Skip to content
Sign in
LONGActiveHigh3Y$TSM$418.51Fri 23 Oct

TSMC: own the bottleneck, not the brand

By MenFem Editorial·Semiconductors·9 June 2026·Methodology·
ai-infrastructuresemiconductorsfoundry
Share
TSMC: own the bottleneck, not the brand

The argument, taken apart

Turn it, hover a ring, a moon or a weight
One call, taken apart
Rings are the falsifier’s legs. Moons are the catalysts it named. Weights are its risks, longer for higher severity.
The claim

Own the bottleneck, not the brand: the chokepoint every AI chip depends on, at a discount to all of them.

What would make it wrong · 2 legs
  1. Leg 1 · Not fired
    Leading-edge or CoWoS capacity moves into surplus
    “Limb 1 not fired on revenue; its margin and pricing half is tested on the Oct 15 print.”
    read 9 Oct 2026
  2. Leg 2 · Can't be checked
    The discount closes because its customers de-rate
    “The second leg is uncheckable from filings.”
    read 6 Sept 2026
The window
9 Jun 2026 to 9 Jun 2029
13% gone · 956 days left
The reviews
Reviewed on 4 days · last 9 Oct 2026
3 Aug · 6 Sept · 23 Sept · 9 Oct
Column = the claim, taller for more conviction · rings = falsifier legs, lit when firing · moons = catalysts · weights = risks, longer for higher severity
  • Leg 1, not fired: Leading-edge or CoWoS capacity moves into surplus, read 9 Oct 2026: “Limb 1 not fired on revenue; its margin and pricing half is tested on the Oct 15 print.”
  • Leg 2, can't be checked: The discount closes because its customers de-rate, read 6 Sept 2026: “The second leg is uncheckable from filings.”
  • Quarterly earnings + capex guide
  • CoWoS capacity doubling + N2 volume
  • high severity: Taiwan geopolitical tail
  • medium severity: Capex / depreciation + overseas-fab dilution

Key Points

  • Fabricates and CoWoS-packages every leading-edge AI accelerator — NVIDIA, AMD, Broadcom, Apple alike.
  • ~22x forward earnings: a discount to NVIDIA, AMD and Broadcom, the names that cannot ship without it.
  • Margins are rising into the build-out; the only first-order risk is exogenous (Taiwan), which is why it is cheap.
Price at Call
$427.92
TSMC

The market obsesses over which AI chip brand wins — NVIDIA versus Broadcom versus the hyperscalers' custom silicon — and pays five trillion dollars for the leader. Meanwhile the one company that monetizes all of them trades at a discount to every customer it supplies. TSMC fabricates and packages every leading-edge AI accelerator on earth; it owns the two literal chokepoints, leading-edge wafers and CoWoS packaging; it is raising prices into a sold-out book; and its margins are rising into the build-out. At roughly 22x forward earnings it is cheaper than NVIDIA, AMD and Broadcom — the names that cannot ship without it. This is the rare case where the highest-quality, most-diversified, most-defensible business in a value chain is also the cheapest. BULLISH, conviction HIGH. The discount is not a business flaw; it is the Taiwan geopolitical tail — real, binary, and the single risk that defines the position. The contrarian call of the cluster: own the bottleneck, not the brand.

Research Log

Reviewed 2026-10-09. Tape: 427.92 → 464.80 (+8.6%). Graded against the page's two limbs, not the Taiwan line (which the page excludes). Q3 revenue from the monthly 6-Ks: NT$1,494.25B, about 2% above the top of the guide at the guide's own NT$32. Limb 1 not fired on revenue; its margin and pricing half is tested on the Oct 15 print. Limb 2 not yet checked. KEEP.

Catalyst and risk text brought up to the current FY26 capex guide, $60–64B (raised from $52–56B, as the 3 Aug entry records). The thesis, conviction and falsifier are unchanged.

Reviewed 2026-09-06. Tape: 427.92 → 428.91 (+0.23%), worst weekly close −6.91%, peak-to-trough −13.80%, vs QQQ −1.74pp — the tape says nothing about the thesis. Falsifier NOT FIRED (the second leg is uncheckable from filings). The 5 Sep dossier: July revenue +44.7% YoY, a single US$29.4B capacity appropriation, H1 capex US$26.8B primary. KEEP.

The margin beat is backward-looking, and the company itself guides it lower.

Q2 2026 gross margin came in at 67.7%, above the top of TSMC's own 65.5–67.5% guide. That is the good news, and it is behind us.

Forward, TSMC guides Q3 gross margin down to 65–67%, the CFO has said the N2 ramp dilutes gross margin by roughly 3 to 4 percentage points in the second half, and FY26 capex was raised to $60–64B from $52–56B. (Margin series previously read directly from TSMC's SEC-furnished releases; not independently re-verified in this pass.)

This is a three-year call on owning the bottleneck, and nothing about the bottleneck has changed. But the honest counterweight to a mild drawdown is that the near-term margin path is guided lower by the company, not merely feared by the market. Raised capex against diluting margins is what building the next node looks like from the inside; it is also what a margin peak looks like from the outside. Both readings are available, and this call is only wrong if the bottleneck itself moves.

Since this call

  1. TSMC · Monthly 6-K for Sep 2026 shareholdings: C.C. Wei (Chairman & CEO) 8,152,610 to 8,356,985 shares (+204,375). Pledge changes, asset deals, bonds and share cancellations: none. · sec.gov
  2. TSMC · 6-K furnishes Exhibit 99.1, the press release issued 15 Oct 2026, and Exhibit 99.2, the 2026 Third Quarter Earnings Conference Presentation Material. · sec.gov
  3. TSMC · 13 insiders filed Form 4s for 7 Oct 2026 ESPP purchases at $79.39, totalling 733 shares; no sales reported. · sec.gov
  4. TSMC · September 2026 consolidated net revenue NT$511,857 million, down 0.6% from August and up 54.6% from September 2025 (NT$330,980m); Jan-Sep 2026 revenue NT$3,898,727m, up 41.1% y/y. · sec.gov
  5. TSMC · August 2026 report: two VPs' holdings rose (Tien +18,051 to 19,051 shares; Yuan +6,650 to 9,000); capital appropriations of US$16,035m, US$4,791m and US$8,616m approved; no pledges or bond issuance reported. · sec.gov
  6. TSMC · Before (deep-dive-2026-07-24-q3-guide-verification.md): (no verdict) After (deep-dive-2026-09-05-refresh.md): The Q2 beat was not a peak — July revenue accelerated to +44.7% YoY, the board appropriated a single-tranche US$29.4B of capacity capital, and H1 capex now clears the primary bar at US$26.8B. The thesis is unchanged and better evidenced; the only thing that moved against it is the tape, which is ~2% below the 2026-07-09 price after a beat. · tsmc
  7. TSMC · Before (deep-dive-2026-07-10.md): The AI complex's single foundry + CoWoS chokepoint just printed a hard 20-F — FY2025 rev $121.4B, 59.9% GM, 50.8% op margin, $54.1B net income, $55B net cash — and the filing DE-RISKS the two loudest bear points: top-2 customers are only 36% of revenue (not 40%+), and its first overseas fab (Arizona) already turned a $514M profit. The only first-order risk stays exogenous (Taiwan), and the Jan-2026 US–Taiwan deal quietly converts TSMC's Arizona spend into tariff immunity. Thesis intact; BULLISH / HIGH / 1Y+ — with one honest downgrade to the "cheapest AI name" line (NVIDIA has de-rated to multiple parity). After (deep-dive-2026-07-14.md): (no verdict) · tsmc
  8. TSMC · Before (deep-dive-2026-06-09.md): The chokepoint that captures the foundry + CoWoS-packaging rent for the ENTIRE AI complex (NVIDIA, AMD, Broadcom, Apple all depend on it), trading at ~22× forward — BELOW every customer it supplies — with the only real risk exogenous (Taiwan geopolitics, the source of the discount). The cleanest, highest-quality, lowest-multiple way to own AI infrastructure. BULLISH / HIGH / 1Y+. After (deep-dive-2026-07-10.md): The AI complex's single foundry + CoWoS chokepoint just printed a hard 20-F — FY2025 rev $121.4B, 59.9% GM, 50.8% op margin, $54.1B net income, $55B net cash — and the filing DE-RISKS the two loudest bear points: top-2 customers are only 36% of revenue (not 40%+), and its first overseas fab (Arizona) already turned a $514M profit. The only first-order risk stays exogenous (Taiwan), and the Jan-2026 US–Taiwan deal quietly converts TSMC's Arizona spend into tariff immunity. Thesis intact; BULLISH / HIGH / 1Y+ — with one honest downgrade to the "cheapest AI name" line (NVIDIA has de-rated to multiple parity). · tsmc

Bull Case

The chokepoint that fabricates and packages every AI accelerator brand, raising prices into a sold-out book at rising margins — at ~22x forward, a discount to every customer it supplies.

Bear Case

Ninety percent of value sits on one island that is the decade's most likely great-power flashpoint — a real, binary, un-diversifiable geopolitical tail.

What would prove this wrong

The reason for holding this stops being true if the chokepoint stops being one — leading-edge wafer or CoWoS capacity moving into surplus, and the pricing power that comes from being sold out inverting. It also stops being true if the discount closes downward: the multiple converging to its customers' because they de-rate rather than because TSMC re-rates would mean the mispricing was real and paid nothing. The Taiwan tail is deliberately NOT the falsifier — it would impair the business without ever showing the thesis was wrong about where the value sits.

Catalysts

Quarterly earnings + capex guideEarnings

>30% USD revenue growth guide and the $60-64B capex into CoWoS and N2.

CoWoS capacity doubling + N2 volumeProduct Launch

CoWoS toward 130-150k wafers/mo and N2/GAA ramp — the supply that gates the whole complex.

Risk factors

Taiwan geopolitical tailHigh

A cross-strait conflict or blockade is a permanent impairment no model survives — real, binary, and the source of the discount.

Capex / depreciation + overseas-fab dilutionMedium

The $60-64B capex converts to depreciation into 2027-28; overseas fabs dilute margin if AI utilization softens.

Related Calls

WATCH$AVGOActive

Broadcom: the right custom-silicon thesis at the wrong price

The right custom-silicon thesis at the wrong price — watching the most expensive, lowest-quality way to play it.

The named winner of 'custom silicon eats commodity inference' — AI revenue +143%, FY27 AI guided >$100B.

Trades at a premium to NVIDIA on every earnings multiple while earning a third of the ROE.

ai-infrastructuresemiconductorscompute
Broadcom
Med1YSemiconductorsUpdated 3d ago9 Jun
LONG$000660.KSActive

SK Hynix: the largest HBM rent, valued like a Korean cyclical

The largest slice of the HBM scarcity rent, valued like a Korean cyclical — with a Nasdaq listing as the re-rating catalyst.

~62% HBM share and ~two-thirds of NVIDIA's HBM4 — the largest rent-capture in the cluster.

72% operating margin and ~61% ROE, yet a single-digit forward multiple — rent in the earnings, not the multiple.

ai-infrastructuresemiconductorsmemory
SK Hynix
Med1YSemiconductorsUpdated 3d ago9 Jun
LONG$MUActive

Micron: priced like a cyclical at the top, contracted like a utility

Priced like a peak-cycle memory-cyclical, contracted like a utility — a re-rating bet, not an earnings bet.

Entire CY2026 HBM output sold under multi-year fixed price-and-volume contracts.

Overtook Samsung into the #2 HBM slot; power-efficiency edge keeps it in every NVIDIA generation.

ai-infrastructuresemiconductorsmemory
Micron
Med1YSemiconductorsUpdated 3d ago9 Jun

Conviction

Conviction History

High→High

3 Aug 2026

Reviewed against the Q2 2026 earnings wave (SK Hynix 29 Jul, Samsung 30 Jul, and the late-July hyperscaler prints). Q2 GM beat the top of its own guide; Q3 is guided lower on N2 dilution and capex is raised. A three-year bottleneck call, unaffected by a near-term margin path. Held — no conviction change.

Weekly closes

entry 427.92

Sources

See all Semiconductors calls