TSM Taiwan Semiconductor

Research status: Primary-source verified through TSMC Q2 2026 results. Valuation basis: $418.95 ADR close on 21 August 2026.

TSMC remains the most important winner-agnostic control point in advanced AI compute. Q2 2026 revenue was $40.2B, gross margin 67.7% and operating margin 60.3%. Q3 guidance calls for $44.6–45.8B revenue with 65–67% gross margin, implying continued extraordinary demand despite a very large base.

The moat is the combination of leading-edge process yield, scale, trusted foundry neutrality, advanced packaging and customer ecosystem. AI/HPC demand and the 2nm ramp support continued above-industry growth, while CoWoS/advanced packaging remains a practical bottleneck.

At $418.95, however, the ADR already reflects TSMC's exceptional quality and AI scarcity. PowerFund views TSM as a benchmark-quality AI-infrastructure asset and a likely correction-buy candidate, but geopolitical tail risk and current crowding require a larger margin of safety than a normal compounder.

Investment case

TSMC manufactures leading-edge logic for nearly every major fabless AI compute platform. Customers compete intensely with one another, but many rely on the same foundry and packaging ecosystem.

> The PowerFund thesis is that TSMC captures winner-agnostic economics from AI/HPC because leading-edge process technology and advanced packaging are scarce, high-switching-cost capabilities. The thesis only works per ADR if process leadership and pricing power remain intact and overseas expansion/geopolitics do not destroy the return on capital.

Verified operating baseline — Q2 2026

• Revenue: $40.2B.
• Gross margin: 67.7%.
• Operating margin: 60.3%.
• Q3 revenue guide: $44.6–45.8B.
• Q3 gross margin guide: 65–67%.
• Q3 operating margin guide: 56–58%.

Strategic growth engines

1. 2nm — next leading-edge node and pricing/mix catalyst.
2. AI/HPC — accelerator, CPU and networking demand.
3. Advanced packaging / CoWoS — practical capacity bottleneck around AI accelerators.
4. Overseas fabs — geopolitical diversification, though with margin/capital tradeoffs.
5. Customer prepayments / capacity commitments — can improve visibility but require scrutiny on cancellation protection.

KPI ladder

PowerFund valuation scenarios

These are PowerFund estimates, not company guidance.

24 months

• Bear (20%) — $280–330: AI growth normalizes, overseas costs dilute margins and geopolitical risk premium rises.
• Base (55%) — $450–520: 2nm/AI demand sustains >15–20% earnings growth with margins remaining exceptional.
• Bull (25%) — $620–720: AI/HPC and packaging scarcity sustain higher pricing/utilization than consensus.

Probability-weighted midpoint is roughly $450–520/ADR.

60 months

• Bear: $300–380.
• Base: $700–900 if TSMC maintains leading-edge share through multiple nodes.
• Bull: $1,100–1,400+ if AI compute intensity remains structurally high and packaging/process leadership deepens.

• Q3 delivery against $44.6–45.8B guidance.
• Clean 2nm volume ramp and customer adoption.
• CoWoS/advanced packaging capacity expansion remaining fully utilized.
• Continued AI/HPC revenue growth above corporate average.
• Price increases / richer mix preserving gross margin.
• Overseas fabs ramping without excessive margin dilution.
• Customer commitments/prepayments increasing visibility.

• Taiwan geopolitical tail risk: irreducible and potentially catastrophic for ADR value.
• Capex intensity: leading-edge fabs/packaging require enormous annual investment.
• Overseas-fab economics: U.S./Japan/Europe diversification may lower returns/margins.
• AI concentration: a capex digestion period would hit TSMC despite customer breadth.
• Customer concentration: the largest AI/mobile customers remain economically important.
• Technology execution: 2nm/yield slips could open share to Samsung/Intel.
• Valuation/crowding: premium quality is widely recognized.
• FX: reporting/cost structure is sensitive to TWD/USD.

Warning — freeze additions

• Gross margin <60% without FX/overseas ramp explanation.
• 2nm yield/ramp materially slips.
• Leading AI customers diversify substantial leading-edge share away.
• AI/HPC growth <15% while capex remains >very high levels.
• CoWoS utilization/lead times normalize abruptly.

Reduce

• Gross margin <55% while utilization remains high.
• Material share loss across two leading-edge nodes.
• Overseas expansion structurally reduces ROIC with no geopolitical benefit capture.
• AI growth weakens sharply while advanced-node capacity continues expanding.

Invalidate

> Invalidate if TSMC loses sustained leading-edge process/packaging leadership or a Taiwan disruption fundamentally impairs the manufacturing base.

Competitive weakness is a slow-burn invalidator; geopolitical disruption is a binary hard invalidator.

TSMC's primary manufacturing alternatives are Samsung Foundry and Intel Foundry. Neither currently matches TSMC's combination of yield, capacity, ecosystem, customer neutrality and advanced packaging.

The moat compounds because leading customers design around proven process libraries and packaging capabilities years ahead of production. The dominant risk is not ordinary competition; it is geopolitics and the cost of geographic diversification.

1. Track 2nm yield, customer tape-outs and production allocation.
2. Quantify CoWoS/advanced-packaging capacity and utilization.
3. Model overseas-fab margin dilution and subsidies.
4. Track customer prepayments/cancellation protections.
5. Build ADR downside cases for blockade/sanctions/geographic-diversification scenarios.
6. Compare normalized FCF yield with ANET/CRDO/MRVL and less-crowded bottleneck names.
7. Define correction-readiness zones; prioritize dislocations rather than chasing quality.

Next full review: after Q3 2026 results or material geopolitical/valuation dislocation.