Taiwan Semiconductor Manufacturing: Record Profits Chart the Future of AI (NYSE: TSM)

TSMC reported Q2 revenue of NT$1.27T (about $40.2B), up 36% YoY, alongside gross margin +910 bps to 67.7% and net profit margin +1290 bps to 55.6%, driving record EPS of NT$27.25 (+77% YoY). Management raised capex to $60B–$64B (from $52B–$56B) and flagged that the next three years’ spend will be “significantly higher,” with an additional $100B planned for Arizona (total $265B) to support 2nm mass production and advanced packaging. For Q3, TSMC guided revenue of $44.6B–$45.8B (37% growth at the midpoint) and gross/operating margins of 66%/57%, reinforcing AI-chip demand despite investor concerns.

Analysis

TSM is effectively signaling that the AI supply chain remains capacity-constrained rather than demand-constrained. The near-term read-through is not just for the foundry itself; it is a stronger-than-expected indicator that hyperscaler and accelerator spending is still outrunning available advanced-node supply, which should keep order books firm for NVDA’s ecosystem and for the broader semiconductor complex over the next 1-3 months.

The market is likely overreacting to the capex headline as if it were a margin warning only. In reality, a step-up in spending raises the barrier to entry for competitors and extends TSM’s pricing power; the bigger second-order effect is that advanced packaging becomes a strategic chokepoint, which should tighten supply for any customer dependent on leading-edge AI accelerators. Over 6-18 months, that favors the most exposed AI compute names and the tools vendors that participate in node transitions, while lower-quality foundry alternatives remain structurally disadvantaged.

The contrarian risk is that investors are extrapolating every capex dollar into lower ROIC without considering that the company is buying future scarcity rents. The main falsifier is not capex itself but any evidence that AI-related revenue mix stalls, gross margin slips materially below the low-to-mid 60s as new fabs ramp, or that customer diversification fails and pricing weakens. If that happens, the stock deserves a de-rating even if revenue keeps growing, because the market is currently paying for both growth and moat preservation.

Key Decisions for Investors

• Buy TSM on any 3-5% post-earnings pullback; 1-3 month setup favors a re-rate if AI demand commentary is confirmed by peers. Risk/reward is attractive so long as gross margin holds near the mid-60s and forward revenue guide is revised up again.
• Pair trade: long TSM / short a weak AI-beneficiary basket or SOXX into the next sector rotation. Thesis is that TSM has the cleanest visibility into true AI capex, while more crowded semiconductor names are more vulnerable to multiple compression if spending broadens but margins don’t.
• Add NVDA only on weakness, not strength, and treat TSM as the cleaner bottleneck trade over the next quarter. If TSM’s capex expansion is validated by customer demand, NVDA benefits through supply assurance, but valuation leaves less room for error.
• Set a hedge trigger: if TSM forward P/E expands materially above ~30x while gross margin trends down for two consecutive quarters, fade the move with a partial short or call spread overwrite. That would signal the market is paying for growth without ROIC protection.