TSM (TSM) Q2 2026 Earnings Call Transcript

The market will likely parse this as a short-term margin headwind wrapped around a longer-duration demand signal. In the next 1-3 months, the risk is that investors over-focus on near-term gross margin dilution from node transition and overseas capacity, which can cap multiple expansion even as the revenue guide moves higher. But mechanically, a higher capex print from the clear market leader is usually a demand-pull indicator, not an overbuild signal; that matters because it validates the earnings power of the entire AI compute stack into 2027-28.

Second-order winners are the suppliers and customers tied to leading-edge capacity, not the mature-node cyclical names. The clearest beneficiary is NVDA: more front-end wafer investment is necessary but not sufficient, so the real bottleneck shifts to advanced packaging, substrate, and tool capacity; any easing there should improve shipment conversion over the next 2-4 quarters. Intel remains the structural loser on the relative-franchise basis: even with policy support, the bar to winning bleeding-edge customer trust keeps rising, while TSM's incremental U.S. buildout removes one of Intel's key narrative advantages.

The contrarian view is that consensus may be too bearish on the capex step-up because it treats it like a cost problem rather than an option on future scarcity pricing. The bigger falsifier is not margin dilution in the second half; it's any sign that AI wafer allocation or customer demand rolls over in the next two quarters, or that packaging/utility constraints force shipment delays. Absent that, the trade is to buy dips in TSM, not chase strength, and use Intel as the cleaner relative short.