Web Watch
Figures converted from TWD at historical FX rates — see data/company.json.fx_rates. Ratios, margins, and multiples are unitless and unchanged.
Web Watch in One Page
The TSMC investment debate distils to one near-term and one structural question: whether the Q1 FY26 gross margin of 66.2% survives the first hyperscaler AI capex digestion, and whether Samsung Foundry — or Intel 18A — can ever land a named HPC customer at 2nm. Those two outcomes anchor the verdict ("Lean Long, Wait For Confirmation"), and three companion signals decide whether the case ages well or breaks abruptly: aggregate hyperscaler AI capex commentary, the high-frequency TSMC monthly revenue tape, and the geopolitical/regulatory perimeter around Taiwan, US export controls, and Arizona conditionality. The five active watches below mirror that hierarchy: one hard-dated earnings test, one moat-breaker probe, one demand-durability probe, one tape signal, and one binary-tail/regulatory monitor. Each is tied directly to a thesis driver or stated failure mode in the report — nothing is here because investors "usually" care.
Active Monitors
| Rank | Watch item | Cadence | Why it matters | What would be detected |
|---|---|---|---|---|
| 1 | TSMC Q2 FY2026 gross margin print and dilution-walk commentary | Daily | The single highest-impact hard-dated catalyst — tests whether 66.2% Q1 GM is the new regime or a cycle peak, and whether the management-disclosed N2 + overseas-fab dilution stack lands inside the 2–3pp guide | Q2 earnings results vs the 65.5–67.5% GM guide, HPC mix %, 2nm wafer-revenue contribution, Arizona/N2 dilution commentary, and any walk-down in FY26 USD growth or Q3/Q4 GM trajectory |
| 2 | Samsung 2nm and Intel 18A leading-edge customer wins | Daily | Driver 1 of the long-term thesis (leading-edge yield monopoly) is load-bearing — a named HPC tape-out (Nvidia, AMD, Broadcom, Qualcomm) at Samsung 2nm or Intel 18A flips the wide-moat read to narrow inside one earnings cycle | Announcements, leaks, or earnings-call commentary committing a major HPC accelerator design to volume at Samsung Foundry 2nm or Intel Foundry 18A; Samsung DS Division foundry segment revenue inflecting back to double-digit growth |
| 3 | Hyperscaler AI capex digestion signals (Microsoft, Google, Meta, Amazon, Oracle) | Daily | HPC is 61% of TSMC Q1 26 revenue and FY26 capex of $52–56B is sized for AI continuation — two consecutive quarters of flat-to-down aggregate hyperscaler AI capex with TSMC capex still rising fires the bear's stated primary trigger | Hyperscaler 10-Q capex disclosures, "digestion" or "pull-forward" language in calls, Nvidia / AMD / Broadcom AI accelerator demand commentary, CoWoS allocation cuts, TPU/ASIC build cuts (Google reportedly cut FY26 TPU target from 4M to 3M units) |
| 4 | TSMC monthly revenue releases | Daily | Highest-frequency pre-print signal for the FY26 "above 30%" USD-growth guide — April 2026 was the first sequential decline (−1.1% MoM) since October 2025, and two more soft months would pre-print Q2 narratively bearish | The 10th-business-day monthly press release on YoY %, MoM %, and YTD cumulative growth; any TSMC commentary attached to the print |
| 5 | Taiwan-strait, US export-control, and Arizona conditionality risk | Every 12 hours | Binary tail (Taiwan strait) and structural risk premium (BIS Sophgo settlement, Section 232 follow-on, possible US equity-stake or IP-licensing conditionality on Arizona) sit outside the moat framework and re-rate the multiple on a single news cycle | Trump–Xi communiqués naming Taiwan; PLA exercise pattern changes; BIS Sophgo / Huawei settlement announcement and dollar figure; Section 232 implementing rules; any forced US equity stake, tech-transfer mandate, or CHIPS Award condition affecting TSMC Arizona |
Why These Five
The five watches map one-for-one to the report's open questions. Monitor 1 is the verdict's own near-term decision marker — Bull's stated catalyst (Q2 GM ≥64% with HPC mix ≥60%) and Bear's stated walk-down trigger both resolve at the same mid-July print. Monitor 2 is the only signal that can flip the long-term thesis from "wide-moat compounder" to "narrow moat at peak cycle" inside one earnings cycle, per Long-Term Thesis Driver 1 and Failure Mode 1. Monitor 3 tracks the variant view's most testable disagreement — whether HPC at 61% of revenue is resilience or the next cycle waiting to happen — and feeds the Bear's primary trigger directly. Monitor 4 is the highest-frequency soft-dated tape, deliberately placed to catch the May, June, and July prints that set the read into the Q2 call. Monitor 5 consolidates the binary geopolitical tail and the regulatory perimeter that the consensus multiple does not visibly discount; it is the one watch whose move would re-rate the stock independent of any operating result.
The four watch items the report flagged but that did not make the cut are deliberate omissions. The Wei-Jen Lo IP lawsuit and Tokyo Electron indictment are slow-burn court calendars whose outcomes shape narrative more than P&L. Apple-specific multi-sourcing rumours fold into Monitor 2 (any flagship customer named at Samsung/Intel triggers detection). Quarterly dividend cadence is mechanical and known. Mature-node price pressure from subsidised China capacity affects ~12% of the profit pool with no moat anyway. Five well-aimed watches beat ten that dilute attention.