TSMC reported third-quarter revenue of $46.72 billion, with a gross margin of 66.4% and an operating margin of 58.7%. The guidance for the fourth quarter is $52.3–54.3 billion at a gross margin of 65–67%.
Read that guidance twice. The company is telling the market it expects to add roughly six billion dollars of quarterly revenue in three months, having already set a record. A foundry cannot do that by selling harder; the capacity has to exist, be qualified, and be filled with work that was committed quarters ago.
The margin is the more interesting half. Holding two thirds gross margin while ramping a new node is unusual, because a ramp normally costs margin: yields start low, depreciation on new tools lands immediately, and the early wafers are expensive. Doing it anyway means the pricing of leading-edge capacity is set by scarcity rather than by cost, which is a market structure rather than a manufacturing achievement.
For anyone further down the chain the consequence is simple arithmetic. A wafer at the leading edge is now priced as a scarce asset, and that price arrives in your bill of materials as a processor or an accelerator you do not get to negotiate. It is also why mature-node parts, where capacity is not scarce, remain the sensible default for anything that does not genuinely need the newest process.
The guidance also sets the pace for everyone downstream. Packaging, substrate and test capacity has to grow at the same rate to convert those wafers into shippable parts, and historically it has not. Advanced packaging has been the real bottleneck for two years, which is why a record foundry quarter does not necessarily mean the part you ordered arrives sooner.
Source: TSMC