Record Quarter Driven by AI Infrastructure Spending
<cite index="8-1,8-2">Taiwan Semiconductor Manufacturing Company (TSMC) reported second-quarter 2026 revenue of approximately NT$1.27 trillion, equivalent to $39.6 billion — a 36% increase over the same period in the prior year and marginally above analysts' consensus forecasts.</cite> <cite index="9-4,9-5">The result landed at the high end of TSMC's own projected guidance range of $39.0–$40.2 billion and exceeded the previous quarterly revenue record of $35.8 billion set in Q1 2026.</cite>
<cite index="6-4">Revenue for the first six months of 2026 reached NT$2.40 trillion, an increase of 35.6% compared to the corresponding period in 2025.</cite> Full earnings — including net income, margins, and node-level revenue breakdown — were scheduled for release on July 16–17, 2026. <cite index="27-12">In the prior quarter, TSMC reported revenue of $35.90 billion, gross margin of 66.2%, and operating margin of 58.1%.</cite>
N3 and CoWoS Capacity Both Fully Booked
<cite index="16-2,16-3">TSMC's N3 manufacturing process and Chip-on-Wafer-on-Substrate (CoWoS) advanced packaging are both sold out through year-end; N3 operates on booking cycles that run 12 to 18 months ahead of delivery.</cite>
<cite index="15-3,15-4">In 2026, nearly every major artificial intelligence (AI) accelerator family — including Nvidia's Rubin, Google's Tensor Processing Unit (TPU) v7/v8, Amazon's Trainium3, and AMD's MI350X — is transitioning to TSMC's N3 manufacturing process simultaneously, producing a demand shock that the foundry cannot expand into quickly enough.</cite> <cite index="15-11,15-12">Effective N3 utilization is expected to surpass 100% in the second half of 2026, with TSMC already shifting certain process layers to other fabs to extract incremental capacity.</cite>
<cite index="11-8">TSMC is sold out on its N3 process, the node used by nearly every leading AI Graphics Processing Unit (GPU) and Central Processing Unit (CPU) shipping this year.</cite> <cite index="13-1">SemiAnalysis models AI accelerators, host CPUs, and networking silicon collectively consuming roughly 60% of all N3 wafer output in 2026, rising to 86% in 2027 — nearly entirely displacing the smartphone and PC chips that previously dominated the node.</cite>
CoWoS — TSMC's proprietary packaging technology that integrates compute dies with High Bandwidth Memory (HBM) on a shared silicon interposer — faces parallel strain. <cite index="22-4">With lead times of 52 to 78 weeks and bookings extending into 2027, CoWoS availability is now the primary constraint on how fast AI accelerators can be deployed.</cite> <cite index="23-5">Nvidia has reportedly booked 60% of TSMC's CoWoS capacity through 2026 plus more than half of the 2026–2027 expansion.</cite>
CEO Warns Supply Will Lag Demand "For Years"
<cite index="12-13,12-14,12-15">At TSMC's June 4, 2026 annual shareholders meeting in Hsinchu, CEO C.C. Wei warned that global chip supply will "lag behind" AI-driven demand "for years," and reaffirmed revenue growth of more than 30% for 2026 in U.S. dollar terms.</cite>
<cite index="17-6,17-7">Advanced-node capacity — the sub-7nm and sub-5nm chips that power Nvidia's latest GPUs and custom AI accelerators — has been under severe constraints since 2024, with demand at leading nodes expected to exceed capacity by 25–30% in 2026.</cite>
Capital Expenditure and Expansion Plans
<cite index="24-8">TSMC has set a record 2026 capital budget of $52 billion to $56 billion.</cite> <cite index="10-10,10-11,10-12">In Taiwan, TSMC plans to add a new 3nm fab to its GigaFab cluster in Tainan Science Park targeting volume production in the first half of 2027; in Arizona, its second fab will also adopt 3nm technologies, with construction already completed and volume production set for the second half of 2027.</cite>
<cite index="26-5">Taiwan officials said the company will also add two more advanced packaging plants in Chiayi, expanding capacity for CoWoS technologies critical for AI chips.</cite>
<cite index="6-7">TSMC, now valued at approximately $2.25 trillion, has become central to the global buildout of AI infrastructure, with major clients including Apple and Nvidia depending on its chips.</cite> <cite index="11-6">The company held 73% of the global pure-play foundry market in the first quarter of 2026, according to Counterpoint Research data.</cite> The June monthly figure that closed the quarter underscored the structural shift: <cite index="8-14">June revenue rose 67.9% year-over-year to NT$442.68 billion</cite>, <cite index="25-4">breaking a four-year pattern of month-over-month June declines that had previously reflected summer softening in consumer electronics demand.</cite>