Competition

Competition

Competitive Bottom Line

TSMC's moat is real, widening, and easy to test against the P&L. The company earns 50.8% operating margin and 44% ROIC at a node-leadership level that exactly one other foundry on earth can attempt — Samsung — and one IDM is paying ~US$13B per year to attempt — Intel Foundry. SMIC, UMC, and GlobalFoundries are economic substitutes only for TSMC's mature-node revenue, not for the 74% of wafer revenue that comes from advanced nodes where price gets set, not taken. The single rival that matters for the leading-edge monopoly is Samsung Foundry, and the most-watched single technical variable is whether Samsung holds competitive yield at 2nm. As of Q1 2026 it does not — Samsung Foundry revenue declined 3.9% in FY2025, and 2nm Gen 1 ramps are still mobile-first rather than HPC. Until that changes, the rest of the peer set is competing for the part of TSMC's business that already cycles like a commodity foundry.

The Right Peer Set

The pure-play and emerging-foundry peer set distills to five comparators chosen to span every economically distinct competitor archetype: Samsung (IDM-owned leading-edge rival), SMIC (China-domestic capped foundry), UMC (Taiwan mature/specialty pure-play), GlobalFoundries (US specialty pure-play), and Intel (IDM pivoting to foundry). Together they cover ~21% of global foundry revenue against TSMC's ~70%, plus Intel's nascent external foundry book. NVIDIA, AMD, Apple, and Broadcom are TSMC customers, not foundry competitors — including them confuses end-demand with supply-side rivalry; AMD is staged separately for end-demand context only.

Where The Company Wins

The four advantages below show up directly in the P&L and capex schedule — they are evidence-based, not narrative.

1. Process leadership is uncontested below 5nm

TSMC is the only foundry in volume production at 3nm (24% of FY2025 wafer revenue) and the only foundry with 2nm in mass production today; 1.6nm-class A16 is in risk production for 2026 with backside power. Samsung Foundry's 2nm Gen 1 commentary in its Q1 2026 deck flags "initiating 2nm Gen 2 mobile product ramp-up" and "1.4nm on track; pursuing large-scale 2nm customer expansion" — language that confirms Samsung's leading-edge revenue is still mobile-first and yield-constrained, not winning HPC sockets. Intel Foundry's 18A is its first credible leading-edge node and is volume-ramping in 2026, but Intel reports negative foundry operating margin while doing it. The economic consequence is in the margin column: TSMC's 59.9% FY2025 gross margin is roughly 30 percentage points above UMC, 35 above GFS, and 39 above SMIC — the spread is the leading-edge premium.

2. Capex scale that nobody else can match

TSMC's FY2025 capex of NT$1.27T (US$40B) is larger than the combined annual revenue of UMC, GFS, and SMIC, and management has guided US$52–56B for FY2026 — top of historical absolute capex. Samsung Electronics' group capex is comparable in magnitude but is spread across memory, displays, mobile, and foundry; Samsung Foundry's share of that capex is the relevant comparison, and on TrendForce and industry-tracker estimates it is well below TSMC's foundry capex. Only Intel comes within an order of magnitude — and Intel is losing money on the spend. Capex productivity, not capex level, is the discriminator: TSMC's revenue / net PP&E ran 1.2x in Q1 2026; SMIC and Intel both run near 0.5x. The dollar that builds an advanced-node fab earns roughly twice as much revenue at TSMC as at any rival.

3. Customer lock-in at the leading edge

Apple, Nvidia, AMD, Broadcom, Qualcomm, and Marvell — TSMC's largest customers — cannot economically multi-source at 3nm or 2nm. A leading-edge tape-out costs hundreds of millions of dollars and 18–24 months of design effort; once committed to TSMC's PDK, switching to Samsung or Intel inside a product cycle is impractical. Samsung's own deck confirms its current foundry pipeline is dominated by mobile Exynos and "AI/HPC design wins" expressed as a forward goal, not a current revenue line. The 75% of TSMC FY2025 revenue from North American customers — and HPC's rise to 61% of Q1 2026 revenue — is captured business that has nowhere else to go for current designs.

4. Mature-node defence via specialty mix and pricing discipline

Even on nodes where UMC and GFS theoretically compete (28nm+), TSMC has held mature-node share via specialty platforms (BCD power, RF-SOI, automotive eNVM, image sensor) and disciplined wafer pricing. The evidence is on the income statement: UMC FY2025 revenue grew 2.3%, GFS revenue grew 0.6%, and both saw operating margins decline year-on-year as China-domestic capacity (SMIC + Hua Hong + Nexchip) pressured mature-node ASPs. TSMC's mature-node business is not growing fast, but it is generating positive margin where specialty competitors are flat-to-declining.

Where Competitors Are Better

The honest read is that competitors do beat TSMC on a handful of specific dimensions — none yet large enough to materially compress economics, but worth naming.

1. Geographic diversification (GlobalFoundries)

GlobalFoundries operates four scaled fabs across three continents — Malta NY, Burlington VT, Dresden, and Singapore — with multi-site qualification of its specialty platforms. TSMC is still ~80%+ Taiwan-concentrated in production capacity through 2027 despite Arizona, Kumamoto, and Dresden ramps. For customers prioritizing supply-chain resilience above leading-edge performance — automotive, defense, industrial — GFS is structurally better positioned. TSMC's commentary explicitly acknowledges 2–3 percentage points of margin dilution from Arizona ramp through 2028, which is the price of partially closing this gap.

2. China-domestic demand capture (SMIC)

SMIC is the explicit beneficiary of "localisation shift in the industrial chain" that SMIC's own 2025 interim report describes as driving "more wafer foundry demand back to domestic market." This is revenue TSMC structurally cannot capture under US export controls. SMIC's FY2025 revenue of ~US$9.3B is up ~+16% YoY per TrendForce (wafer revenue up +24.6% in H1 2025 per SMIC's own interim report) on subsidized capacity expansion; ~74% of revenue is China-domestic — that is a captive demand stream growing faster than the global foundry market average. The trade-off is real economics: 21% gross margin, 4% ROE, and capex/revenue of 71.5% — building capacity it cannot earn returns on. But the demand pool is captive and TSMC has no access to it.

3. Vertical integration with memory/HBM (Samsung)

Samsung Foundry's structural advantage is co-location with Samsung memory, which gives it a path to bundled foundry + HBM offerings for AI accelerators. Samsung's Q1 2026 deck flags "Ramping mass production of 4nm memory products and LPU for AI/HPC applications" and "Targeting earnings improvement on higher HBM4 B-die supply." TSMC works around this via partnerships (CoWoS for HBM integration) but does not own the memory side. For a customer whose AI accelerator design depends on tightly co-optimized logic + HBM, Samsung has a vertically integrated offer TSMC structurally cannot match.

4. Capital efficiency relative to ambition (UMC)

UMC's capex/revenue ran 20.1% in FY2025 against 33.4% for TSMC — UMC is not trying to lead the leading edge, but it is generating 22% FCF margin against TSMC's ~26% on a fraction of the capital intensity. For investors who weight cash-return durability over reinvestment runway, UMC's lighter capex model produces a higher dividend yield (5.9%) and a payout ratio of 89% versus TSMC's ~28%. The cost is forfeiting the leading-edge premium — but UMC's gross margin (29%) and ROE (11%) are stable, and the business is not bet-the-company on the next node.

Threat Map

The shape of the threat map is the bull case in disguise: the high-severity items either depend on a single execution gap closing (Samsung 2nm yield) or on regulatory action that to date has been net favourable to TSMC (Sec 232 advantaging US-built TSMC capacity, Jan 26 US–TW deal). The structural risks are real and concentrated; none are imminent in the central case.

Moat Watchpoints

Five measurable signals that tell an investor whether the competitive position is widening, holding, or fragmenting. All are quarterly-observable in filings or public disclosure.