TSM · US
TSM.US: Q2'26 update — advanced foundry monopoly & AI capacity pricing
TSMC posted Q2 2026 revenue of US$40.20B (+33.7% YoY, +12% QoQ), net income NT$706.56B (+77.4%), 67.7% GM and 60.3% OM. Advanced nodes (≤7nm) are 77% of wafer revenue; HPC 66%. Q3 guide US$44.6–45.8B; FY26 revenue growth slightly above +40% USD; CapEx raised to $60–64B. The advanced foundry monopoly for AI GPUs/ASICs is intact, but CoWoS packaging bottlenecks and Taiwan geopolitics remain structural discount factors. See also NVDA, ASML, AMD, INTC.
Cite a section with a deep link, e.g. /en/r/tsm-us-research#thesis
Q2 2026 snapshot
- Q2 revenue
- USD 40.2B
- ADR EPS
- USD 4.31
- Gross margin
- 0.7
- Operating margin
- 0.6
Q3 / FY26 guidance
- Q3 revenue midpoint
- USD 45.2B
- Q3 GM guide
- 1
- FY26 revenue YoY
- 0
- FY26 CapEx midpoint
- USD 62.0B
Thesis
TSMC is the only foundry capable of scale delivery at advanced nodes for global AI compute infrastructure. Q2 2026 revenue US$40.20B (+33.7% YoY, +12% QoQ), net income NT$706.56B (+77.4%), GM 67.7% and OM 60.3% validate a core proposition: pricing power for AI GPUs/ASICs sits with TSMC, not the chip designers.
The premium holds only if:
- Advanced node monopoly — N3 30%, N5 33%, advanced (≤7nm) 77% of wafer revenue; Samsung Foundry and Intel Foundry still lag on yield and customer ecosystem;
- HPC platform dominance — 66% platform revenue maps directly to NVIDIA Blackwell, Broadcom custom ASICs, and AMD Instinct shipment slopes;
- CoWoS advanced packaging — capacity beyond logic dies (CoWoS-L/S) is a key bottleneck for AI cluster delivery; TSMC's ramp sets the industry ceiling.
C.C. Wei maintained FY26 revenue growth slightly above +40% USD and raised CapEx to $60–64B—a strong signal the AI CapEx cycle is still accelerating.
Business
Model: Pure-play foundry + advanced packaging (CoWoS), monetizing node leadership and utilization.
Flagships: N3/N5 logic; N2 ramp; CoWoS packaging.
Competitiveness: ~90%+ advanced-node share; packaging capacity and trust moat.
Strategy: N2 ramp, geographic diversification, CapEx ~$60–64B for AI demand.
Value chain
Mid-stream advanced manufacturing between tool vendors and chip designers.
Upstream: ASML EUV. Downstream: NVDA / AVGO / AMD / Apple. Risks: Taiwan concentration; EUV/CoWoS ramp.
Valuation
TSMC ADR (TSM) has long traded in a band of "advanced foundry monopoly premium minus Taiwan geopolitical discount." Post Q2 (~2026-07-16), expectations for N3/N2 ramp and CoWoS expansion rose further, but geopolitical risk cannot be fully diversified via Arizona / Japan / Germany fabs.
Valuation frame
| Metric | Status | Read |
|---|---|---|
| 营收增长 | 强劲 | Q2 +33.7% YoY；FY26 指引略超 +40% USD |
| 毛利率 | 高位 | Q2 67.7%；Q3 指引 65–67% |
| 竞争壁垒 | 垄断级 | 先进制程代工份额 ~90%+；N3/N5 主导 AI 芯片 |
| CapEx | 上修 | FY26 $60–64B；N2/Arizona/Japan 扩产 |
| 地缘风险 | 结构性折价 | 台湾集中产能；ADR 长期含风险溢价 |
| CoWoS 瓶颈 | 缓解中 | 封装产能扩张是 AI 出货斜率关键约束 |
Versus NVDA: TSMC is the "pick-and-shovel" play—GM 68% vs NVDA ~75%, but TSMC does not carry product-cycle or software ecosystem risk; multiples typically below NVDA but above traditional semis.
Versus ASML: ASML is the sole EUV equipment supplier; TSMC is its largest customer—a dual monopoly upstream/downstream in the AI capacity expansion chain.
Financial trend (~24 months)
Eight-quarter revenue and gross margin with YoY and QoQ. Semi/AI hardware seasonality is secondary to supply, ASP, and order timing.
Revenue · last 8 quarters
Interactive chart available in the reader.
Gross margin · last 8 quarters
Interactive chart available in the reader.
Operations
Q2 2026 operations
Q2 P&L and structure
| Item | Q2 2026 | Note |
|---|---|---|
| 营收（USD） | $40.20B | +33.7% YoY；+12% QoQ |
| 营收（NT$） | NT$1,270.38B | — |
| 净利润（NT$） | NT$706.56B | +77.4% YoY |
| EPS（NT$） | NT$27.25 | — |
| EPS（ADR） | US$4.31 | — |
| 毛利率 | 67.7% | 先进节点 mix 驱动 |
| 营业利润率 | 60.3% | 运营杠杆显著 |
| HPC 占比 | 66% | AI GPU / ASIC 为主力 |
Node mix
| Node | Mix | Watchpoint |
|---|---|---|
| N2（2nm） | 3% | 2025 量产起步；2026 爬坡 |
| N3（3nm） | 30% | Apple / NVDA / AMD 主力节点 |
| N5（5nm） | 33% | 成熟先进；HPC 仍大量出货 |
| N7（7nm） | 11% | 逐步向 N5/N3 迁移 |
| 先进节点（≤7nm） | 77% | wafer revenue 占比 |
| HPC | 66% | platform revenue；AI 驱动 |
Takeaways:
- HPC 66% directly maps AI demand—NVIDIA GPUs, Broadcom AI ASICs, AMD Instinct all in N3/N5 production;
- N2 3% starts mass production; 2026–2027 is the critical ramp—N2 yield and capacity define the 2027+ AI chip performance curve;
- GM 67.7% at historical highs reflects advanced node ASP and utilization;
- OM 60.3% shows operating leverage—revenue +34% while profit +77%.
Node mix
TSMC's node mix is a key AI supply-chain metric:
| Node | Q2 mix | Main customers / use |
|---|---|---|
| N2 | 3% | 2025 mass-prod start; Apple / NVDA next-gen |
| N3 | 30% | NVDA Blackwell, Apple A/M, AMD |
| N5 | 33% | Mature advanced; heavy HPC volume |
| N7 | 11% | Migrating to N5/N3 |
| ≤7nm total | 77% | wafer revenue |
N3 + N5 at 63% is the main manufacturing node for AI chips. N2 starting at 3% will ramp through H2 2026–2027—the performance base for next-gen AI accelerators (Rubin and beyond).
AI & HPC demand
How the AI compute cycle flows through TSMC:
- GPUs — NVIDIA Blackwell in N4/N3 production, largest single driver of HPC 66%;
- Custom ASICs — Broadcom, Google TPU, Amazon Trainium/Inferentia, Microsoft Maia all tape-out at TSMC advanced nodes;
- CPUs — AMD EPYC and Intel server CPUs provide the base, but AI elasticity comes from GPU/ASIC;
- CoWoS packaging — HBM + GPU die 2.5D/3D packaging is the physical bottleneck for AI clusters; TSMC CoWoS ramp sets industry delivery ceiling.
Key chain: ASML EUV delivery → TSMC N3/N2 capacity → CoWoS packaging → NVDA GPU shipments → SMCI / CLS server assembly. TSMC sits at the "capacity switch" in the mid-chain.
Competition
Strategy / customers / risk map
| Category | Detail | Implication |
|---|---|---|
| 先进代工 | N3 30%、N5 33%；N2 3% 起步 | AI GPU/ASIC 产能定价权在 TSMC |
| 先进封装 | CoWoS / SoIC 扩产 | 封装瓶颈 narrative 仍是行业约束 |
| 地缘分散 | Arizona、Japan、Germany 建厂 | 成本更高；无法完全复制台湾效率 |
| 核心客户 | NVDA、AVGO、AMD、Apple、Qualcomm | HPC 66% 反映 AI 资本开支周期 |
| 设备依赖 | ASML EUV 独家供应 | 设备交付节奏影响节点爬坡 |
| 竞争 | Samsung、Intel Foundry 追赶 | 份额差距仍大；Samsung 良率/生态弱 |
Three axes:
| Axis | TSMC | Samsung Foundry | Intel Foundry |
|---|---|---|---|
| Advanced nodes | N3/N2 in production; 77% ≤7nm | 3nm yield/customer weak | 18A catching up |
| AI customers | NVDA, AVGO, AMD, Apple | Some Qualcomm | Internal + limited external |
| Packaging | CoWoS leader | I-Cube catching up | EMIB/Foveros |
| Geo | Taiwan concentrated + overseas | Korea + US | US-focused |
Samsung still faces yield and customer trust issues on GAA 3nm; Intel Foundry 18A has not won external mass-production orders. TSMC's advanced foundry share remains ~90%+ with no credible challenger near-term.
Peer comparison
| Company | Share / role | GM | Strength | Weakness |
|---|---|---|---|---|
| TSMC (self) | Adv. foundry ~90%+ | ~68% | N3/N2 + CoWoS | Taiwan concentration |
| Samsung Foundry | Distant #2 | Lower | GAA process bet | Yield / trust gap |
| Intel Foundry | Catch-up | Subscale | US capacity + CHIPS | External wins still thin |
Management
CEO C.C. Wei long-tenured; finance/comms bench stable. No material operating C-suite upheaval in 24m (board chair transitions watched). Stability: steady.
Key management (24m)
| Role | Name | Since | 24m change |
|---|---|---|---|
| CEO | C.C. Wei | 2018-06 | No change |
| CFO / Spokesperson | Wendell Huang | 2019 | No material change |
| Chair | Mark Liu (through transition) / board | — | Chair succession monitored; ops continuity via CEO |
Outlook
Company guide · Q3 / FY26
| Item | Guide | Note |
|---|---|---|
| Q3 营收 | US$44.6–45.8B | 中值 ~$45.2B；+38% YoY 量级 |
| Q3 毛利率 | 65–67% | N2 爬坡初期略稀释 mix |
| Q3 营业利润率 | 56–58% | — |
| FY26 营收增长 | 略超 +40% USD | AI 需求持续上修 |
| FY26 CapEx | $60–64B | 较此前指引上修 |
Framework:
- Q3 revenue US$44.6–45.8B (midpoint ~$45.2B) implies ~+38% YoY—AI demand not slowing;
- Q3 GM 65–67% slightly below Q2 67.7%, reflecting early N2 ramp mix dilution;
- FY26 CapEx $60–64B confirms TSMC is still in "expand, not contract" mode—N2, Arizona Fab 21, Japan JASM are priorities;
- Full-year revenue growth "slightly above +40% USD" is an unusually strong guide, implying H2 2026 still has acceleration room.
Scenarios
| Scenario | Conditions | Implication |
|---|---|---|
| Bull | Q3 >$46B; N2 yield beats; CoWoS bottleneck eases; FY26 >+45% | Capacity pricing power strengthens; ADR continues up |
| Base | Q3 ~$45B; FY26 +40% USD; CapEx $62B on track | Steady growth; hold as core allocation |
| Bear | AI CapEx slows; N2 yield issues; Taiwan Strait escalation | ADR -15–25%; geo discount widens |
Positioning frame (research only): core AI infrastructure holding; geopolitical risk requires position-size discipline. Not an all-in bet but a solid anchor in a semi portfolio.