TSMC (2330 TT) is sharply accelerating its capacity expansion as Apple (AAPL), NVIDIA (NVDA), AMD (AMD), Qualcomm (QCOM), MediaTek (2454 TT) and others increase their 2nm bookings by 10-20% amid strong AI and HPC demand

TSMC (2330 TT) is sharply accelerating its capacity expansion as Apple (AAPL), NVIDIA (NVDA), AMD (AMD), Qualcomm (QCOM), MediaTek (2454 TT) and others increase their 2nm bookings by 10-20% amid strong AI and HPC demand.

Capacity acceleration at the leading edge is the clearest forward signal a foundry gives, since bookings of this kind are typically placed well ahead of wafer output and reflect customers locking in node allocation rather than reacting to spot demand. Past node ramps have followed a familiar sequence: early capacity is rationed among the largest clients, yields and pricing improve as volume builds, and the first customers to commit secure the most favourable windows, which is why a broad-based increase in bookings across mobile, GPU and HPC names matters more than any single client. The transmission runs beyond the foundry itself, into the equipment makers and packaging and substrate suppliers that follow capex with a lag, and into the competitive read-through for rival foundries whose ability to match the node has historically determined pricing power at the leading edge. The distinction worth drawing is between demand-led acceleration, where bookings rise across an entire customer set, and inventory-driven pull-forward, which tends to be concentrated in one or two names and unwinds quickly; a move spanning this peer set sits closer to the former. Worth noting next is whether management formalises the acceleration in capex guidance and how allocation is split between the named clients, since that ordering has tended to preview which product cycles hit their windows.