TSMC Founder Morris Chang

Episode Summary

Executive Summary: This episode centers on Morris Chang’s account of TSMC’s rise through disciplined foundry strategy, learning-curve economics, and critical relationships with NVIDIA and Apple. Chang explains how TSMC navigated manufacturing setbacks, avoided layoffs, committed heavily to 28nm/20nm, and used customer trust plus massive scale to become the world’s leading semiconductor foundry.

Main Topics: TSMC and Jensen Huang / NVIDIA partnership (Priority: 5/5): Chang recounts how Jensen Huang first reached out in 1997 from a struggling NVIDIA, how TSMC responded, and how that relationship later became a foundational customer partnership despite temporary disputes over 40nm yield and quality issues. Leadership crisis, layoffs, and Chang retaking CEO duties (Priority: 5/5): Chang describes the 2008-2009 downturn, the outgoing CEO’s performance-based layoffs, the protests at Chang’s home, and his decision to retake CEO control to restore TSMC’s culture and fix the business. 28nm strategy, R&D spending, and capital intensity (Priority: 5/5): Chang explains the decision to lock in an 8% of revenue R&D budget and dramatically expand capex to win the 28nm node, describing it as the company’s strategic sweet spot and a key inflection point in TSMC’s leadership. Apple’s entry and the 20nm/16nm tradeoff (Priority: 5/5): The transcript details how Apple, via Jeff Williams, approached TSMC for iPhone chips, prompting a costly but ultimately successful 20nm commitment that delayed 16nm and briefly led Apple to sample Samsung before returning to TSMC. Pure-play foundry model and customer non-competition (Priority: 5/5): A major theme is TSMC’s core identity as a dedicated pure-play foundry that does not compete with customers, contrasted with integrated firms like Intel and the strategic advantage this created over time. Learning curve theory and scale economics (Priority: 4/5): Chang reflects on his TI-era work with BCG/Bill Bain and how learning-curve thinking shaped TSMC’s pricing, capacity planning, and long-term strategy to win volume and reduce unit cost faster than rivals. The Taiwan semiconductor ecosystem and industrial policy (Priority: 4/5): The discussion closes with the importance of Hsinchu Science Park, the cluster of customers, suppliers, universities, and talent in Taiwan, and why this ecosystem cannot easily be replicated elsewhere.

Key Arguments: TSMC’s pure-play foundry model was strategically superior because it never competed with customers and could fully align around manufacturing excellence. Customer trust matters as much as technology; Intel’s reputation and culture made it hard for it to succeed as a foundry even when it had the technical lead. TSMC’s 40nm issues were not just a customer-specific dispute but a broader manufacturing/yield challenge that required executive intervention and compensation. Layoffs based on subjective performance reviews damage morale and credibility; TSMC’s culture favored rehiring/transfer over firing. Raising R&D to 8% of revenue removed annual budget fights and enabled faster node execution. The 28nm node was a key strategic “sweet spot” that justified aggressive capex and established TSMC’s leadership trajectory. The Apple 20nm deal was worth it even though it delayed 16nm, because it secured a massive customer relationship and expanded long-term scale. Learning curve economics reward the largest-volume player; TSMC’s strategy was to get to scale faster than competitors and then compound the advantage. Taiwan’s science-park ecosystem, universities, and tight supplier/customer proximity are a critical, non-transferable asset for TSMC. TSMC’s success was enabled by Moore’s Law tailwinds and the broader reorganization of the semiconductor value chain into specialized companies.

Data Points: NVIDIA founding age at first TSMC outreach: 4 years old - Chang says NVIDIA was a small, early-stage company when he received Jensen Huang’s 1997 letter. TSMC revenue in 1995: Over $1 billion - Chang notes TSMC had already passed $1B in revenue by 1995, before the NVIDIA relationship scaled. TSMC scale vs NVIDIA in 1997: A few thousand employees vs. 50–60 at NVIDIA - Illustrates the imbalance between the companies when the relationship began. Major customer threshold: At least $50 million annual revenue - Chang says a customer had to generate roughly this much to count as a major customer for TSMC. 40nm affected period: 2009 - The transcript centers on yield and quality problems at the 40nm node during this period. Employees laid off by prior CEO: 600–700 - Chang says the prior CEO terminated this many workers based on poor performance reviews. Protest police presence: 50–60 officers - Police were sent to Chang’s home when the laid-off employees protested. 2008-2009 R&D budget target: 8% of revenue - Chang made this a fixed policy to avoid yearly budget fights and accelerate R&D. Prior R&D spend: 6%–7% of revenue - Chang says TSMC was already around this range before he set the 8% target. Capex ramp: From about $2B–$2.5B annually to almost $6B in 2010 - TSMC sharply increased fab spending to pursue 28nm leadership. Apple initial gross margin offer: 40% - Jeff Williams suggested this margin in early discussions; Chang notes TSMC was already above it. TSMC gross margin at the time: 45% - Chang says TSMC was already around this level during the Apple negotiations. Target gross margin: 50% - Chang says he had been trying for years to push TSMC to this level. Apple demand reduction: Half of Apple’s requested volume - TSMC decided to fulfill only half the original Apple demand due to financial prudence and capacity constraints. Debt financing: Billions of dollars - TSMC chose borrowing as the main way to fund the Apple-driven expansion. TSMC market cap scale: Trillion-dollar company - The hosts note TSMC is one of the few trillion-dollar companies outside the U.S. West Coast. Global semiconductor market in 1987: $26 billion - Hosts cite this as the market size when TSMC was founded. Global semiconductor market last year: $527 billion - Used to show the massive industry tailwind since TSMC’s founding. R&D director and business development team size: 60–70 people - Chang says the combined business development function was relatively small but strategically important. TSMC/Taiwan ADR premium: 20% premium - Chang notes the ADR trades above the Taiwan-listed shares.

Pivotal Quotes: "“be a dedicated pure play foundry”" — Morris Chang: Referenced as an original TSMC business-plan principle and the core of the company’s strategic identity. "“Intel just does not know how to be a foundry.”" — Tim Cook (as reported by Chang): Cook reassured Chang during Apple’s evaluation of suppliers, helping TSMC’s case versus Intel. "“There’s a tide in the affairs of man, which, taken at its flood, leads on to fortune.”" — Morris Chang: Chang uses Shakespeare to describe the timing and conviction behind the 28nm bet.

Implications: The episode shows that semiconductor leadership is built on specialization, trust, and relentless scale—not just chip design. For the industry, it reinforces why TSMC’s foundry model is hard to replicate and why ecosystem, capital, and execution matter as much as process technology.