Taiwan Semiconductor reported a record $40 billion quarter and raised its AI chip spending plans. Investors cheered the revenue, then sold the stock anyway. The episode exposes the real story behind the AI boom.
When TSMC, the world's largest chip foundry, posted Q2 2026 revenue of about $40.2 billion and net profit of $22 billion up 77 percent year over year, the results were unambiguously strong. Then the stock fell about 4 percent. The culprit was not earnings. It was guidance: TSMC lifted its 2026 capital expenditure forecast to $60–64 billion, up from an earlier $52–56 billion.
For two years, investors tolerated sky-high AI spending on faith: build the data centers now, and the revenue will follow. That patience is fraying. Big-tech AI infrastructure spending is set to climb another roughly 50 percent in 2026, well over $600 billion, and the gap between what is being spent and what is being earned is now wider than the telecom bubble of 2001.
The physics of making the next generation of AI chips is genuinely expensive. An advanced fab costs tens of billions of dollars, consumes the power of a small city, and takes years to build. TSMC simply has to spend this money to keep supplying the chips that run every frontier AI model. The buildout is real, but the market now wants to see returns matching the outlays.
Analysts have noted that even in the selloff, the underlying fundamentals of the semiconductor companies remain strong: AI chips already account for roughly 61 percent of TSMC's revenue, and demand from hyperscalers shows no sign of collapsing. The correction may be a healthy stress test rather than a bubble bursting. Either way, it marks a shift in the AI story: the era of spending without scrutiny is over.