TSM · Taiwan Semiconductor Manufacturing Company Ltd. · Stock research

Current Thesis

The Q2 print on 2026-07-16 settled the demand question and opened a new one. Revenue came in at $40.20B, up 33.7% YoY and 12.0% sequentially, at the top of guidance. Gross margin expanded 150bp QoQ to 67.7%, operating margin hit 60.3%, net margin 55.6%, and GAAP EPS of $4.31 cleared consensus by $0.37. Management lifted full-year 2026 revenue growth to "slightly above 40%" from "above 30%" and guided Q3 to $44.6–45.8B another ~12% sequential step. On the demand side there is nothing to argue about.

What the market sold was the funding of it. CFO Wendell Huang raised 2026 capex to $60–64B from $52–56B, roughly 15% more spend, and flagged that the 2nm ramp will dilute gross margin by 3–4 percentage points in H2. Taipei shares fell 7.3%, the ADR cut through the late-June breakout pivot near the published invalidation level and closed 2026-07-17 around $397.75 at a two-month low the sharpest drawdown since 2022. The peer tape confirmed the direction rather than contradicting it: SoftBank fell 11%, Korean semis dragged the group lower, and Navitas and Sandisk sold off on the same capex read.

The narrative leg an investor buys here is unchanged every leading-edge AI accelerator on the planet is fabbed in one company's tools, and that company just raised its own growth guide by ten points. The problem is that the price structure that made the trade tactical no longer exists. This is a broken chart attached to accelerating fundamentals, which historically resolves upward but not immediately. The setup is a pass until a higher low prints.

Bullish and bearish views on Taiwan Semiconductor Manufacturing Company Ltd.

The model's bull view on Taiwan Semiconductor Manufacturing Company Ltd. (TSM), in brief: 2026-07-16 Q2 revenue $40.20B, +33.7% YoY / +12.0% QoQ, top of guide. The bear view: 2026-07-16 capex raised to $60–64B from $52–56B. Both cases follow in full.

Bull Case

• 2026-07-16 Q2 revenue $40.20B, +33.7% YoY / +12.0% QoQ, top of guide. Gross margin 67.7% (+150bp QoQ) against a guide that assumed less. Net profit up roughly 77% YoY to about $22B.
• 2026-07-16 FY2026 revenue growth guide raised to "slightly above 40%" from "above 30%." A ten-point full-year raise mid-year is the single most bullish datapoint in the release and is the opposite of a peaking cycle.
• 2026-07-16 Q3 guide $44.6–45.8B, gross margin 65–67%. The midpoint implies another double-digit sequential step; the margin band absorbs the 2nm dilution and still sits near record levels.
• 2026-07-15 June monthly revenue reported +68% YoY. The interim data series is running hotter than the quarterly average, meaning the exit rate into Q3 is accelerating rather than flattening.
• 2026-07-17 Barclays maintains Overweight, PT to $650; DA Davidson Buy, PT to $500. Sell-side raised targets into a down tape, which is the sequence that usually precedes a base rather than a top.
• 2026-07-16 additional $100B Arizona commitment, taking US investment to $265B across 10 fabs, 2 advanced-packaging plants and an R&D center. Trump amplified it publicly on 2026-07-17. Over a multi-year horizon this is the only credible dilution of the single-island concentration discount embedded in the multiple.
• 2026-07-17 Pat Gelsinger's warning that a China-induced Taiwan power cut would trigger a crisis "worse than the Great Depression." Framed as risk, it is also the clearest third-party statement that no substitute capacity exists at the leading edge.

Bear Case

• 2026-07-16 capex raised to $60–64B from $52–56B. A ~15% step-up in spend against a fixed revenue base compresses free cash flow, and the reaction shows the market is now pricing TSMC on cash generation rather than revenue growth. This is a genuinely new bear vector, not a recycled one.
• 2026-07-16 2nm ramp guided to dilute gross margin 3–4pp in H2 2026. The pricing-power story that carried the stock from April to June now runs into a mix headwind management itself quantified.
• 2026-07-17 ADR at a two-month low near $397.75, worst selloff since 2022. The entire June advance is gone. Price sits below the late-June pivot near the published invalidation level with no higher low yet; a 52-week range of $223.70–$479.00 puts the stock roughly 17% off the high.
• 2026-07-17 TD Cowen maintains Hold, PT $440. A held Hold rating during a PT-raise wave marks where the sell-side thinks fair value actually sits, and $440 is above spot but well under the bull targets.
• 2026-07-16/17 sector-wide rout: SoftBank -11%, Korean semis lower, Navitas and Sandisk sold off on TSMC's capex read. Cluster confirmation is running against the long side right now.
• 2026-07-18 the week's most-discussed retail tickers were NFLX, ASTS, SK Hynix, MU and SpaceX; TSM absent. Narrative attention inside the AI-silicon complex is rotating toward memory, which drains the marginal buyer.
• 2026-07-09 tech-stock volatility at a 23-year high, Nasdaq swings outpacing the S&P. Position sizing has to assume wider daily ranges than the last two quarters implied.
• Structural Taiwan Strait concentration. Unhedgeable, non-diversifiable, and permanently caps the multiple regardless of execution.

Setup & Price Structure

The chart flipped from continuation to repair in two sessions. Price broke the late-June pivot near the published invalidation level on the earnings reaction and closed 2026-07-17 near $397.75, a two-month low, below both the 20- and 50-day averages with the 52-week mean around $330.76 still far beneath. That gap between spot and the long-run average is the reason the drawdown has room to extend there is a large unfilled distance to any prior consolidation.

The relevant structure now is the May shelf in the high $380s to low $400s. Holding it turns this into a capex-anxiety flush inside an intact uptrend; losing it on a weekly basis converts a pullback into a trend reversal and hands the tape to the free-cash-flow bears. Repair requires two things in sequence: a higher low above the May shelf, then a weekly close back above the published invalidation level to reclaim the broken pivot. Buying before the first of those is buying a falling knife in a group where every peer is also selling off. Adding on the way down would be compounding a structural break with size.

Catalyst Calendar (next 30 days)

• ~2026-08-10 (est.) July monthly revenue release. The highest-signal interim datapoint; June ran +68% YoY, so anything under ~+50% would be the first crack in the demand series.
• ~2026-07-29/30 (est.) FOMC decision. Rate path drives the discount rate applied to a name whose bull case is now explicitly a long-duration capex story.
• ~2026-08-26 (est.) Nvidia's quarterly print. The cleanest read-through on whether hyperscaler AI capex digestion is real or a narrative artifact; a strong NVDA guide directly refutes the free-cash-flow bear on TSM.
• ~2026-09-10 (est.) August monthly revenue release.
• ~2026-10-15 (est.) next quarterly report. Outside the window, but it is the date on which the 3–4pp H2 margin dilution gets marked to actuals.

Elapsed catalysts

• Ongoing Commerce Department and Arizona follow-through on the 2026-07-16 $100B expansion; individual fab milestone announcements have moved the stock before. _(passed 3d ago)_

What Would Change Our Mind

• A weekly close below $380 breaks the May consolidation shelf and reclassifies this from pullback to reversal. That level, not the earnings reaction itself, is the line that matters.
• A Q3 gross-margin print or forward guide under 65% would mean the 2nm dilution is running worse than the 3–4pp management quantified on 2026-07-16, validating the margin bear.
• Monthly revenue decelerating below roughly +40% YoY in the August or September releases would break the accelerating interim series that underwrites the full-year raise.
• Any confirmed top-5 customer moving leading-edge volume to Intel 18A or Samsung 2nm. The foundry-monopoly premise is the entire thesis; this is a hard break rather than a data point to monitor.
• On the upside a higher low above the May shelf followed by a weekly close above the published invalidation level restores the structure and makes the capex selloff a buyable dislocation rather than a top.

Correlation Notes

TSM is the denominator of the AI-silicon complex, so it correlates tightly with NVDA, AVGO and AMD on the customer side and with ASML, AMAT and Lam on the equipment side the 2026-07-16 capex raise is directly bullish for the latter group even as it pressured TSM itself, and that divergence is the cleanest way to hold the theme without holding the Taiwan tail. Memory names (MU, SK Hynix, Sandisk) trade on the same AI-infrastructure impulse but currently absorb more of the retail narrative flow, which showed up on 2026-07-18. SoftBank has become a high-beta proxy for AI-momentum positioning generally; its 11% single-day drop on 2026-07-17 is a sentiment thermometer for the whole complex.

Idiosyncratic drivers that do not travel with the sector: USD/TWD, Taiwan Strait headlines, and US–Taiwan trade and tariff policy. Because the Taiwan concentration risk is unhedgeable, pairing this exposure with a second island-based foundry compounds a single tail rather than diversifying it; US-fab and equipment exposure is the structurally sounder pairing.