TSMC Q2: Still One Of The Best AI Stocks To Own

TSMC reported robust Q2 results led by strong demand and top-line growth, alongside sharp YoY gross and operating margin expansion, reinforcing a Buy rating. 2nm/3nm now contribute one-third of revenue, while high performance computing represents 66% of sales. Management flagged near-term gross-margin pressure as new technology ramps and CapEx increases, but the earnings momentum is clearly positive for the stock.

Analysis

TSM is increasingly functioning like the toll booth for AI compute rather than a cyclical semiconductor manufacturer. The key implication is that leading-edge mix is now the real moat: as 2nm/3nm share rises, smaller peers face a widening cost/performance gap they cannot easily close, while fabless leaders become even more dependent on TSM for time-to-market. That supports continued share gains for TSM and should keep valuation premium intact unless customer concentration begins to bite.

The near-term risk is not demand, but sequencing: capex and new-node ramps can suppress gross margin for 1-2 quarters even when end-demand is healthy. That creates an opportunity for the market to overreact on any slight margin guide-down, especially if investors confuse temporary absorption effects with weakening pricing power. The more durable read-through is that equipment vendors tied to advanced-node expansion should stay well bid on every capacity add cycle.

Contrarian take: consensus may be underestimating how much of the AI buildout is already embedded in expectations for the fabless ecosystem, while still underpricing TSM’s own strategic leverage. If AI demand normalizes faster than capex commitments, the penalty lands first on customers with weak differentiation, not on the foundry that sits at the center of supply. The thesis is falsified if leading-edge revenue share stalls, gross margin guidance cuts by more than ~150 bps, or order commentary points to a 2H demand air pocket rather than a temporary ramp issue.

Key Decisions for Investors

• Buy TSM on 3-5% post-earnings pullbacks; 3-6 month thesis is that any margin wobble from new-node ramp is temporary, while mix keeps improving. Falsify if management guides to sustained gross margin compression beyond ~150 bps or delayed 2nm adoption.
• Pair trade: long TSM / short INTC over 6-12 months. The spread should widen if foundry execution and advanced-node demand remain intact, while INTC continues to pay for capacity without comparable pricing power.
• Add to SOXX or SMH on weakness, but prefer a barbell with TSM as the core name and ASML/AMAT as secondary beneficiaries of the capex cycle. This captures both the operating moat and the equipment spend from incremental wafer capacity.
• Set an alert for any commentary on AI/HPC order digestion in the next 1-2 quarters; if utilization softens while capex stays elevated, reduce exposure because margin risk would shift from transitory to structural.