Taiwan Semiconductor Manufacturing reported on Thursday that third quarter profit rose about 69% from a year earlier. It was the sixth straight quarter of record earnings. Revenue grew roughly 51%. The company guided the current quarter above what analysts had expected. And the U.S.-listed shares fell about 2% on Thursday as a sector-wide selloff deepened.
If you only read the headlines, that looks like a contradiction. Great numbers, falling stock. But the gap between the two is exactly where individual investors should be spending their time right now, because it says something about how the market is pricing the AI buildout. The business is not the question anymore. The price paid for it is.
What TSMC Actually Said
Start with the results, because they were strong. According to reporting from Yahoo Finance, net income rose nearly 69% and revenue grew 51%. TSMC guided fourth quarter revenue to a range of $52.3 billion to $54.3 billion, per Benzinga, which works out to growth of about 58% at the midpoint. CEO C.C. Wei said demand for the company’s leading-edge technologies continues to be “extremely robust,” adding, “We did not see any sign of slowdown.” He tied that to what he described as rising adoption of agentic AI as the industry moves from chatbots to AI assistants.
Two details stand out. First, gross margin came in at 66.4%, down from 67.7% in the second quarter, with fourth quarter guidance of 65% to 67%. That is still a remarkable margin for a company that physically manufactures things, but it is a small step down, and a market priced for perfection notices small steps. Second, the share of wafer revenue coming from 2-nanometer chips rose from 3% in the second quarter to 8% in the third, according to Benzinga. That is the newest and most expensive technology, and customers are clearly lining up for it.
TSMC also held its 2026 capital budget at $60 billion to $64 billion. CFO Wendell Huang pointed to continued multi-year demand as the reason, and in July the company said capital spending over the next three years would be “even more significantly higher” than over the past three years. That is a company telling you it sees a long runway, and is willing to bet tens of billions of dollars on it.
So Why Did the Stocks Fall?
The selloff did not start with TSMC. Benzinga, citing Reuters, reported that the Philadelphia Semiconductor Index fell 4.3% on Wednesday and was down about 11% over six sessions. TSM itself closed at $430 on October 14, down 3.4%, per Investing.com, though Yahoo Finance noted the stock was still up more than 41% year to date and had hit all-time highs earlier this month.
That context matters. A stock that has run that far does not need bad news to pull back. It only needs the good news to be fully expected. When a company has beaten estimates for six straight quarters, a seventh beat stops being a surprise and becomes the baseline. The stock is no longer paying you for being right about demand. It is asking whether demand can keep exceeding what is already assumed.
One analysis published by FXStreet argued the problem is financial rather than technological. It pointed to a US 10-year Treasury yield that has recently traded around 5.25%, and to roughly $50 billion of debt at AI cloud provider CoreWeave, whose bonds now yield around 15%, as signs that investors are growing less willing to pay any price for growth. Treat that as one writer’s read, not settled fact. But the logic is sound on its face. Higher yields lower the present value of profits that arrive years from now, and the AI trade is built on exactly that kind of profit. When safe money pays more, speculative money gets pickier.
The Case That This Cycle Is Different
Chip investors have seen this movie before. Demand surges, everyone builds capacity, supply catches up, prices collapse, and the stocks fall long before the earnings do. The old semiconductor cycle was brutal, and anyone who bought near a peak remembers it.
Steve Williams, who spent more than 25 years at Applied Materials, told Benzinga that investors are still stuck in fear of that old pattern, and that the current AI buildout “has broken the old demand cycle.” His argument rests on supply constraints rather than optimism. Advanced packaging has been the immediate bottleneck, he said, but fixing it would likely move the squeeze back to 2-nanometer capacity. And expansion there is limited by the equipment makers. In his words, “They can only go so far before ASML can’t make what they need.” Benzinga also reported that ASML has said its planned supply will not fully meet customer demand in 2027 or 2028.
There is a useful detail hiding in the flat capital budget too. Williams noted that an unchanged capex figure does not signal weaker demand, because equipment arriving this year was ordered months ago and some large tools carry lead times of six months or more. In other words, the spending you see today was decided long before today’s market mood.
Still, a bottleneck is not a guarantee. Constrained supply protects pricing only as long as the customers buying the chips keep earning enough to pay for them. That is the weak link, and TSMC cannot fix it.
Where the Real Risk Sits
TSMC sells to a handful of enormous buyers, and Nvidia and Apple are among the customers named in coverage of this quarter. Those buyers, in turn, depend on cloud providers and AI developers spending heavily on data centers. Gartner expects AI providers to spend nearly $1 trillion on AI infrastructure in 2026, though estimates vary widely depending on who is counting. The point is not the exact figure. The point is that the foundry’s order book is only as durable as the spending of the companies downstream.
That is why the tone around 2027 deserves attention. Wei said consumer and price-sensitive end markets are being challenged by rising component prices and macroeconomic uncertainty, and that TSMC is “being prudent in our business planning going into 2027.” Wei himself said the company is also evaluating Texas and in talks with Terafab, a fab project tied to Elon Musk, which tells you TSMC is still planning to build, not to retrench. Both things can be true. A company can expect strong demand and still acknowledge that the world around it is uncertain.
Markets are also watching a quieter policy thread. Bloomberg reported on October 16 that Taiwan plans to curb AI chip exports to China as part of a trade deal with the U.S. If that materializes, it would change the mix of who can buy, and at what volume. The details are still emerging, so watch how it is implemented before pricing it in.
What Retail Investors Can Take From This
The lesson is not that TSMC is cheap or expensive. It is that good news and good investments are different things. A record quarter tells you the business is healthy. It does not tell you what the stock should be worth, and the two can drift apart for months in either direction.
A few questions are worth asking whenever an AI chip name reports. Is the surprise in the numbers, or only in the narrative? What is the stock already assuming? What happens to the story if interest rates stay high, or if a major customer slows its spending? And how much of your own portfolio is quietly riding on the same handful of themes?
Prediction markets offer one rough sentiment gauge. Benzinga reported that Polymarket traders put about a 6% chance on an AI bubble bursting by December 31, on a contract that requires several severe events within 90 days rather than a mere stock correction, and 19% by June 30, 2027. Those are crowd guesses on a betting platform, not forecasts you should lean on. But they show how much of the fear is about timing rather than whether AI demand exists.
The honest read is mixed. The demand evidence is strong, the supply constraints are real, and the financing environment is less forgiving than it was a year ago. A falling stock after a record quarter is not a verdict on AI. It is the market asking what it costs to own the best business in the chain, and nobody has a settled answer to that yet.
This article is written for educational and informational purposes only and does not constitute financial or legal advice. The views and analytical frameworks presented draw on publicly available information and reported commentary from industry participants. Readers are encouraged to consult primary sources and form their own informed views on these complex topics.
