Taiwan Semiconductor Manufacturing Company (TSMC) reported record second-quarter 2026 results on July 16, posting net revenue of $39.62 billion—a 36% surge year-over-year—as the world's largest contract chipmaker continues to benefit from relentless AI infrastructure spending by the world's biggest technology companies.
Key Financial Highlights
- Q2 Revenue: $39.62 billion USD, within the guidance range of $39.0–$40.2 billion
- June Revenue: NT$442.68 billion, up 67.9% year-over-year — the largest single-month revenue gain in TSMC's four-decade history
- H1 2026 Revenue: NT$2,404.48 billion ($74.99 billion), up 35.6% compared to the first half of 2025
- Expected EPS: $3.83, representing approximately 55% year-over-year growth
- Gross Margin Guidance: 65.5–67.5% — maintaining premium economics despite record capital expenditure
- YTD Stock Performance: Up approximately 52% year-to-date
The results mark nine consecutive quarters in which TSMC has met or beaten Wall Street estimates, cementing its position as the definitive bellwether for the global AI hardware cycle.
AI Chips Now Drive the Majority of Revenue
High-performance computing (HPC)—the segment that includes AI accelerators for Nvidia, AMD, Google, Amazon, and Apple—accounted for 61% of TSMC's total Q1 2026 revenue, and that figure is expected to hold or climb in Q2 results. TSMC manufactures virtually every advanced AI chip on the market, giving its quarterly earnings an outsized signal value for the entire technology sector.
Chief Executive C.C. Wei reiterated TSMC's full-year 2026 growth target of more than 30% at the company's annual shareholders' meeting. He struck a blunt tone on supply constraints: "It will be a long time before we can meet customer demand."
N3 Sold Out, N2 Ramping
TSMC's advanced 3-nanometer process (N3) remains entirely sold out through the end of 2026, with N2 production—the company's next-generation 2-nanometer node—already ramping ahead of a commercial rollout expected in the second half of 2026. Customers secured on N2 engineering runs include major AI chip designers looking to lock production slots before demand further outstrips capacity.
The supply crunch extends beyond logic silicon. CoWoS advanced packaging — a critical interconnect technology that bonds high-bandwidth memory layers to AI accelerators — was running at roughly 13,000 wafer-equivalents per month in late 2023. Industry projections now estimate capacity could reach 120,000–140,000 wafer-equivalents per month by year-end 2026, a more than tenfold expansion in under three years. Even so, CoWoS demand continues to exceed supply, underscoring the scale of AI infrastructure investment underway.
Record Capital Expenditure
TSMC's 2026 capital expenditure plan of $52–56 billion is tracking toward the upper end of guidance, with approximately 70–80% of investment directed toward advanced nodes — N3, N2, and beyond. The scale of this commitment reflects confidence that AI-driven demand will remain durable through the end of the decade.
Analysts noted ahead of the earnings call that customers are "pivoting from software to hardware," pointing to a structural shift in AI development from model training compute toward inference chips and on-device AI acceleration — a trend that plays directly into TSMC's advanced packaging and leading-edge node strengths.
Market Position and Competitive Context
TSMC commands approximately 73% market share among advanced foundry customers as of Q1 2026. The company's dominance spans every tier of the AI chip supply chain: Nvidia's H200 and B300 AI GPUs, Google's TPU v6, Amazon's Trainium 2, Microsoft's Athena custom accelerator, and Apple's M4 neural engine are all manufactured at TSMC facilities.
The results arrive as competition in custom silicon intensifies. OpenAI unveiled its Jalapeño inference chip built with Broadcom in June 2026, and Anthropic is in early-stage talks with Samsung to develop dedicated inference processors for Claude. Rather than threatening TSMC's position, these moves underscore the structural boom in advanced semiconductor manufacturing demand — every new custom chip design ultimately requires a foundry to manufacture it, and TSMC remains the only facility capable of delivering at the leading edge.
What's Next
Eyes are on TSMC's Q3 2026 guidance, delivered during the July 16 earnings call. Analysts expect Q3 revenue to show continued sequential growth, with particular interest in whether N2 ramp timelines are accelerating and whether CoWoS capacity expansion is on track. Any signals on 2027 pricing for advanced nodes — a potential indicator of further margin expansion — will be closely watched.
The results confirm a structural reality driving technology markets in 2026: AI infrastructure investment is not slowing, and TSMC sits at the center of everything being built.
Source: TSMC Investor Relations