66.2% Gross Margin and a 10% Sequential Q2 Guide Signal AI Demand Escape Velocity — Buy the Sell-Off
Initial Read: Comprehensive beat across all metrics — revenue at the top of guidance ($35.9B), gross margin 66.2% (120bp above the guidance ceiling), EPS $3.49 per ADR vs. $3.31 consensus; Q2 guide of $39.0–40.2B (+10.3% QoQ) implies an FY26 run-rate well above $160B. The 2.5–3% post-release stock decline is macro-driven, not fundamental — add on weakness.
Key Takeaways
- Revenue of $35.9B (NT$1,134.10B) landed at the top of the company's $34.6–35.8B guidance range and beat LSEG consensus by 0.6%; the real story is the gross margin of 66.2%, which shredded the guidance midpoint by 220bp and the guidance ceiling by 120bp — a degree of outperformance essentially unheard of for a company that guides conservatively by design.
- HPC/AI platform surged to 61% of wafer revenue (up 20% QoQ), driving a 58.3% year-over-year jump in net profit to NT$572.48B; management raised the AI accelerator revenue CAGR estimate for 2024–2029 to 54–56% from 45% previously — a structural upgrade to the demand outlook, not a routine tweak.
- Q2 2026 guidance of $39.0–40.2B (midpoint +10.3% QoQ) is exceptional for a company of this scale; with H1 at ~$75.5B, FY26 of >$160B appears well within reach if H2 sustains even modest momentum; capex guidance was nudged to "toward the high end" of $52–56B, signaling management's conviction in sustained demand visibility.
- N2 node is in high-volume manufacturing with solid yields at Fab 20/22; N3 gross margin will cross the company average in H2 2026, eliminating a structural headwind that has compressed margins for six quarters — the margin expansion story has a meaningful second leg.
- Rating: Maintaining Outperform. Results and guidance strongly reinforce the AI-driven investment thesis; any weakness on macro or "sell the news" dynamics is an accumulation opportunity, not a red flag.
Results vs. Consensus
| Metric | Actual | Consensus / Guide | Beat/Miss | Magnitude |
|---|---|---|---|---|
| Revenue (USD) | $35.9B | ~$35.5B (LSEG) | Beat | +$0.4B (+1.1%) |
| Revenue (NT$) | NT$1,134.1B | NT$1,127B (LSEG) | Beat | +NT$7.1B (+0.6%) |
| Net Income (NT$) | NT$572.48B | NT$543.32B | Beat | +NT$29.2B (+5.4%) |
| EPS per ADR (USD) | $3.49 | $3.31 | Beat | +$0.18 (+5.4%) |
| EPS per Share (NT$) | NT$22.08 | ~NT$20.94 implied | Beat | +NT$1.14 (+5.4%) |
| Gross Margin | 66.2% | 63.0–65.0% (guide) | Beat | +120–320bp vs. guide |
| Operating Margin | 58.1% | 54.0–56.0% (guide) | Beat | +210–410bp vs. guide |
| Net Margin | 50.5% | ~47–49% (implied) | Beat | +150–250bp |
Year-Over-Year Comparisons
| Metric | Q1 2026 | Q1 2025 | YoY Change |
|---|---|---|---|
| Revenue (NT$) | NT$1,134.10B | NT$839.25B | +35.1% |
| Revenue (USD) | $35.9B | ~$25.5B | +40.6% |
| EPS per Share (NT$) | NT$22.08 | ~NT$13.95 | +58.3% |
| EPS per ADR (USD) | $3.49 | ~$2.21 | +57.9% |
| Gross Margin | 66.2% | ~53–55% (est.) | +~1,100–1,300bp |
Quarter-Over-Quarter Comparisons
| Metric | Q1 2026 | Q4 2025 | QoQ Change |
|---|---|---|---|
| Revenue (USD) | $35.9B | ~$33.7B | +6.4% |
| Gross Margin | 66.2% | 62.3% | +390bp |
| Operating Margin | 58.1% | ~53.7% | +410bp |
| Net Margin | 50.5% | ~48% (est.) | +~250bp |
Quality of Beat
Revenue: Organic and AI-demand driven. No acquisitions distort the figure. Revenue landed at the very top of the guidance range ($35.8B ceiling; actual $35.9B), reflecting both execution and a favorable demand mix shift. The modest top-line beat relative to consensus (~0.6%) understates the quality of the quarter — the bigger story is what happened below the revenue line.
Margins: The 66.2% gross margin is the defining data point of Q1 2026. The 390bp sequential expansion from Q4 2025's 62.3% reflects three reinforcing dynamics: (1) technology mix richness — 5nm and 3nm together are 61% of wafer revenue and those nodes are approaching or at margin parity with older nodes; (2) exceptionally high fab utilization driven by AI demand that is capacity-constrained, not demand-constrained; and (3) better-than-expected management of overseas ramp dilution (new fabs typically drag margins 2–3% during tool installation phases). The operating margin beat of 210–410bp above guidance demonstrates that scale leverage is flowing through with minimal operating expense creep.
EPS: The 58.3% year-over-year EPS jump is operationally driven — revenue up 40.6% YoY compounded by significant margin expansion. Share count has been stable; there are no material below-the-line distortions evident in preliminary data. The FX tailwind is modest but real: TWD strengthened ~2.8% against USD year-over-year, meaning USD ADR EPS grew slightly faster than NT$ EPS.
Segment Performance
Platform Revenue Mix
| Platform | % of Revenue | USD (~) | QoQ Growth | Assessment |
|---|---|---|---|---|
| HPC (incl. AI accelerators) | 61% | ~$21.9B | +20% | Dominant driver; accelerating |
| Smartphone | 26% | ~$9.3B | -11% | Normal Q1 seasonal; not alarming |
| IoT | 6% | ~$2.2B | +12% | Recovering from trough |
| Automotive | 4% | ~$1.4B | -7% | Inventory correction ongoing |
| DCE | 1% | ~$0.4B | +28% | Small base; encouraging |
HPC / AI — The Engine
HPC revenue of ~$21.9B (61% of total) grew 20% sequentially in a quarter that saw total company revenue grow only 6.4% — meaning the rest of the portfolio contracted in aggregate while AI drove explosive gains. The implied Q4 2025 HPC revenue was ~$18.2B (~54% of ~$33.7B total), confirming an ongoing structural shift in the mix toward AI. Management raised the AI accelerator revenue CAGR for 2024–2029 to 54–56%, up from 45% previously — this is not routine guidance refinement; it reflects an observation that AI inference and training capacity deployments are accelerating faster than the already-aggressive prior forecast.
Assessment: HPC at 61% of revenue is a structural milestone. Two years ago, mobile chips were the plurality. The demand environment for AI accelerators is management's word: capacity-constrained, not demand-constrained. This matters enormously for pricing power and margin sustainability.
Smartphone — Seasonal, Not Structural
The 11% QoQ decline in smartphone revenue is Q1 seasonality — mobile OEMs routinely reset inventory in the January–March period after the holiday ramp. No competitive displacement is evident. Mobile still represents 26% of revenue, and Q2 typically sees modest recovery.
Assessment: Not concerning. Monitor Q2 for normalization back toward 28–30%.
Automotive — Inventory Correction Persists
Automotive declined 7% QoQ to ~$1.4B (4% of revenue). The automotive semiconductor inventory correction that began in 2024 continues. TSMC has limited direct automotive customer exposure relative to its downstream chip design customers (Infineon, NXP, Renesas design for the auto OEMs), but the trend suggests the correction is not yet complete.
Assessment: Automotive represents only 4% of revenue; any recovery would provide marginal upside but is not thesis-determinative.
Technology Node Revenue Mix
| Node | % of Wafer Revenue | USD (~) | Characterization |
|---|---|---|---|
| 3nm (N3/N3E/N3P) | 25% | ~$9.0B | Strong ramp; approaching margin parity |
| 5nm (N5/N4) | 36% | ~$12.9B | Mature node; likely above company avg. margin |
| 7nm (N7/N6) | 13% | ~$4.7B | Workhorse; high-margin mature node |
| Advanced (≤7nm combined) | 74% | ~$26.6B | 74% of revenue at leading-edge pricing |
| Older nodes (>7nm) | 26% | ~$9.3B | Stable; mature node utilization |
The 74% advanced-node revenue mix is a record and drives the margin profile directly. The 5nm node (36%) is now sufficiently mature that it likely runs above the company's gross margin average, while 3nm (25%) is approaching that crossover — management expects N3 to cross the company average in H2 2026. N2, just entering high-volume manufacturing, will follow the same ramp curve N3 did: dilutive initially, then margin-accretive as yields and utilization improve.
Notable Items in the Release
1. AI Accelerator CAGR Raised to 54–56%
Management raised the revenue CAGR for AI accelerators (2024–2029) from 45% to 54–56%. This is a substantial upward revision — over a five-year period, the compounding difference between 45% and 55% CAGR produces revenue that is roughly 65% higher at the endpoint. The revision reflects observed demand from hyperscalers, AI labs, and increasingly from model inference workloads, which are proving to be a second major demand wave behind training.
Assessment: This is the most strategically significant disclosure in the release. A CAGR revision of this magnitude, by a company that has historically been conservative in its forward estimates, signals genuine demand visibility rather than aspirational marketing.
2. N3 Gross Margin Crossing Company Average in H2 2026
Management guided that N3's gross margin will cross the corporate average in the second half of 2026. N3 nodes have been margin-dilutive since their introduction due to high initial tooling costs and suboptimal yield curves. Once N3 crosses the company average, one of the two primary drags on gross margin (N3 + overseas fabs) is eliminated. The gross margin expansion story therefore has a second leg beyond what Q1 2026 already demonstrated.
Assessment: Highly significant for the margin outlook. If N3 crosses average in H2 2026, gross margins could sustain in the 66–68% range even as N2 ramps — the N2 ramp dilution would be offset by N3 maturity.
3. Capex Moves to High End of $52–56B Range
TSMC's original 2026 capex guidance was $52–56B (issued at Q4 2025 results in January). The company has now indicated capex will be "toward the high end" of that range, implying $55–56B. This reflects demand confidence: TSMC only accelerates capacity investment when it has customer-contracted visibility. The incremental capex at the high end (~$2–4B above midpoint) is consistent with pulling in equipment deliveries for AI-specific capacity ahead of schedule.
Assessment: Bullish signal on the demand environment. Companies don't commit an extra $2–4B in capex without high-confidence customer demand commitments, typically in the 2–3 year lead time window.
4. US-Taiwan Tariff Relief
A US-Taiwan trade accord reached on January 15, 2026 reduced tariffs on Taiwanese goods from 20% to 15% and granted duty-free treatment for semiconductor manufacturing equipment imported during construction phases. For TSMC's Arizona buildout — which involves importing billions of dollars of specialized tooling — this equipment exemption materially reduces the effective cost premium of US-based manufacturing. The remaining 15% tariff on finished goods (chips) is relevant to TSMC's chip-fabbing customers but less so to TSMC directly, as TSMC sells wafers rather than finished devices.
Assessment: Modest but real positive. The equipment duty-free provision is directly margin-accretive for the Arizona buildout over the next 3–5 years. The broader tariff reduction reduces headline risk but doesn't fundamentally change TSMC's cost structure since its primary cost base is in Taiwan.
5. N2 High-Volume Manufacturing Confirmed
N2 (2nm-class) entered high-volume manufacturing in Q4 2025 and is ramping at Fab 20 (Hsinchu) and Fab 22 (Kaohsiung) with solid yields. This is important context for the gross margin trajectory: N2's initial ramp will create some dilution (as all new nodes do), but management's ability to guide 65.5–67.5% gross margin in Q2 despite N2 ramping suggests the dilution is well-contained.
Assessment: Eliminates a key technology execution risk. TSMC's N2 node is in production ahead of any meaningful competitive threat. Intel 18A remains in limited qualification and Samsung's 2nm is not in volume production.
6. TSMC Arizona: $165B Commitment, Third Fab Under Construction
The $165B Arizona commitment (up to 12 fabs) continues to build out: Phase 2 construction is complete (tool installation begins 2026; production H2 2027); a third fab is under construction; a fourth fab and advanced packaging facility have applications submitted. The scale of this commitment — both in dollars and in number of fabs — is a multi-decade statement on the permanence of US semiconductor manufacturing policy support.
Assessment: The Arizona expansion is strategically and geopolitically necessary rather than purely economically optimal (US fabs cost ~2x Taiwan). The tariff relief and CHIPS Act subsidies narrow the cost gap. Long-term, geographic diversification reduces the Taiwan-concentration risk that has historically applied a valuation discount to TSM shares.
Guidance & Outlook
| Metric | Q1 2026 Actual | Q2 2026 Guide Low | Q2 2026 Guide High | Midpoint Change |
|---|---|---|---|---|
| Revenue (USD) | $35.9B | $39.0B | $40.2B | Midpoint $39.6B, +$3.7B (+10.3%) |
| Gross Margin | 66.2% | 65.5% | 67.5% | Midpoint 66.5%, roughly flat QoQ |
| Operating Margin | 58.1% | 56.5% | 58.5% | Midpoint 57.5%, -60bp QoQ |
The Q2 2026 revenue guide is the defining forward data point. A $39.0–40.2B quarter for TSMC — up 10.3% sequentially and up more than 40% year-over-year — is not a normal seasonal pattern. Q2 is historically a period of moderate smartphone recovery but not dramatic sequential expansion at the company level. The 10%+ QoQ guide implies AI/HPC demand acceleration into Q2 is absorbing not just seasonal smartphone recovery but is providing significant incremental uplift.
Implied Full-Year 2026 Scenario: Q1 actual $35.9B + Q2 guided midpoint $39.6B = H1 total ~$75.5B. TSMC's full-year guidance is ">30% USD growth" from FY25's $122.4B, implying a minimum of ~$159.1B. To hit $159.1B on an H1 of $75.5B, H2 needs to deliver $83.6B, or ~$41.8B per quarter. With Q2 already at $39.6B and AI demand described as "extremely robust," H2 quarters averaging $41–43B appears comfortably achievable and potentially conservative. Our working estimate for FY26 revenue is $163–168B, implying 33–37% growth.
Guidance Style: TSMC has a well-established pattern of conservative guidance, particularly on gross margins, where it has beaten the midpoint in 9 of the last 12 quarters. The Q2 GM guidance midpoint of 66.5% is approximately in line with Q1 actuals (66.2%), suggesting the company is guiding for stability rather than mean-reversion — itself a bullish signal, as the initial expectation was that N2 ramp dilution would compress margins in H1 2026.
Capex Guidance: "Toward the high end" of $52–56B translates practically to ~$55–56B. For context, FY25 capex was approximately $40.9B, so FY26 capex represents a ~35–37% increase year-over-year. This magnitude of investment growth, made in a single year, signals multi-year demand confidence. TSMC does not commit this level of capital without contracted customer commitments extending to 2028 and beyond.
Questions for the Call
- N2 Gross Margin Trajectory: Management guided N3 to cross the corporate average in H2 2026. When does N2 cross the average? N2 entered HVM in Q4 2025 — if N3 took approximately 4–6 quarters to reach margin parity, N2 could cross in 2027. A bullish answer (earlier than expected crossover) would be meaningful for the FY27 margin outlook. A bearish answer (N2 margin pressured by new fab construction costs in Hsinchu and Arizona) would introduce incremental caution.
- Tariff Scenario Modeling: The US-Taiwan tariff deal reduced tariffs to 15% from 20% on finished goods. How does TSMC assess residual risk from additional tariff escalation? Are there scenarios in which semiconductor tariffs specifically targeting foundry services could be applied? Bullish: management characterizes the current deal as stable and multi-year. Bearish: management hedges on future tariff scenarios, particularly as geopolitical tensions evolve.
- Arizona Cost Premium — Current P&L Impact: TSMC's US fabs cost approximately 2x Taiwan to operate. With Phase 2 tooling starting in 2026 and production in H2 2027, how much of this cost premium is flowing through the P&L now (during construction) versus what will flow through at production ramp? Bullish: CHIPS Act grants and equipment duty-free provisions substantially close the cost gap. Bearish: the ramp cost is larger than disclosed.
- N2 vs. N3 Capacity Allocation — Hyperscaler Pull-Forward: Are AI hyperscalers already booking N2 capacity for 2027 production in meaningful volume? This matters because N2 demand pull-forward would indicate the next leg of the AI capacity wave is already committed, reducing uncertainty in 2027 revenue. Bullish: N2 is oversubscribed for the first year of volume production. Bearish: N2 demand is solid but not yet at the level that N3 was when it entered HVM.
- Intel/Terafab and Intel Foundry Partnership: On March 21, 2026, Terafab (Tesla/SpaceX/xAI) announced a joint venture with Intel to build 1 terawatt of AI compute capacity in Austin, with Intel contributing manufacturing expertise. What is TSMC's posture on advanced packaging or co-manufacturing partnerships with Intel? Bullish: TSMC sees this as a customer opportunity (Intel using TSMC wafers for the compute cores, with Intel handling packaging). Bearish: TSMC views any Intel manufacturing partnership as competitive encroachment.
- FY26 ">30%" Revenue Guide — Floor or Midpoint? With Q1 already at $35.9B and Q2 guided to $39.6B midpoint, the ">30%" full-year guide appears very conservative (our model implies 33–37% growth). Is this guide a floor, a midpoint, or a ceiling? Bullish: management acknowledges the conservative framing and notes demand remains "extremely robust" with supply as the binding constraint. Bearish: management reaffirms ">30%" as the appropriate range, suggesting visible macro uncertainty in H2.
Market Reaction
After the release, TSM ADR shares declined approximately 2.5–3.0% from a pre-announcement close of ~$377, trading down to the ~$365 range. This decline occurred despite a broad beat across all financial metrics and exceptional forward guidance. Two dynamics appear to be driving the move:
- "Sell the News" Positioning: TSM had appreciated significantly in the weeks heading into the print, as AI semiconductor sentiment remained elevated. Investors who bought in anticipation of a strong report took profits on confirmation.
- Macro and Geopolitical Headwinds: Broader market concerns — tariff uncertainty (despite the US-Taiwan relief), geopolitical risk premium on Taiwan-exposed names, and general risk-off sentiment — weighed on the tape independent of TSMC-specific fundamentals.
Our assessment: The decline is fundamentally disconnected from the results. A company that just delivered 66.2% gross margin (beating the guidance ceiling by 120bp), raised its AI demand growth forecast, guided to 10% sequential revenue expansion in Q2, and moved capex to the high end of its range does not deserve to trade down. This is the kind of noise that creates entry points. We would view any further weakness into the $355–365 range as an attractive accumulation opportunity for long-term holders.
Model Implications
| Item | Prior View | Post-Earnings View | Reason |
|---|---|---|---|
| FY26 Revenue (USD) | ~$155–158B | ~$163–168B | Q1 beat + Q2 guide imply H1 of ~$75.5B; H2 $88–93B on AI trajectory |
| FY26 Gross Margin | 63.5–64.5% | 65.5–67.0% | Q1 66.2% sets new floor; N3 margin crossover in H2 provides second leg |
| FY26 Operating Margin | 55–56% | 57–58.5% | Gross margin upside flows through with stable OpEx |
| FY26 EPS per ADR (USD) | ~$14.00–15.00 | ~$16.00–17.50 | Revenue + margin beats compound significantly at this scale |
| FY26 Capex (USD) | ~$53–54B | ~$55–56B | Management guided "toward high end" of $52–56B range |
Valuation: At ~$365 per ADR and revised FY26 EPS of ~$16–17.50, TSM trades at approximately 21–23x forward earnings — a premium warranted by the AI demand duration, TSMC's foundry monopoly at leading edge, and the structural margin expansion story. Relative to the S&P 500's ~21x forward P/E, TSM is not egregiously expensive for a company with 35–40% revenue growth and expanding margins. On a PEG ratio basis (P/E divided by growth rate), TSM looks outright cheap relative to peers.
Thesis Scorecard
| Thesis Point | Status | Notes |
|---|---|---|
| Bull: AI is a structural demand driver, not a cyclical peak | Confirmed | HPC at 61% of revenue (up 20% QoQ); AI CAGR raised to 54–56% through 2029 — structural acceleration, not deceleration |
| Bull: TSMC has an unassailable process technology lead | Confirmed | N3 at 25% of revenue, N2 in HVM with solid yields; Intel 18A and Samsung 2nm remain in limited qualification — 2+ node gap intact |
| Bull: Margin expansion as advanced nodes mature | Confirmed | 66.2% GM beats guidance ceiling by 120bp; N3 margin crossover in H2 2026 provides second leg of expansion |
| Bull: Geographic diversification reduces geopolitical risk discount | Partial | $165B Arizona commitment with 3rd fab under construction is directionally positive; US-Taiwan tariff deal reduces trade friction; but military risk premium on Taiwan-sited capacity is unchanged |
| Bull: AI accelerator CAGR durability through 2029 | Confirmed (upgraded) | CAGR revised from 45% to 54–56% — prior estimate was already aggressive; the upgrade signals management sees incremental demand beyond its own prior forecast |
| Bear: Geopolitical Taiwan risk remains unquantifiable | Neutral | US-Taiwan trade accord reduces economic friction, but a military scenario is unchanged as a tail risk; Arizona buildout is the primary structural hedge |
| Bear: Overseas fab cost premium compresses margins sustainably | Challenged | Q1 margins of 66.2% demonstrate that overseas ramp dilution is well-contained at this stage; equipment duty-free imports reduce Arizona cost premium; N3 margin crossover in H2 offsets new-node dilution |
| Bear: ASML EUV supply constraints limit capacity expansion | Challenged | Capex moved to "high end" of $52–56B range — TSMC is pulling in equipment ahead of schedule, implying supply chain bottlenecks are easing. No equipment constraint language in results. |
| Bear: Customer concentration in AI hyperscalers creates demand volatility | Neutral | HPC at 61% of revenue is concentrated in a handful of hyperscalers; this creates upside leverage but also risk if any major customer pulls back orders. No evidence of pullback in Q1 or Q2 guide. |
Overall: Thesis strongly reinforced. Every bull thesis point was confirmed in Q1 2026; all bear thesis points were either challenged or remain theoretical. The AI demand acceleration, margin outperformance, and guide raise collectively represent the strongest single-quarter data point in the investment case since initiation.
Preliminary Action: Add on weakness — the post-earnings decline appears fully disconnected from fundamentals and represents one of the cleaner "buy the dip on a sentiment-driven pullback" opportunities in a name with strong visibility. Maintaining Outperform with increased conviction. Full recap with earnings call color to follow.