TSMC 2024: how the Taiwanese chip foundry's structural advantage in advanced-process technology became the AI infrastructure foundation while geopolitical complexity intensified
Taiwan Semiconductor Manufacturing Company (TSMC) emerged as the most strategically important company in technology through 2024. The world's largest contract chip foundry produces approximately 90%+ of the world's advanced (3nm and 5nm) chips for customers including Apple (A18 chips), Nvidia (H100/H200/Blackwell), AMD (Instinct GPUs), Qualcomm (Snapdragon), Broadcom (custom AI chips for Google/Meta), and dozens of others. Q3 2024 revenue $23.5B (+36% YoY); AI-related revenue grew exceptionally; gross margins ~57%. The company's strategic position is unique: a Taiwan-headquartered company that produces the chips powering the global AI infrastructure boom while operating under ongoing US-China geopolitical tension that constrains both customer relationships (Huawei restrictions, broader US export controls) and capacity expansion (Arizona Fab construction, Japan Fab, European Fab). Under CEO C.C. Wei (since June 2023), TSMC has continued sustained operational discipline and CapEx-intensive capacity expansion. The TSMC 2022-2024 chapter is studied as the worked example of foundry-monopoly positioning in the AI era.
- Story: TSMC produces ~90%+ of world's advanced (3nm/5nm) chips for Apple, Nvidia, AMD, Qualcomm, Broadcom, others. Q3 2024 revenue $23.5B (+36% YoY) with ~57% gross margin. CEO C.C. Wei (since June 2023) continues sustained operational discipline. CapEx 2024 $30-32B. AI demand from Nvidia H100/Blackwell, custom AI silicon (Google TPU, Meta MTIA via Broadcom) consistently exceeds TSMC capacity. Geographic expansion: Arizona Fab (delayed late 2024/early 2025), Japan Fab (operational late 2024), Germany Fab (under construction). $6.6B CHIPS Act funding awarded. Geopolitical: US-China export controls limit Chinese customer access; Taiwan strait risk continues.
- Why it matters: TSMC 2022-2024 is the worked example of foundry-monopoly positioning during major technology category emergence: sustained R&D and CapEx investment compounds over decades into structural moats competitors can't easily displace.
- Takeaway: Compound technology-infrastructure investment over decades produces structural moats single strategic initiatives can't replicate.
- Takeaway: Pure-foundry business model produces customer-relationship advantages over integrated competitors.
- Takeaway: Geopolitical risk is structural concern that operational excellence alone cannot fully address.
TSMC AI dominance + geopolitics — the four-step story
TSMC AI dominance + geopolitics at a glance
Quick facts
The TSMC strategic moat and the AI demand acceleration
TSMC was founded in 1987 by Morris Chang in Hsinchu, Taiwan. Chang's strategic insight: separate chip design from manufacturing, allowing fabless companies (Apple, Nvidia, AMD, Qualcomm, etc.) to focus on design while TSMC focused on manufacturing excellence. The pure-foundry model produced sustained competitive advantage:
- Pure-foundry business model: TSMC doesn't design competing chip products. Customers can engage with TSMC without competitive-design conflicts that integrated competitors (Samsung, Intel) face.
- Sustained advanced-process leadership: TSMC's 3nm (production 2022) and 5nm (production 2020) process nodes have maintained leadership over Intel (delayed 18A) and Samsung (yield issues at 3nm). 2nm production scheduled for 2025.
- Customer concentration on largest customers: top 10 customers represent ~80%+ of revenue. Apple is largest single customer at ~25%+ of revenue. Nvidia, AMD, Qualcomm, Broadcom, MediaTek also major.
- AI demand acceleration 2023-2024: Nvidia's H100 and Blackwell GPU production is exclusively on TSMC advanced nodes. Demand has consistently exceeded TSMC capacity throughout 2023-2024.
- CoWoS advanced packaging: TSMC's chip-on-wafer-on-substrate packaging is the binding constraint for high-bandwidth-memory (HBM) integration on AI GPUs. CoWoS capacity has been growing aggressively but consistently undersupplied.
- 2024 capacity build-out: TSMC continued aggressive capacity expansion with N3 (3nm) production scaling and N2 (2nm) production preparation.
The 2024 financial performance and the AI-driven growth
TSMC's 2024 financial performance reflects the AI demand acceleration:
- Q3 2024 revenue $23.5B (+36% YoY): substantial growth despite mature category baseline.
- 2024 full-year revenue trajectory $90B+: significant growth from 2023 $69.3B.
- HPC (high-performance computing) revenue share growing: from ~40% of revenue in 2022 to ~50%+ in 2024. HPC includes AI training/inference chips.
- Smartphone revenue share declining: from ~40% to ~30% of revenue. Reflects HPC growth rate exceeding smartphone.
- Q3 2024 gross margin ~57%: substantial profitability despite massive CapEx.
- Net income strong: Q3 2024 net income ~$10B.
- CapEx discipline: 2024 CapEx $30-32B (vs $30B-$32B 2023). Substantially higher than peer foundries but justified by continued advanced-process leadership.
- Dividend stability: TSMC continues conservative dividend policy alongside CapEx-intensive growth.
- Stock recovery: TSMC stock recovered substantially through 2024 from 2022 lows. Market cap exceeded $1T at peak.
The geopolitical complexity and the geographic expansion
TSMC's strategic position has been complicated by US-China geopolitical tension:
- US-China export controls: October 2022 US export controls (and subsequent tightening) prevented TSMC from selling advanced chips to certain Chinese customers. Huawei, SMIC, various AI-related Chinese companies restricted.
- Taiwan strait risk: Taiwan-China relationship remains tense; potential military escalation is structural risk that affects TSMC strategically. Some customers have begun diversifying supply away from Taiwan-only production.
- Arizona Fab: TSMC announced Arizona Fab construction in May 2020 with $40B+ subsequent investment. Initial Arizona production targeted 4nm 2024; delayed multiple times to late 2024/early 2025. Construction-cost overruns and labor-relations issues compounded.
- Japan Fab: Kumamoto facility operational late 2024 for older-node (16nm-28nm) production primarily for Sony/Denso/other Japanese customers. Sustainable but less strategically central than advanced-process production.
- Germany Fab: Dresden facility under construction; 28nm-22nm production planned 2027. Subsidized by EU CHIPS Act.
- CHIPS Act funding US: TSMC awarded $6.6B in direct CHIPS Act funding (April 2024 announcement, finalized late 2024) for Arizona expansion.
- Customer-base diversification: Apple, Nvidia, AMD all have public statements supporting diversification of TSMC capacity to US/Japan/Germany alongside Taiwan production.
- Structural production-cost advantage Taiwan: Taiwan production remains ~30% cheaper than US/Japan/Germany production due to labor, regulatory, and supply-chain factors. Advanced-process production still concentrated in Taiwan.
The C.C. Wei CEO transition and the strategic continuation
C.C. Wei became TSMC CEO on June 6, 2023, succeeding Mark Liu (who had been co-CEO 2018-2023). Wei had been at TSMC for 25+ years in senior R&D and operations roles:
- Internal-promotion succession: Wei's TSMC tenure provided continuity of operational discipline.
- Strategic-direction continuity: Wei's framework continues Mark Liu and Morris Chang strategic priorities. No major strategic-direction reset.
- Continued advanced-process leadership push: N2 (2nm) production preparation 2024-2025; A16 (1.6nm equivalent with backside power delivery) targeted for 2026 production.
- Geographic-expansion continuation: Arizona, Japan, Germany expansion all continued.
- CoWoS capacity prioritization: aggressive expansion of advanced packaging to support AI GPU customers.
- Customer-relationship management: Wei has personally engaged with Apple, Nvidia, AMD CEO-level relationships.
- Morris Chang continued influence: founder (92 years old in 2023) continues advisory role; his strategic-direction framework remains foundational.
How RGM thinks about foundry-monopoly positioning in technology cycles
TSMC's 2022-2024 chapter is the worked example of foundry-monopoly positioning during major technology category emergence. The structural elements: sustained R&D and CapEx investment over decades produced advanced-process technology leadership; pure-foundry business model produced customer-relationship advantages over integrated competitors; geographic concentration in Taiwan produced cost advantages but also geopolitical risk; AI demand acceleration produced exceptional financial performance; geographic-expansion strategy addresses geopolitical risk but at higher production costs.
Our framework for clients in similar technology-infrastructure-leadership situations: structural advantages compound over decades when sustained R&D and CapEx investment is maintained. TSMC's position in AI infrastructure reflects 35+ years of consistent operational discipline. The position is contestable in principle (Intel's IDM 2.0 strategy was specifically designed to challenge it; Samsung continues investment) but practically very difficult to displace. Geopolitical risk is the structural concern that no operational excellence can fully address. We tell clients in technology-infrastructure categories that compound investment discipline produces structural moats that competitors typically can't match through any single strategic initiative. TSMC's foundry monopoly is the canonical example.
Frequently asked questions
Could Samsung or Intel realistically catch up?
Difficult. Samsung Foundry has struggled with 3nm and 4nm yield issues; major customers (Qualcomm, Nvidia in selected cases) have shifted business to TSMC. Intel's IDM 2.0 strategy was specifically designed to compete on advanced-process technology but Pat Gelsinger's December 2024 departure and 18A delays have raised questions about Intel's competitive position. Most analysts expect TSMC to maintain advanced-process leadership through 2030+ unless geopolitical disruption forces customer-base shifts.
What about Taiwan strait risk?
Real but managed. Taiwan-China military escalation would disrupt TSMC and global semiconductor supply chains dramatically. The probability and timing is genuinely uncertain. TSMC's geographic-expansion strategy (Arizona, Japan, Germany) addresses tail-risk but advanced-process production will remain concentrated in Taiwan for the foreseeable future. Most major customers maintain Taiwan-production-dependency despite the risk.
How profitable can TSMC be at this scale?
Substantially. Q3 2024 gross margin ~57% and net margin ~42% are exceptional for any business at $23B+ quarterly revenue. Continued advanced-process leadership and CoWoS premium pricing support sustained margins. AI demand has produced pricing power TSMC hasn't always had at earlier-cycle peaks.
What's the Arizona Fab status?
Delayed but progressing. Initial Arizona production target was 2024 4nm; pushed to late 2024/early 2025. Construction-cost overruns and labor-relations issues compounded. TSMC has continued investment but Arizona production costs are reportedly ~30% higher than Taiwan equivalents. Whether Arizona ever achieves Taiwan-comparable economics is uncertain. Continued US government support (CHIPS Act, $6.6B direct funding) sustains the strategic commitment.
Is TSMC's AI revenue sustainable?
Probably yes for several years. AI infrastructure capex (hyperscaler buildouts) is multi-year program. Nvidia, AMD, Broadcom custom AI silicon for Google/Meta all rely on TSMC. New AI customers continue emerging. The structural demand for advanced-process chips will likely continue through 2027-2030 at minimum. Cycle moderation eventually inevitable but timing uncertain.
Sources & references
- TSMC Q3 2024 earnings — TSMC investor materials and quarterly earnings.
- CHIPS Act funding announcement — Commerce Department April 2024 TSMC funding announcement.
- C.C. Wei CEO appointment — TSMC CEO succession communication.
- Arizona Fab delays coverage — WSJ coverage of Arizona project status.
- AI semiconductor demand context — Bloomberg analysis of TSMC AI revenue dynamics.