TSMC

Taiwan Semiconductor Manufacturing Company (TSMC) is the world's largest dedicated independent semiconductor foundry, responsible for manufacturing the majority of the advanced chips that power the global artificial intelligence industry. Founded in 1987 by Morris Chang in Hsinchu, Taiwan, TSMC pioneered the pure-play foundry business model — manufacturing chips designed by its customers rather than selling its own branded semiconductors — and has grown into the most critical node in the global AI hardware supply chain.^[1]

Founding and Business Model

TSMC was established in 1987 as a spin-off from Taiwan's Industrial Technology Research Institute (ITRI). Morris Chang, a former Texas Instruments executive, recognised that as semiconductor design became increasingly complex, a market would emerge for companies that specialised purely in manufacturing — allowing "fabless" chip designers to focus on design while outsourcing fabrication.

This pure-play foundry model transformed the semiconductor industry. Companies like NVIDIA, AMD, Qualcomm, and later Google (for its TPU AI accelerators) could design cutting-edge chips without owning billion-dollar fabrication plants. TSMC became the manufacturing partner of choice, investing aggressively in advanced process nodes that individual fabless companies could not justify alone.

Role in the AI Industry

TSMC manufactures the silicon for nearly every major AI accelerator on the market:

• NVIDIA GPUs: TSMC fabricates NVIDIA's entire GPU line, including the Hopper (H100/H200), Blackwell (B100/B200/GB200), and Rubin architectures that dominate AI training and inference in data centres worldwide.^[1]
• Google TPUs: Google's custom Tensor Processing Units, used for Gemini model training and Google Cloud AI services, are manufactured by TSMC.^[1]
• AMD accelerators: AMD's MI300 and MI400 series AI accelerators are TSMC-fabricated.
• Apple silicon: The M-series and A-series chips that power Apple Intelligence are manufactured by TSMC.

This concentration means that TSMC's production capacity, technology roadmap, and financial performance are leading indicators of the entire AI industry's health. As one analyst noted, "TSMC's receipt drives the AI trade."^[2]

Financial Performance and AI Demand

TSMC's financial results in 2026 have demonstrated the explosive demand generated by the AI buildout:

• July 2026 revenue: NT$467.58 billion (US$14.5 billion), up 44.7% year-on-year.^[3]
• Q2 2026 revenue: NT$1.27 trillion, a 36% year-on-year rise, with June 2026 setting a record as the best sales month in company history at NT$442.68 billion (67.9% YoY growth).^[3]
• H1 2026 revenue: NT$2.4 trillion (US$74.99 billion), running 35.6% ahead of the previous year.^[3]
• High-performance computing (HPC), the segment where TSMC books AI chip revenue, accounted for 66% of Q2 2026 revenues.^[3][4]

TSMC raised its 2026 revenue guidance to slightly above 40% growth in US dollar terms (up from a prior forecast of over 30%) and increased its capital expenditure projection to US$60–64 billion for the year. Chairman C.C. Wei stated: "AI-related demand continues to be extremely robust."^[3][4]

Analysts' average 2026 revenue estimate for TSMC is approximately US$5.2 trillion (NT$), which would represent a 36.5% gain over 2025.^[5]

Advanced Process Technology

TSMC's competitive advantage rests on its leadership in advanced semiconductor manufacturing nodes:

3-Nanometre (N3)

The N3 process node is in high-volume production and is used for current-generation AI accelerators and premium consumer silicon. TSMC's N3 technology delivers significant improvements in power efficiency and transistor density over the preceding N5 (5nm) generation.

2-Nanometre (N2) Ramp

TSMC is aggressively ramping its N2 (2-nanometre) process, the industry's most cutting-edge production technology. The N2 transition is expected to introduce Gate-All-Around (GAA) transistors, also known as Nanosheet FETs, replacing the FinFET architecture that has dominated advanced nodes since the 2010s. The ramp has caused some near-term gross margin pressure — a normal phenomenon during early yield optimisation — but is expected to deliver substantial performance and density gains for next-generation AI chips.^[4][5]

CoWoS Advanced Packaging

TSMC's Chip-on-Wafer-on-Substrate (CoWoS) advanced packaging technology is critical for AI accelerators, which combine multiple silicon dies (compute, high-bandwidth memory, and interconnect) into a single package. CoWoS capacity has been a bottleneck for AI chip supply, and TSMC has invested heavily in expanding this capability to meet demand from NVIDIA, AMD, and Google.^[4]

Geopolitical Context

TSMC's dominance in advanced chip manufacturing has made it a focal point of geopolitical tension between the United States and China, and a national security concern given Taiwan's contested political status. The concentration of over 90% of the world's most advanced chip production on a single island has been described as the semiconductor industry's greatest single point of failure.

In response, TSMC has expanded production internationally:

• United States: TSMC is building fabrication plants in Arizona (Fab 21) to produce N3 and eventually N2 chips for the US market, supported by CHIPS Act funding.
• Japan: TSMC's JASM joint venture operates a fab in Kumamoto, Japan, with a second facility under construction.
• Germany: TSMC is building a fab in Dresden as part of the European Chips Act strategy.

>See Also

• NVIDIA
• AMD
• Intel
• High-Bandwidth Memory
• AI Data Centres Malaysia
• National Semiconductor Strategy Malaysia

References