edge manufacturing and packaging facilities in Arizona.
The company also raised its 2026 capital expenditure guidance to between $60 billion and $64 billion, up from a prior range of $52 billion to $56 billion.
Wei said AI demand from cloud customers remains “extremely robust” and that he expects strong demand “all the way to probably 2029, 2030.”
The market reacted by selling the stock anyway.
Why the Sell-Off Actually Makes Sense
This is where it gets interesting. The reaction is not irrational — it is actually quite logical given where we are in the cycle.
The capex raise is the problem. A jump from $52–56 billion to $60–64 billion means more spending, which compresses near-term free cash flow. The $265 billion Arizona commitment, while strategically significant, is a multi-year spend that adds uncertainty to the balance sheet. Markets tend to punish near-term cash flow compression even when the long-term logic is sound.
At the same time, the Philadelphia Semiconductor Index has pulled back meaningfully from its late-June high. After a run like that, even a perfect quarter is not enough to hold the price. The valuation premium the sector built up has been evaporating, and TSMC — despite its fundamental quality — is caught in that same tide.
For the first time since late April, semiconductors no longer trade at a premium valuation to the broader Nasdaq. The PEG ratio for the semiconductor sector now sits at 1.26x — the lowest reading since 2016 — compared to 1.56x for the Nasdaq 100. Semiconductors are now expected to grow faster while costing less per unit of that growth.
What the Numbers Actually Mean
High-performance computing, which includes AI chips, generated 66% of TSMC’s second-quarter revenue. The 2nm node contributed to wafer revenue for the first time. Q3 2026 revenue guidance came in at $44.6 billion to $45.8 billion, another step up from the record Q2.
Wall Street had estimated NT$632.64 billion in net income for Q2. TSMC delivered NT$706.56 billion. That is not a small beat. Analysts who expected 30% full-year revenue growth now have to model 40%+ instead.
The AI demand thesis is not cracking. It is accelerating. TSMC’s own customers — every major hyperscaler and chip designer on the planet — are providing what Wei called “very strong signals and positive outlook.” The company does not speculate on demand. It reports what Nvidia, AMD, Apple, Google, and Amazon are actually ordering.
The Real Question Right Now
The debate is not whether TSMC’s business is strong. It clearly is. The debate is whether the market can look past near-term free cash flow pressure and re-rate the stock on the strength of what a $265 billion domestic chip industry actually means for the next five years.
Alphabet reports next week. Amazon reports July 30. Microsoft reports July 29. Those three earnings calls will either confirm or challenge the AI capex cycle in real time. If hyperscalers hold spending guidance, every argument for the semiconductor sell-off being a valuation reset rather than a cycle turn gets stronger.
TSMC just showed the most convincing evidence yet that the AI infrastructure boom is intact. The market sold it anyway. That gap between evidence and price action is worth sitting with for a moment.
Sector-wide selloffs have a way of creating opportunities in the highest-quality businesses within that sector. Whether TSMC at current levels is one of those is the question every semiconductor investor is running this weekend.