Register for a premium account to gain access to Sterling AI.
Get StartedThings you can ask Sterling:
Taiwan Semiconductor Manufacturing Company Limited manufactures, packages, tests, and sells integrated circuits and other semiconductor devices worldwide.
World's largest semiconductor foundry, manufacturing virtually all leading-edge AI chips for NVIDIA, AMD, Apple, and others. Over $90B annual revenue with 60%+ global foundry market share and a monopoly on cutting-edge nodes at 3nm and 2nm. Diversifying geographically with new fabs in Arizona, Japan, and Germany while maintaining its process technology lead.
The world's most important company you've never heard of — manufactures virtually every cutting-edge chip.
N3E / N3P
3nm ProcessCurrent leading-edge for Apple, NVIDIA, AMD
N2
2nm ProcessNow in volume production, powering next-gen chips
CoWoS Packaging
Advanced PackagingCritical for stacking HBM on AI chips
InFO
Packaging TechnologyFan-out packaging for mobile chips
Market Share
~60% of global foundry revenue, ~90% of leading-edge (<7nm)
Competitive Moat
5+ years process tech lead, massive scale, deep customer co-design relationships
Key Risk
Taiwan geopolitical risk; single point of failure for global chip supply
Every NVIDIA GPU, every Apple chip, every AMD processor is made by TSMC. If something happened to TSMC's Taiwan fabs, the entire AI revolution — and most consumer electronics — would grind to a halt. That's how critical they are.
Key Milestones
Founded February 21 in Hsinchu, Taiwan by Morris Chang as the world's first pure-play foundry; backed by Taiwan government (48%), Philips (28%) and Taiwanese conglomerates, the model would upend integrated chipmakers worldwide.
Began commercial production at Fab 1 in Hsinchu Science Park on 3-micron CMOS, the foundry's first revenue node; early customers included VLSI Technology and Adaptec.
IPO on Taiwan Stock Exchange and NYSE October 8 (ADR); first major foundry public listing and the financial engine for the multi-fab buildout that followed.
Began 130nm production at Fab 12, drawing first lead-customer wins from Altera and Xilinx; revenue topped $5B as PC graphics, modems and FPGAs migrated to the foundry model.
Began 90nm production at Fab 12 in Hsinchu; first foundry to lead a node ahead of IDMs, putting TSMC on a structural path past UMC and IBM.
Began commercial 40nm production; early yield issues at TSMC delayed key NVIDIA Fermi GPUs by 6+ months and triggered a strategic review of the foundry partnership.
Began 28nm volume production at Fab 14, the high-K-metal-gate node that made TSMC the foundry of record for fabless mobile chips; ramped past 10K wafers/week within a year.
Began 16nm FinFET volume production, the foundry’s first 3D-transistor node; enabled Apple A10 and full mobile SoC migration off 28nm planar transistors.
First foundry to begin volume production of 7nm FinFET (N7) at Fab 15 in Tainan; Apple A12 was lead customer at iPhone XS launch September 2018, validating TSMC's lead over Intel.
Surpassed 1B 7nm chips shipped milestone with Apple A12 as flagship customer at iPhone XS launch; cumulative N7 wafer volume hit 1M by mid-2019.
Announced $12B Arizona fab (Fab 21) for 5nm production, the largest US foreign greenfield FDI; later expanded to three fabs and $65B+ commitment under CHIPS Act incentives.
US Commerce Department added Huawei to Entity List with foreign-direct-product rule; TSMC stopped Huawei HiSilicon Kirin shipments September 2020, removing 10-15% of TSMC revenue.
5nm (N5) entered volume production, debuting in Apple A14 for iPhone 12 with 1.8x density vs N7; AMD Zen 4, NVIDIA H100 and Apple M-series followed through 2022-23.
Broke ground on Phoenix Arizona Fab 21 with 5nm process target, kicking off the largest US foreign greenfield FDI; ground-breaking attended by VP Kamala Harris and Tim Cook.
Began JASM Kumamoto Japan fab construction in partnership with Sony and Denso; Japan government provided ~$3.5B in subsidies as part of supply-chain reshoring policy.
October 7 BIS rules forced foundries to flag advanced AI/HBM chip orders destined for China; TSMC stopped accepting orders from sanctioned Chinese AI startups including Biren and Moore Threads.
Began 3nm (N3) commercial production at Fab 18 in Tainan with Apple as lead customer; N3 wafer pricing reached ~$20K vs ~$17K for N5, reflecting the EUV-heavy mask layer cost.
Joint venture ESMC announced with Bosch, Infineon and NXP for a 10B+ euro Dresden fab on 28/22nm and 16/12nm; Germany committed up to 5B euros in subsidies for European chip sovereignty.
Opened JASM Fab 1 in Kumamoto, Japan February 24, the first TSMC majority-owned overseas fab on 12-40nm process; second JASM fab broke ground later in 2024 for 6nm/7nm production.
Awarded preliminary $6.6B CHIPS Act funding plus $5B in loans, with plans for a third Arizona fab on 2nm/A16; Taiwan government urged TSMC to keep most-advanced node in Taiwan despite US pressure.
Phoenix Fab 21 ramp delayed by labor and equipment-install issues; first wafers slipped to late 2024 from initial 2024 H1 target, intensifying scrutiny of US fab competitiveness.
Began N2 risk production at Hsinchu Baoshan, the foundry's first nanosheet GAA node; targeted N2 mass production for H2 2025 with Apple, AMD and NVIDIA as initial customers.
Doubled CoWoS advanced-packaging capacity to ~35K wafers/month, struggling to meet AI accelerator demand; CoWoS booking lead time stretched to 18 months by Q4.
Finalized $6.6B CHIPS Act direct funding award November 8; Phoenix Fab 21 began ramping 4nm chips in volume to lead customer Apple ahead of formal opening.
Completed N2 risk production with ~5,000 wafers at Hsinchu, on track for H2 2025 mass production; Apple A20 and AMD MI400 confirmed as lead N2 customers.
Pulled Phoenix Fab 21 third fab forward to 2027 to host 2nm and A16 production for US AI customers; first time TSMC committed to leading-edge node outside Taiwan.
Began N2 volume production at Hsinchu Fab 20 December 31, lead customers Apple and AMD on CDNA5; output target was 50K wafers/month exiting Q1 2026, ramping to 100K+ in 2026.
CoWoS-L/S capacity reached ~75K wafers/month; both lines fully booked through 2026 driven by NVIDIA Rubin and AMD MI400 demand.