Six consecutive record quarters. Net profit up by close to 70% year on year. And by the closing bell, down about 2%.
Taiwan Semiconductor Manufacturing Company reported its third quarter and beat its own guidance, which it had already raised. The demand driver is the same one powering every interesting line item in global technology right now, which is to say artificial intelligence and the advanced packaging and leading-edge nodes that make it physically possible. Nobody else can do this work at this volume at this yield. That is not marketing, it is a measured fact of the industrial world, and it is the reason a single company in northern Taiwan sits across the throat of a trillion dollars of ambition.
So the market sold it, on two things. Margin guidance came in softer than hoped, and capital expenditure is going up again.
Consider what those two complaints actually are. The first says that building out capacity at this speed costs money in the near term. The second says that the company intends to keep building out capacity. These are the same sentence, read twice in different tones of voice.
Anyone who has watched a serious barista dial in a new single origin will recognise the pattern. Six shots in a row come out clean and sweet and the room goes quiet in the right way. Then the barista reaches for the grinder and moves the setting, because the next bag is already in the hopper and it behaves differently, and for one shot the extraction wobbles. The impulse of the inexperienced customer is to say the coffee has got worse. What has actually happened is that the person behind the machine is spending a shot now to protect the next hundred. That is not a decline. That is the entire craft.
TSMC is spending margin now to protect the next decade of nodes. The capex line that spooked investors is the mechanism by which a company remains the only one capable of the thing everybody needs. Reading rising capital expenditure at a foundry as a negative signal requires believing that the demand will stop before the plants are finished, which is a position somebody can hold, but they should hold it out loud rather than smuggling it in as a concern about gross margin.
What the coverage underplayed is the structural asymmetry this creates. A near-70% profit increase at a company with TSMC’s scale does not just make money, it makes the next round of capital cheaper and the next process generation more defensible. The competitors chasing leading-edge capacity have to fund their catch-up out of worse economics. Every record quarter widens the gap it is being measured against, which is why the sixth one in a row matters so much more than the first. This is compounding in its industrial form, slow and heavy and extremely difficult to reverse.
There is also a geopolitical fact sitting underneath the earnings release that no margin discussion touches. The most contested sovereignty question in Asia now has a balance sheet attached to it, and that balance sheet just got stronger for the sixth quarter running. Governments on three continents are spending public money to build alternatives, and the alternatives are, at present, further behind than they were a year ago. That is the real result in the filing.
Six records. Seventy percent. Sold off 2% because the company intends to do it again.
Somewhere in Hsinchu, somebody read the share price, went back to the fab, and ordered more machines.
Source: RTE, “TSMC posts record quarterly profit on AI demand”
