Wall Street’s explanation was simple. Taiwan Semiconductor Manufacturing ADR (NYSE: TSM) beat earnings, then the stock fell, and many headlines hastily pinged it on the management raising capital spending by 15%, and investors concluding that higher spending today meant lower returns tomorrow. I think they’re looking at the wrong problem entirely.
TSM reported EPS of NT$27.25, or US$4.31 per ADR, on NT$1.27 trillion (US$40.2 billion) in second-quarter revenue. Revenue climbed 36% year-over-year, net income surged 77.4%, gross margin expanded to 67.7%, operating margin reached 60.3%, and net profit margin finished at 55.6%. None of these figures suggest that the business is straining under the weight of excessive investment. From my experience, you only get these from a company whose competitive advantage is becoming more expensive to build, and that gap is exactly where most investors are confused right now.
How The Market Scapegoated “CapEx”
Over the past year, investors have watched nearly every major AI company promise unlimited demand, unlimited spending, and unlimited returns on both. Skepticism has followed, reasonably, and so when TSM announced 15% increase in capital expenditure guidance, the market placed it in the same bucket as every other AI infrastructure spending story it has grown tired of trying to underwrite. That reaction is understandable. It’s also the wrong frame for this specific business, because TSM isn’t spending billions hoping customers eventually materialize. Its customers are already here, already committed, and already driving utilization levels that produced a 67.7% gross margin in the same quarter the spending increase was announced.
Look past the headline and the business composition tells you something the selloff obscures. High Performance Computing now generates 66% of total revenue, up from 61% last quarter and 60% a year ago, while smartphone revenue has fallen to 22% as AI infrastructure keeps absorbing a larger share of the wafer mix. The technology breakdown reinforces the same trend; 2nm has already reached 3% of wafer revenue despite only beginning production, 3nm increased to 30%, and 5nm remained a substantial 33%, with advanced nodes collectively representing 77% of total wafer revenue. That’s what additional capital spending is buying, a wider technological lead in the nodes that every serious AI hardware program on earth currently depends on, not more factories
Investors Are Measuring Returns At Exactly The Wrong Moment.
The easiest way to judge capital spending is to look at what it does to free cash flow today. The harder and more important question is what it enables the business to earn three years from now, and conflating those two time horizons is where most of the analytical error in this selloff is concentrated.
TSM deployed US$15.7 billion in capital expenditures during the quarter, bringing year-to-date investment to US$26.8 billion, and free cash flow declined as a direct result because capital expenditures rose faster than operating cash flow in the period. That’s the number that triggered the reaction. But the numbers sitting right beside it deserve equal weight in any honest assessment of what’s actually happening to this balance sheet. Operating cash flow climbed to NT$783.4 billion. Cash and marketable securities increased to NT$3.52 trillion. Net cash reserves reached NT$2.49 trillion. The current ratio held at a healthy 2.5x.
Like aforementioned, TSM isn’t stretching its balance sheet to fund expansion, it’s funding expansion from one of the strongest cash-generating businesses anywhere in the global semiconductor industry, and the margin profile during this investment cycle makes that even more striking. Even overseas fabs, which management acknowledged diluted consolidated margins, couldn’t prevent gross margin from expanding to 67.7%, because utilization remained exceptionally strong and manufacturing efficiency kept improving at the same time.
It’s simple, higher spending, higher margins, and higher profitability in the same quarter is not the financial signature of a company making a reckless bet. But the signature of a business investing from a position of structural strength, and those two setups produce very different outcomes over a three-to-five year horizon.
The Chart And The Conclusion
The post-earnings selloff looks dramatic when you focus on the daily candle. The longer-term structure looks considerably less concerning when you step back and read the full picture. Even after the decline, TSM remains comfortably above its 200-day moving average at $351.08, while price closed at $402.22, below the 20-day average of $438.68 but recovering after testing an intraday low of $397.21… a reversal that suggests buyers stepped in at a level that held rather than gave way.
Daily volume reached 1.24 million shares, reflecting institutional participation rather than the thin, panicked selling that accompanies a genuine thesis break. Price has pulled back toward the rising long-term trendline intact since April, and that’s historically where institutions make the decision between two very different interpretations: a broken business, or a more attractive entry point into an unbroken one.
The market spent the day debating whether US$15.7 billion in quarterly capital spending is too much for investors to absorb. I spent it asking a different question; if AI demand keeps accelerating, if 2nm follows the same adoption curve 3nm has already demonstrated, and if High Performance Computing keeps displacing smartphones as TSMC’s dominant revenue engine, today’s spending levels may look surprisingly conservative three years from now.
That’s why I wouldn’t remember this quarter as the one where TSMC raised CapEx. I’d remember it as the quarter where the market temporarily confused investment with risk. The two can look identical in the short term. Over a full cycle, they’re almost never the same thing.
