Finance · Daily Deep-Dive · 2026-07-17

TSMC's Profit Just Hit a Record +77%. The Stock Fell Anyway.

On July 16, TSMC reported second-quarter net profit up 77.4% year over year to a record high, beating estimates, and raised full-year revenue growth guidance from "more than 30%" to "slightly above 40%." On paper, a flawless quarter. Yet TSMC's US-listed shares fell roughly 3% intraday. What actually moved the market wasn't the demand number — it was the exact, percentage-point-precise margin erosion schedule the company handed investors on the same call.
+77.4%
Q2 net profit y/y · NT$706.6B, a record, beat NT$632.6B forecast
67.7%→65-67%
Q2 gross margin → Q3 guide · N2 (2nm) ramp cost drags ~1.7pp
$64B
Raised 2026 capex · up from a prior ~$52-56B
~40%
Raised 2026 revenue growth guide · up from "more than 30%"
-3% / +1%
US ADR intraday drop · Taiwan-listed shares closed up same day
3-4pp
Management's own guide: 2H26 margin drag from the N2 ramp

What happened: a "beat, then guided down" quarter

On July 16, TSMC reported second-quarter net profit of NT$706.6 billion, up 77.4% year over year — a record high, beating the NT$632.6 billion consensus. Management also raised full-year 2026 revenue growth guidance from "more than 30%" to "slightly above 40%," and announced an additional $100 billion investment in Arizona. On the traditional playbook, this is a report that should have sent the stock higher.

Instead the market's reaction split down the middle: TSMC's US-listed ADR fell roughly 3% intraday, while Taiwan-listed shares actually closed up more than 1% the same session. The same day, all three major US indices closed lower, memory-chip stocks (SK Hynix, SanDisk) — already in a bear market — fell more than 7% further, and semiconductor/AI-hardware names broadly weighed on the tape.

Why: this is a margin erosion schedule the company chose to disclose

This wasn't a surprise cost overrun — it was a structural trade-off management disclosed proactively and precisely on the same call. TSMC guided Q3 gross margin to 65-67%, down about 1.7 percentage points sequentially from Q2's 67.7%, with CFO Wendell Huang explicitly attributing the decline to N2 (2nm) ramp costs. Looking further out, management flagged 2nm ramp-up will dilute gross margin by 3-4 percentage points in the second half of the year.

A second, separate dilution source is the overseas fab build-out. Huang said gross margin dilution from overseas fab ramp-up is expected to run 2-3 percentage points in early stages, widening to 3-4 percentage points in later stages as the scale of overseas expansion grows — not a one-off cost, but a structural drag that persists and expands over the next several years. On the node mix, 2nm made up just 3% of Q2 wafer revenue (its first materially meaningful quarter), 3nm was 30%, and 5nm was 33%; advanced nodes together were 77% of total wafer revenue.

CEO C.C. Wei's tone on the call stayed optimistic, describing the moment as "witnessing a kind of a new industry called AI industry," expecting it to expand into automotive and humanoid robotics, and emphasizing that 3nm lines are running above 100% utilization — with the real bottleneck being CoWoS advanced-packaging capacity, not wafer fabrication itself.

Second-order effects: who gains, who loses

The strongest bull and bear cases right now

Bull: demand is real and was arguably under-modeled — 3nm lines are running above 100% utilization, and the genuine constraint is CoWoS packaging, not wafer fabrication itself; raising the revenue growth guide from "more than 30%" to "slightly above 40%" in one quarter is a rare upward revision for a company this size; the margin dilution is a disclosed, scheduled capacity trade-off, not an unexplained cost blowout.

Bear: the dilution stacks from two separate sources — N2 ramp and overseas fab build-out — and the CFO explicitly used the word "widen" to describe where the overseas piece is headed over the next several years. If AI capex growth merely slows from extraordinary to just strong, rather than reversing, TSMC's already-locked-in high-capex base converts directly into persistently compressed margin the moment revenue growth can't keep pace.

What the narrative is missing

Most coverage flattens this into "beat estimates but stock fell, investors are taking profits, the reaction seems irrational." That framing undersells what actually happened: TSMC handed the market a rare, management-disclosed, percentage-point-precise schedule of its own margin erosion — 1.7 points sequentially in Q3, another 3-4 points from the N2 ramp in the second half, and 2-3 points widening to 3-4 points from overseas fabs over the next several years. That is more informative than a single day's stock chart. The real question worth asking isn't why the stock fell — it's whether ~40% revenue growth can outrun as much as 8 stacking percentage points of margin dilution over the next two to three years. That's a compounding-math question, not a settled bull case.

One takeaway

When you see "beat estimates, guidance also raised, yet the stock still fell," don't dismiss it as noise, and don't treat it as proof the thesis broke either — read the specific guidance numbers management gave on the call, not the day's stock chart. Use the next two to three quarters of actual results to check whether revenue growth is outrunning the margin drag management itself disclosed, before drawing a conclusion.

Sources

  1. CNBC — TSMC invests additional $100 billion in Arizona as second-quarter profit soars 77% (2026-07-16)
  2. Benzinga — TSMC Stock Slides Despite Crushing Q2 Estimates With Fifth Straight Record Profit (2026-07-16)
  3. Invezz — Why is TSMC stock falling despite record Q2 earnings? (2026-07-16)
  4. BigGo Finance — TSMC Lifts 2026 Capex to $64 Billion, But Margin Caution Sends Mixed Signals (2026-07-16)
  5. The Motley Fool — TSM Q2 2026 Earnings Call Transcript (2026-07-16)
This is informational market commentary, not investment advice. Data as of 2026-07-17; figures may vary slightly across data providers. Semiconductor equity markets can be highly volatile, and losses of principal are possible.