TSMC second-quarter profit jumps 23%, beating estimates, on high-end chip boom

TSMC reported Q2 profit up 23.4% and net income of NT$706.56B vs. NT$632.64B expected, sharply beating estimates. Revenue rose to NT$1.27T (+36% YoY) versus NT$1.264T expected, with advanced (7nm and under) technologies contributing 77% of wafer revenue. The company reiterated momentum from AI chip demand, and shares rose 1.23% Thursday after record-setting results.

Analysis

This print is less about a one-quarter earnings surprise and more about confirmation that the AI supply chain is still capacity-constrained at the highest-value nodes. That matters because the marginal winner is not just the foundry itself; it is the small set of customers with priority access to advanced wafers and packaging, which should keep shipment risk low for NVDA and AVGO into the next couple of quarters. The structural loser is any would-be share taker without comparable process technology or customer lock-in, especially lower-end foundry capacity that cannot participate in the same mix.

Near term, the market may already own much of the good news, so the first reaction can be muted even when fundamentals remain strong. Over 1-3 months, the key catalyst is whether management commentary and customer capex plans confirm that advanced-node demand is still running ahead of supply; if so, estimate revisions should flow through to NVDA and AVGO faster than to TSM, which is already valued as a high-quality bottleneck. Over 6-18 months, the real risk is not demand collapse but demand normalization combined with rising geopolitical/export-control friction, which would compress multiples across the AI semiconductor complex.

The contrarian read is that the consensus is treating this as a generic AI-demand check, but the more important signal is that scarcity remains intact enough to preserve pricing power and allocation discipline. That is bullish for the ecosystem, yet it also means the upside is increasingly concentrated in names that can monetize every incremental wafer more efficiently than TSM itself. Falsifiers would be a guide-down in advanced-node mix, a pause in customer capex, or any evidence that AI orders are simply being pulled forward rather than expanded.

Key Decisions for Investors

• Buy NVDA on a 2-4% post-print pullback; 1-3 month horizon. Thesis: TSM’s execution reduces shipment-delay risk and should support near-term estimate revisions. Prefer call spreads over outright calls to cap premium if the stock is already extended.
• Add AVGO on weakness, with a 1-3 month view. The read-through is that custom silicon and networking demand should stay supported as advanced-node supply remains tight. Risk/reward is best if the stock fails to re-rate on the first move and offers a better entry after the initial flush.
• Hold TSM, but only add on a 3-5% fade rather than chasing strength. The stock is high quality, but much of the AI bottleneck narrative is already embedded; invalidation would be any sign that advanced-node mix or forward commentary stops surprising to the upside.
• Pair trade: long NVDA / short INTC for 1-3 months. Use TSM as evidence that leading-edge capacity is still concentrated with the best operator, which should keep the relative gap wide if AI demand remains intact.
• Set an alert on the next TSM customer capex update and next-quarter guidance. If management fails to reaffirm strong advanced-node demand or if the semi complex sells off on volume after the print, treat the move as digestion rather than a fresh leg higher.