TSMC Q3 2026 earnings: $46.7 billion in revenue, 2nm at 8% and a 44% growth forecast for the year

TSMC, the contract chipmaker that manufactures AI accelerators for chip designers and cloud companies, reported its third-quarter 2026 results on October 15. Revenue was US$46.72 billion, up 41.2% from a year earlier, and net income reached a record NT$763.06 billion. Its newest 2nm process made up 8% of wafer revenue. High-performance computing, the segment that includes AI accelerators for AI datacenters, was 68% of revenue. TSMC guided fourth-quarter revenue to $52.3–54.3 billion and raised its full-year growth forecast to about 44%.

This story is based on TSMC's October 15 earnings release and quarterly management report, TSMC's transcript of the earnings call and reporting by Taiwan's Central News Agency (Focus Taiwan), checked October 16, 2026.

The short version

• Revenue: NT$1,494.24 billion (US$46.72 billion), up 50.9% year over year in NT dollars and 41.2% in US dollars.
• Profit: net income of NT$763.06 billion, up 68.7%; diluted EPS of NT$29.43 (US$4.60 per ADR).
• Margins: gross 66.4%, operating 58.7%, net 51.1%.
• AI signal: HPC was 68% of revenue, and HPC revenue rose 22% from Q2, per TSMC's management report.
• 2nm: 8% of wafer revenue, up from 3% in Q2.
• Capex: US$13.84 billion in Q3 and US$40.64 billion so far in 2026.
• Outlook: Q4 revenue of $52.3–54.3 billion; 2026 growth of about 44% in US dollars; 2026 capital budget unchanged at $60–64 billion.

TSMC Q3 2026 results at a glance

Last verified: October 16, 2026 (TSMC earnings release and quarterly management report, October 15, 2026)

The gap between NT-dollar and US-dollar growth reflects currency moves over the year, not a different business. TSMC reports in NT dollars and gives guidance in US dollars.

How much of it is AI?

TSMC doesn't report AI revenue as its own line. The closest measure is its HPC platform, which covers AI accelerators, CPUs and other datacenter chips. TSMC's quarterly management report puts HPC at 68% of Q3 revenue, up from 66% in Q2 and 57% a year earlier, with HPC revenue up 22% from the previous quarter. Smartphones were 21% (revenue up 10% from Q2), IoT 4%, automotive 4%, consumer electronics 1% and others 2%.

Chairman and CEO C.C. Wei said AI demand "is being supported by strong underlying demand for AI accelerators" and that "We did not see any sign of slowdown." Asked about TSMC's earlier forecast for AI accelerator revenue growth in the mid-to-high 50s percent a year through 2029, he said it "probably will be higher." Those are management's forecasts, not results.

The comments arrived in the middle of a sharp sell-off in AI stocks and credit, which our analysis of the mid-October sell-off covers. TSMC's US-listed shares closed down 3.4% on October 15, Investing.com reported. Results beat the analyst consensus Investing.com cited ($4.42 per ADR and $45.41 billion in revenue).

What does the 2nm share tell us?

TSMC's release breaks wafer revenue down by process node:

Last verified: October 16, 2026

2nm rose from 3% in Q2, according to TSMC's management report. It's the node behind new datacenter chips such as Microsoft's Maia 300, whose compute chiplets use TSMC's N2P process (our Maia 300 story). Wei said TSMC expects its N2 family's annual capacity "to more than double in 2027."

New nodes cost margin early. CFO Wendell Huang said the 2nm ramp was the main reason gross margin slipped 1.3 points from Q2. He put the margin dilution from overseas fabs at closer to 2 points for 2026 and said it will widen to 3–4 points over the next few years as more US fabs ramp.

What is TSMC guiding for Q4 and 2026?

Last verified: October 16, 2026 (TSMC release; TSMC call transcript; Focus Taiwan)

At the midpoint, the Q4 range is up 58% year over year and 14% from Q3, Focus Taiwan calculated. Through Q3, TSMC had spent US$40.64 billion of its 2026 capital budget, according to its management report. Huang said capital spending from 2026 to 2028 would be significantly higher than in the previous three years.

Why it matters

TSMC sits under almost every AI build-out. Nvidia's Vera Rubin systems now in production (our Vera Rubin story), custom chips from cloud providers and the gigawatt-scale datacenters in our tracker all depend on its leading-edge capacity and packaging. Its quarterly results are one of the few hard, company-reported data points on how much AI hardware is actually being built and paid for, rather than announced.

They measure chip demand, though, not whether AI customers will earn back what they spend. TSMC's customers are chip designers and cloud giants. The financing stress in mid-October hit the companies further downstream that borrow to build datacenters.

What's confirmed, reported and uncertain

• Confirmed (TSMC release and management report, October 15): revenue, net income, EPS, margins, node shares (including 2nm at 3% in Q2), HPC at 68% of revenue, Q3 capex and Q4 guidance.
• Stated on the earnings call (management, per TSMC's transcript): the 44% growth forecast, the $60–64 billion capital budget, the N2 capacity plan and the AI demand comments.
• Reported: the Q4 midpoint math (Focus Taiwan/CNA); share moves and consensus estimates (Investing.com).
• Uncertain: TSMC doesn't break out AI accelerator revenue, so how much of HPC is AI-specific isn't disclosed. Management's long-range AI growth view is a forecast.