TSM | Earnings Review

Taiwan Semiconductor Manufacturing Company | 2026Q2 reported July 16, 2026 | Analysis date: July 19, 2026 | Daloopa company_id 911
EPS Beat
+12.8%
$4.31 ADR vs $3.82 Street — largest surprise in 12Q; +77.4% YoY (NT$27.25)
Revenue
$40.2B
Top end of $39.0-40.2B guide; +0.9% vs Street; +33.7% YoY USD / +36.0% NT$
Gross Margin
67.7%
Record; above 65-67% guide; +910 bps YoY, +150 bps QoQ
FY2026 Guide
Above 40%
Raised from >30% USD; Q3 +12% QoQ / +37% YoY; capex to $60-64B
TSM delivered a clean, high-quality beat and raised nearly everything that matters. Revenue of $40.2B landed at the top of the $39.0-40.2B guide (+33.7% YoY USD, +36.0% NT$), gross margin hit a record 67.7% (above the 65-67% guide, +910 bps YoY), and diluted EPS of NT$27.25 (NT$706.6B net income) grew +77.4% YoY — the +12.8% ADR EPS surprise was the largest in 12 quarters. Management lifted the FY2026 revenue-growth guide from "above 30%" to "slightly above 40%" (USD), raised 2026 capex to $60-64B, hiked the dividend +33% to NT$24, and announced +$100B of incremental Arizona investment. The single caveat that worsened: 2nm ramp dilution to gross margin widened to 3-4pp in 2H. Read: decelerating on the surface (tough comp + TWD strength), accelerating underneath (record margins, +77% EPS, re-accelerating NT$ revenue).
What is new this quarter
ThemeWhat changed
FY2026 revenue guideRaised from "above 30%" to "slightly above 40%" (USD) — third upward revision in six months.
CapexLifted to $60-64B (from ~$56B high end); next 3 yrs "even more significantly higher." Driven by demand pull + tool inflation.
MixHPC rose to 66% of revenue (+20% QoQ); advanced nodes (7nm and below) at 77% of wafer revenue; 3nm a record 30%; 2nm entered at 3%.
ToneMaterially more confident QoQ — AI conviction "very high," demand "extremely robust." Q1's Middle East / materials-supply risk retired.
Caveat (worse)2nm ramp GM dilution widened to 3-4pp in 2H (from "2-3% full-year"); overseas-fab dilution 2-3% early → 3-4% at scale.
ContradictionCompetition is "not a consideration" (Q1) then "number one" (Q2) in capacity planning — one confirmed cross-quarter reversal (see below).
Capital returnDividend hiked +33% to NT$24/sh; committed to further increases in 2027.
Next catalyst~mid-Oct 2026 Q3 print (guide $44.6-45.8B, +37% YoY); Jan-2027 possible AI-CAGR and 3-yr capex-guide upside.
Key Metrics Trends
Nine-quarter trend. Revenue on TSM's reported USD net-revenue line; EPS on native NT$. YoY shown from 25Q1 (first quarter with a same-quarter prior-year base in this window).
Metric 24Q224Q324Q425Q125Q225Q325Q426Q126Q2
Revenue (USD) $20.8B $23.5B $26.9B $25.5B $30.1B $33.1B $33.7B $35.9B $40.2B
Revenue YoY % (USD) ---+35.3%+44.4%+40.8%+25.5%+40.6%+33.7%
Gross margin 53.2% 57.8% 59.0% 58.8% 58.6% 59.5% 62.3% 66.2% 67.7%
Gross margin YoY (bps) ---+570+540+170+330+740+910
Operating margin 42.5% 47.5% 49.0% 48.5% 49.6% 50.6% 54.0% 58.1% 60.3%
Diluted EPS (NT$) $9.56 $12.54 $14.45 $13.94 $15.36 $17.44 $19.50 $22.08 $27.25
Diluted EPS YoY % ---+60.2%+60.7%+39.1%+34.9%+58.4%+77.4%
HPC revenue mix 52% 51% 53% 59% 60% 57% 55% 61% 66%
Record gross margin 67.7% (+910 bps YoY) and +77.4% EPS growth on +33.7% USD revenue — a ~44-point EPS-over-revenue spread from margin expansion, utilization and advanced-node pricing. Data sourced from Daloopa.
Beat / Miss — Last 8 Quarters (EPS, ADR/USD)
Quarter Est EPS Actual EPS Surprise Result
2026Q2 ← this quarter$3.82$4.31+12.8%BEAT (largest in 12Q)
2026Q1$3.31$3.49+5.4%BEAT
2025Q4$2.90$3.09+6.6%BEAT
2025Q3$2.63$2.92+11.0%BEAT
2025Q2$2.38$2.50+5.0%BEAT
2025Q1$2.07$2.14+3.4%BEAT
2024Q4$2.20$2.19-0.5%MISS (razor-thin)
2024Q3$1.79$1.95+8.9%BEAT
Last 4 Quarters
4 / 4 · 100%
Avg EPS surprise ~+9.0%
Last 12 Quarters
11 / 12 · 91.7%
Avg ~+6.3%; only miss was 2024Q4 (-0.5%)
Magnitude Trend
Improving
Recent 4Q ~+9.0% vs prior 4Q ~+4.2%
Consistent beater with improving magnitude — beats are getting bigger, not just more frequent, as the AI/HPC mix lifts revenue and margin above conservatively-set guidance. Consensus/surprise from FMP (ADR/USD); actuals from Daloopa.
Guidance Deep Dive
Metric Prior (Apr/Jan) New Guide (Jul) vs Consensus Signal
Q3 2026 revenue (USD)n/a (guides 1Q out)$44.6B-$45.8B (mid $45.2B)~+2.7% above+12% QoQ / +37% YoY
Q3 2026 gross marginn/a65-67% (mid 66%)~in line-170 bps QoQ on 2nm ramp
FY2026 revenue growth (USD)"above 30%" (~31%)"slightly above 40%" (~41%)~in line (~+41%)Third raise in 6 months
FY2026 capex (USD)high end of $52-56B$60B-$64B (mid $62B)n/a+$6B; demand + tool inflation
2H26 GM dilution — 2nm2-3% full-year3-4pp in 2Hn/aOnly caveat that worsened
Dividend / share (NT$)18 (2025 actual)24 (+33% YoY)n/aConcrete capital-return step-up
Guidance-change waterfall
FY2026 revenue growth (USD, %) 0 ~25% Jan 31% Apr 41% Jul 41% Cons.
FY2026 capex (USD $B) 0 $54B Jan mid ~$56B Apr high $62B Jul mid
Three consecutive upward revisions in six months on both revenue growth and capex. Guidance implies revenue RE-ACCELERATION into 2H (implied ~+50% Q4) while GM steps down sequentially on widened 2nm dilution. Sourced from FY2026Q2 / FY2026Q1 transcripts; revenue-guide range cited to Daloopa.
Historical Performance & Inflection Points
20% 40% 60% 80% Q3'24Q4'24Q1'25Q2'25Q3'25Q4'25Q1'26Q2'26 Rev peak +44.4% Rev trough +25.5% (hard comp) EPS peak +77.4% EPS trough +34.9% Revenue YoY % EPS YoY %
Revenue YoY peaks at +44.4% (Q2'25), bottoms at +25.5% (Q4'25) on a hard prior-year comp, then re-accelerates to +40.6% (Q1'26) and settles at +33.7% (Q2'26). The single most important read: the Q4'25 deceleration was a base effect, not a demand crack — growth snapped right back the next quarter. EPS YoY plateaus in the ~54-61% band through Q2'25, dips to +34.9% (Q4'25), then surges to a series-high +77.4% (Q2'26). EPS has outrun revenue in every one of the last eight quarters and the gap is widening — margin expansion (gross margin high-50s% to 67.7%) plus utilization and advanced-node pricing compounding on volume. Net: decelerating on the surface, accelerating underneath. Data sourced from Daloopa (company_id 911).
Key Catalysts
Catalyst Timing Consensus / Watch Implication
Q3 2026 print (next report)~mid-Oct 2026Guide $44.6-45.8B (+37% YoY); Street revised up post-Q2Watch for another beat + FY raise
2nm (N2) steep volume ramp2H26 → 200k wpm 2027Biggest volume driver AND main margin drag (-3-4pp 2H26)3% of wafer revenue in Q2; Apple/Nvidia/AMD lead
AI 5-yr CAGR revisionJan 2027 annual guidePrior 'mid-to-high 50%'; Wei: 'stronger and stronger'Formal upward revision — under-modeled upside
Capex / possible 3-yr guideJan 20272026 $60-64B; next 3 yrs 'even more significantly higher'Reinstating a multi-year guide = strong demand signal
Arizona +$100B & tariff frameworkMulti-year~4 more fabs; overseas dilution 2-3% → 3-4% at scaleJan-2026 framework exempts AZ from Section 232 tariff
Advanced packaging (CoWoS) capacityTight nowPackaging is 'limiting my customers' growth' todayReal gating constraint on 2027 AI revenue
Agentic-AI / CPU resurgenceEmerging, multi-yearx86/Arm/RISC-V 'almost all TSMC customers'New TAM not fully in Street models
Mature-node / consumer softnessOngoingConsumer / price-sensitive 'challenged'; smartphone -4% QoQClearest near-term downside watch
The bull case is management-led and not yet fully Street-believed (agentic-AI CPU, AI-CAGR and capex-guide upside). Downside is bounded to self-inflicted 2nm + overseas-fab margin dilution; CoWoS packaging is the real gating constraint. Sourced from FY2026Q2 transcript + web search.
Street Q&A
Q&A Pairs
24
12 distinct analysts
Well Answered
12
Substantive on every qualitative topic
Deflected
12
Every forward number withheld (evasive-positive)
Analyst Question Management response Assessment
Charlie Chan (MS)Foundry competition — Samsung, Intel, ASML EUV; losing slots?"No shortcuts" — tech, manufacturing, trust are the moat.Well answered
Gokul Hariharan (JPM)Foundry less profitable than memory — pricing philosophy?Trustable partner; jealous of memory 86% GM; "~68% I'd be happy."Well answered
Felix Pan (KGI)Where is the ~$10B capex-guide upside coming from?Two reasons: demand continues to increase; tool inflation.Well answered
Sunny Lin (UBS)3-year capex outlook like the 2021 super-cycle?"We do not have a number" — but 'even more significantly higher.'Deflected (positive)
Charlie Chan (MS)Update to 5-yr AI-semi revenue CAGR?"Not a number, but stronger and stronger... stronger than before."Deflected (positive)
Robert Sanders (DB)Is unconstrained 3nm-and-below demand 30-50% above supply?"We don't have a number to share... the gap is very big."Deflected (positive)
Jim Fontanelli (Arete)Customer-concentration risk — top-5 exposure at record high?"That's not our concern." Dismissed the premise.Deflected (red flag)
Management was substantive on every qualitative topic (moat vs Samsung/Intel, capacity philosophy, packaging strategy, ~68% GM pricing aspiration) but systematically declined ALL forward quantification — 3-yr capex, 5-yr AI CAGR, multi-year growth, 3nm demand/supply gap, Arizona timeline, packaging capex split. Notably the dodges were evasive-POSITIVE: every declined number came with 'bigger/stronger than before' — a bullish tell from a high-guidance-accuracy team (the classic "only management is contrarian" set-up). The one genuinely unhelpful dodge is the customer-concentration dismissal, the standing risk to flag.
Contradictions
Contradiction · Confirmed · Management-quality flag
Competition: "not a consideration" (Q1) vs. "number one" (Q2)
Statement A — FY2026Q1 call (2026-04-16), C.C. Wei to Sunny Lin (UBS)
"we prepare the capacity to meet customers' demand, not because of our competitor or not because of other considerations. The most important one is our customers' demand."
Statement B — FY2026Q2 call (2026-07-16), C.C. Wei to Gokul Hariharan (JPMorgan)
[Asked directly whether capacity expansion also considers competitive pressure] "Definitely, every time when we think about the business, we consider the competition. That's number one."
Why incompatible: Both answers respond to the same question — what inputs drive TSMC's capacity/CapEx build. A factor cannot be both "not a consideration" and "number one." Most likely message drift as competitor threats (Samsung, Intel, ASML EUV) escalated Apr→Jul, but a genuine inconsistency to reconcile with IR — the exact pattern investing-principles flags as a poor-management signal.
Apparent inconsistency · Period-mismatched · Verify
2nm GM dilution: "2-3% full-year" (Q1) vs. "3-4pp in 2H" (Q2)
Statement A — FY2026Q1 call, Wendell Huang (CFO)
"the initial ramp-up of our 2-nanometer technology will start to dilute our gross margin in the second half of this year, and we expect between 2% and 3% dilution for the full year of 2026."
Statement B — FY2026Q2 call, Wendell Huang (CFO)
"we expect the steep ramp-up of our 2 nm to dilute our gross margin by about 3 percentage points-4 percentage points in the second half of the year."
Why flagged (and why it is not a clean contradiction): The figures are quoted against different measurement periods — Q1 gave a full-year 2026 number (2-3%), Q2 gave a 2H-only number (3-4pp). Since the ramp is concentrated in 2H, a full-year 2-3% is roughly consistent with 3-4pp in 2H, so it does not numerically conflict. The issue is disclosure hygiene: switching the reference period obscures an apples-to-apples read. Worth a direct IR question; not a hard contradiction.
Guidance raises (FY revenue >30% → slightly above 40%; capex $52-56B → $60-64B) are management-flagged revisions, not contradictions. Sourced from FY2026Q2 and FY2026Q1 transcripts.
Indirect Read-Throughs
Company / Theme Read-through Why it matters
ASML / semicap (AMAT, KLAC, LRCX)PositiveCapex to $60-64B (70-80% advanced process) + tool inflation = more litho/WFE orders; EUV expansion validates multi-year demand
HBM / memory (Micron, SK Hynix, Samsung)PositiveMemory a large, rising share of AI system cost; Wei cites memory 86% GM — pricing power confirmed
CPU designers — x86/Arm/RISC-V (AMD, Arm)PositiveAgentic AI drives CPU resurgence; TSMC is ISA-agnostic and captures all of it as customers
Hyperscalers / CSPsPositiveCSPs provide 'very strong signal and positive outlook'; corroborates sustained cloud AI capex
OSATs (ASE) / advanced packagingPositiveCoWoS capacity-constrained; TSMC welcomes third-party packaging to offload back-end
Intel / Samsung (foundry)Neutral-to-cautiousFramed as non-threatening or helpful; Intel EMIB-T a 'relief valve,' not a share-taker; newcomers dismissed
PMIC / sensor suppliers (analog)PositiveAI data-center power drives PMIC and sensor shortage; rest of mature-node remains soft
Consumer / commodity semisCautiousConsumer / price-sensitive 'challenged' on rising component prices + macro; smartphone -4% QoQ, 22% of mix
Data-center power / grid / coolingWatchTSMC checking AI data-center build progress; power buildout gates AI compute deployment
Net: bullish for the AI / leading-edge / data-center complex (ASML, HBM/memory, CPUs across all ISAs, PMIC/sensor, OSATs, hyperscaler capex), cautious on consumer / commodity semis. Competitors (Samsung, Intel, newcomers) are framed as non-threatening or actually helpful to TSMC's constrained front-end wafer business.
Reported financials and the captured revenue-guidance range sourced from Daloopa (company_id 911). Consensus EPS / surprise history from FMP (ADR/USD basis). Guidance, tone, Q&A, contradictions and read-throughs from the FY2026Q2 (2026-07-16) and FY2026Q1 (2026-04-16) earnings-call transcripts. Data sourced from Daloopa.