ASML Vs. Taiwan Semiconductor: Who Can More Fully Satisfy AI Demand?

ASML reported Q3 revenue of $13.06B, up 55.2% year over year, and EPS of $14.48 versus $10.66 consensus, although a $570.65M one-time investment gain boosted EPS. TSMC earned $4.60 per ADR versus $4.46 consensus; HPC contributed 68% of revenue, while gross margin was 66.4% as 2nm production represented 8% of wafer revenue. Both companies say AI demand exceeds available capacity: TSMC plans $60B–$64B of 2026 capex, while ASML says planned 2027–28 supply will not meet demand and tool price increases take full effect from 2028. The article favors TSMC’s capacity-expansion flexibility and lower forward multiple—about 21x earnings versus ASML’s 29x—while noting ASML’s pricing power and exposure to China export controls.

Analysis

The bottleneck is likely to migrate, not disappear: more lithography capacity only translates into sellable AI accelerators if TSMC can also ramp yields, advanced packaging and customer-ready capacity. That favors TSMC if execution holds, but makes headline fab additions a poor proxy for near-term shipments. Watch packaging throughput and leading-edge utilization alongside wafer capacity. ASML’s installed-base upgrades are a meaningful counterweight: they can lift customer output without waiting for entirely new fabs, while negotiated price increases support earnings later rather than solving the near-term supply gap. The risk is paying today for pricing power whose full financial contribution is deferred.

Over the next 1–3 months, the market may reward visible AI capacity and penalize margin dilution; TSMC’s new-node economics and overseas-fab costs matter more than demand commentary. Over 6–18 months, delivery schedules, yield learning and packaging availability determine whether TSMC converts capex into profitable volume. ASML’s China exposure and any export-control tightening remain downside catalysts; a weaker AI buildout or memory downturn would also challenge equipment demand. The contrarian point: ASML’s tool scarcity does not guarantee it captures more value than the foundry—customers can defer expansions, while TSMC can monetize constrained capacity through mix and utilization. But TSMC’s lower relative multiple is not protection if the ramp dilutes margins longer than expected.

Key Decisions for Investors

• Relative-value watch: favor TSM over ASML on pullbacks, rather than treating both as interchangeable AI exposure. Initiate only if TSMC’s next guidance preserves its capacity ramp without a material deterioration in gross-margin outlook; use ASML as a partial hedge, recognizing that an unexpectedly larger tool-capacity plan could reverse the relative trade.
• For the next earnings cycle, monitor TSMC advanced-packaging throughput, leading-edge yields/utilization and overseas-fab dilution. If capacity grows but shipments or margins fail to follow, reduce the TSM tilt; these are more useful falsifiers than demand commentary alone.
• Keep ASML exposure sized for delayed catalysts: its pricing gains are weighted toward 2028, while China export restrictions can impair orders sooner. Reassess after any material export-control change or evidence that customers are pushing out deliveries.
• Avoid extrapolating announced capex into near-term revenue. Verify customer delivery schedules, packaging capacity and node-level profitability before adding to either name; if AI spending or memory demand weakens, both may face order deferrals, with equipment orders potentially more cyclical.