Taiwan Semiconductor Manufacturing Company Limited (TSM)

• Cash and quality are exceptional: free-cash-flow yield is 42.22% with 64.2% gross and 50.8% operating margins, backed by net cash and 285.4x interest coverage.F3F4F5F6Q1
• Valuation signals are mixed: trailing P/E is 29.6x and roughly in-line with peers (29.7x median) and above its own 5‑year median band (1.0x vs 0.9x; higher than on 73% of days); the flagged 0.6x forward P/E is anomalous and not relied upon.F7M2V5V2
• Setup into 10/15 earnings is two-way: TSM has beaten EPS in 4/4 quarters and FY25 revenue grew 33% YoY, but options imply ±6.1% (7D) and ±9.4% (30D) moves and price a typical ±3.4% earnings-day swing, consistent with event risk rather than one-way optimism.E1F2O3O4O5C1

Sur quoi le comité s'est appuyéChaque étape a lu un même dossier de preuves construit à partir des documents de la société, des données de marché, des estimations de consensus et des conclusions des agents spécialisés. Chaque affirmation devait citer ses éléments, et chaque citation a été vérifiée dans le dossier.

Les points de désaccord entre les deux thèsesUn point crucial est une question à laquelle les deux thèses répondent différemment, et dont la réponse changerait l'avis. Chaque partie a argumenté chaque point une fois, avec des preuves nouvelles.

1. 1. Does TSM's trailing P/E of 29.56x represent a genuine valuation discount to peers, or does it trade in-line with the peer median of 29.7x?

   Éléments équilibrés

   Peer reads conflict: the sector snapshot shows TSM roughly in-line on trailing P/E (29.56x vs 29.7x median), while the valuation desk cites a 44.0x peer median. Without a reconciled peer set, the evidence does not clearly establish a trailing P/E discount; strong FCF yield supports quality but does not settle the earnings-multiple question.M2V2V3V5

   Ce qui trancherait : Independent peer-group recomputation showing whether the 44.0x peer median cited in V1 reflects the same peer set and time period as the 29.7x median in M2, or clarification of which peer set and lookback period is correct for trailing P/E comparison.

   Lire les deux arguments

   Thèse haussière

   TSM trades at a real trailing P/E discount to peers when using the peer set and methodology that anchor on earnings power: 29.6x for TSM versus a 44.0x peer median, as laid out by the valuation desk’s peer read, not the sector snapshot that uses a different construction.V1V2 This gap persists despite an exceptional 42.22% free-cash-flow yield, an unusually strong cash return profile that is rare at this multiple and argues the market is underpricing durability of cash conversion.V3 The PEG ratio of 0.56 further indicates the current price is not fully capitalizing the growth and margin structure, reinforcing the discount case on an earnings-growth adjusted basis.V4 Importantly, the peer-comparison specialist also flags a meaningful valuation discount relative to semiconductor peers alongside strong profitability, corroborating the discounted positioning narrative beyond a single model.M4 The short side’s best point—an in-line P/E of ~29.6x vs 29.7x—relies on a different peer lens; under the 44x median set, the discount is substantial and strategically relevant.V2

   Concède : Sector-level comp tables show TSM roughly in-line on trailing P/E (29.56x vs 29.7x), so the discount is not visible under that specific peer set and metric mix.

   Thèse baissière

   The crux is trailing P/E, not EV/EBITDA. On that like-for-like basis TSM sits at 29.56x versus a peer median of 29.7x — essentially in-line, not a 33% discount.M2 The long side's own peer-comparison desk concedes the discount language is driven by EV/EBITDA and P/S framing, not earnings multiples: TSM actually carries a P/S premium and only an EV/EBITDA discount, so calling this an earnings-multiple bargain conflates metrics.M4M2 The forward P/E in the dataset (0.6x) is flagged as anomalous and unusable, so there is no credible forward anchor to rescue the 'cheap on earnings' claim beyond trailing figures that already show parity.F7 Layer on the own-history read: today's price-to-trailing-earnings ratio sits above its 5-year median, richer than 73% of trading days, meaning TSM is expensive versus itself even as it is merely average versus peers. The FCF yield is real, but FCF yield and P/E measure different things — citing one does not establish a discount on the other.

   Concède : The 42.22% FCF yield and EV/EBITDA discount are genuine data points the short case does not explain away, even though they do not establish a trailing P/E discount.

2. 2. Will Q3 2026 earnings (due 2026-10-15) beat consensus EPS and raise FY27 guidance materially above the $709.65 consensus while maintaining or expanding margins?

   Penche vers la thèse baissière

   Beat cadence and 33% growth argue for execution, but options imply notable two-way risk and near-term momentum has softened; with FY27 EPS consensus at 709.65, the burden for a material raise is high ahead of multiple catalysts. Today’s evidence tilts against assuming a clear beat-and-raise outcome.E1F2E2O3O4M5C1T2T3
   Lire les deux arguments

   Thèse haussière

   I take the long side: Q3 is more likely to beat and lift FY27 guidance while sustaining elite margins. The beat cadence (4/4) and 33% FY25 revenue growth demonstrate execution momentum into a sector tailwind, supporting upside to consensus EPS and a guidance raise as AI demand persists at scale.E1F2M3 The key short rebuttal is that a +49% YTD move leaves no room for error; however, options markets are not signaling extreme crowding or a breakneck risk premium into the print: crowding is only moderate and the implied one-sigma ranges of ±6.1% (7D) and ±9.4% (30D) are consistent with a typical event setup rather than euphoric pricing, suggesting room for a positive surprise to reprice without needing perfection.O2O3O4 Additionally, sector-rotation data indicate continuing inflows into tech with in-line relative strength, a constructive backdrop for rewarding beats and guidance lifts.M5 With 64.2% gross and 50.8% operating margins, maintaining or expanding profitability alongside AI-led growth is plausible on mix and scale.F4

   Concède : There is no company-filed preannouncement narrowing Q3 ranges or an explicit FY27 guidance framework, and options-implied moves do not predict direction.

   Thèse baissière

   The bull case leans on a flawless beat cadence, but options markets—not narrative—are the best real-time gauge of what's priced in, and they show elevated two-way risk, not confidence in a clean beat-and-raise: a ±9.4% one-sigma 30-day range and a specific ±6.1% 7-day range imply the market itself is hedging a disappointment scenario, not just upside.O3O4 Sector-rotation data corroborate this fragility: while XLK's one-month spread versus SPY is positive, its one-week momentum has already turned negative (-0.52%), suggesting the AI-capex enthusiasm that powered the +49% re-rating is decelerating right before the print.M5 Layer on a FOMC decision two weeks later and NVIDIA's own Q3 report in November, and TSM's guidance becomes just one of several catalysts that could reprice the stock if it merely meets rather than exceeds the aggressive bar implied by $709.65 FY27 consensus EPS.T2T3 After this much re-rating, 'in line' reads as a miss, and a conservative FY27 guide — plausible given elevated expectations — would be the more likely near-term outcome than a material raise.

   Concède : If TSM both beats Q3 consensus and raises FY27 guidance materially while holding margins, the short thesis's near-term disappointment call would be directly falsified.

3. 3. Is the institutional holder erosion (3,434 closed positions, net -3,579 holders) a signal of rational profit-taking after a +49% run, or does it reflect deteriorating fundamentals conviction?

   Éléments équilibrés

   13‑F data show a large net decline in holders with rising days-to-cover trend, but absolute short interest is still low and the strong 12‑month run makes mechanical rebalancing plausible. Without value-weighted flow detail, conviction vs. profit‑taking cannot be decisively distinguished.H1H2P1

   Ce qui trancherait : The next 13-F cycle (Q4 2026 filing, due in mid-2027) showing whether institutional net additions resume, stabilize, or continue declining; and contemporaneous short-interest data trend reversal or continued rise through year-end 2026.

   Lire les deux arguments

   Thèse haussière

   The 13-F holder erosion reads as post-rally rebalancing more than fading fundamental conviction. After a +49.2% YTD and +48.9% 12-month move, many institutions will systematically trim smaller or non-core lines to manage risk and mandate weights without changing their thesis.P1 The strongest bear point is the sheer count of closures (3,434) and a rising short-interest trend; however, absolute short interest remains low at 0.6% of float (3.1 days), which does not signal a broad-based bet against the fundamentals.H1H2 Moreover, insider cluster-buys with no offsetting sales point to internal confidence that contrasts with the mechanical nature of 13-F churn.H3 Options positioning shows only moderate crowding, not capitulation or euphoric piling-in, which is consistent with portfolio rebalancing rather than a thesis break.O2 A continuing cash-return framework, evidenced by an ongoing dividend (0.94% yield), further underpins steady sponsorship even if some holders rebalance.F8 Finally, third-party news flow has skewed positive over the last 30 days, aligning sentiment with resilience rather than deterioration (sentiment only).N1

   Concède : We do not have net dollar flow data by holder cohort, so the 13-F count changes cannot be tied to value-weighted selling or buying pressure.

   Thèse baissière

   The bull case leans on insider cluster-buys as a conviction signal, but the fundamentals desk itself flags that the insider-transaction feed is marked UNTRUSTED_DATA, meaning the very evidence used to offset the 13-F erosion is of unverified reliability.A1 That leaves the 13-F erosion standing largely unrebutted: 3,434 closed positions against 88 added and a net decline of 3,579 holders is a far larger and more systematic signal than a handful of discretionary purchases, and it coincides with a rising short-interest trend rather than a falling one.H1H2 The sector specialist also classifies TSM as a 'challenger' rather than a dominant incumbent in its sector positioning, which cuts against the market's apparent assumption of unassailable moat-driven conviction.M1 Separately, a straightforward DCF built on company filings prices TSM 71% below the current quote, underscoring that even conservative fundamental models see far less value than the price embeds — consistent with large holders de-risking into strength rather than reaffirming conviction.F9 Low short interest reflects limited bearish positioning, not confirmed bullish conviction among the exiting institutional base.

   Concède : It is possible some of the 3,434 closed positions reflect routine index-fund rebalancing or fund closures rather than a deliberate fundamentals-driven exit, which this argument cannot fully distinguish from the data available.

Revues du comitéDes relecteurs indépendants mettent le dossier à l'épreuve avant la rédaction de la note : un pré-mortem de chaque thèse, une revue du positionnement et de la volatilité, et un audit de chaque citation.

Ce qui amènerait le comité à revoir cet avisDéclencheurs de recherche : des évolutions observables, datées lorsqu'un catalyseur le permet, qui rouvriraient l'avis. Ce ne sont pas des instructions de négociation.

• 15 oct. 2026FY27 EPS guidance comes in below 709.65 at or after Q3 2026, signaling undercut to the consensus barE2C1
• 15 oct. 2026Gross margin deterioration from elite levelsF4C1
• 15 oct. 2026Operating margin deterioration from TTM levelsF4C1
• En coursPrice breaks below options’ 30‑day one‑sigma lower bound, indicating negative repricing beyond implied riskO4

Ce sont des déclencheurs de recherche, pas des instructions de négociation.