TSMC seen riding AI boom to fifth straight quarter of record profit

TSMC is expected to report Q2 net profit of T$632.6B ($19.65B), up 59% year-on-year, following a 36% jump in Q2 revenue—both driven by strong AI-chip demand for leading-edge nodes (3nm/2nm) and CoWoS advanced packaging. Investors will focus on whether TSMC raises full-year revenue growth (currently “above 30%”), with analysts suggesting upside and potential increases to 2026 capex from the $52B–$56B range (some forecasting ~$58B). The company is already investing $165B in U.S. (Arizona) chip fabs and its stock is up 56% YTD, slightly ahead of the broader market’s 54% gain.

Analysis

This is still a bottleneck story, not just a demand story. The market has already rewarded TSM for being the gatekeeper to AI capacity, so the next leg in the stock depends less on the profit beat and more on whether management signals that supply remains the constraint through 2025; a capex raise would reinforce that and likely widen the relative premium to the broader semi complex. The flip side is that higher capex is a mixed blessing: it confirms demand, but it also pulls forward depreciation and can cap free-cash-flow upside, which is where crowding risk starts to matter.

NVDA is the cleaner second-order winner because any evidence that TSM is still capacity-limited extends the duration of accelerated GPU shipments and reduces the odds of a near-term digestion phase. AAPL is the more exposed customer in relative terms: not because demand collapses, but because foundry allocation is increasingly skewed toward AI infrastructure rather than mature-device demand, so it is likely to lag on a relative basis if the print reinforces a priority shift. Memory names can also benefit if the market extrapolates the capex cycle, but that enthusiasm can reverse quickly if the industry interprets this as a late-cycle equipment arms race rather than durable end-demand.

The key catalyst window is the next 1-3 trading sessions for the headline reaction and 1-3 months for follow-through as consensus revises capex and FY revenue targets. The contrarian risk is that this is already well-owned: if TSM only reiterates current guidance, the stock can still sell off even on record results because the market wants proof of incremental supply expansion, not just confirmation of an already-strong AI narrative. The thesis is falsified if capex stays at the low end of the existing range or if management soft-pedals 2025 demand comments, which would be a warning sign that AI order flow is normalizing faster than bulls expect.

Key Decisions for Investors

• Buy TSM only on confirmation: if management raises FY revenue growth and nudges 2026 capex toward the high end or above current guidance, use a 1-3 month horizon with a tight stop on any post-earnings failure back below the breakout level; risk/reward is attractive because the market is paying for evidence, not just the beat.
• Relative value: long TSM / short AAPL for 1-3 months. Thesis is that foundry allocation and AI packaging scarcity will drive higher earnings revision momentum at TSM while AAPL remains a lower-beta customer with limited direct upside from the AI capex cycle.
• Add NVDA on TSM confirmation, but prefer call spreads rather than outright stock for the next 4-8 weeks. Upside comes from extended supply tightness; the risk is multiple compression if investors conclude the supply chain is merely keeping pace rather than accelerating.
• If TSM reiterates capex and does not lift full-year guidance, fade the initial rally in NVDA and TSM on the first strong bounce. That outcome would signal the AI trade is entering a digestion phase, and crowded longs could unwind quickly over days to weeks.
• Watch MU and memory peers for a separate trade only if pricing data improves; this TSM print alone is not enough to justify a durable long in the memory complex. Treat any move there as a confirmation alert, not a standalone buy signal.