"When I introduced myself, said this is Morris Chang, he immediately shouted at those people that were making noises. He said, quiet. Morris Chang is calling me." — Morris Chang [on his first call with Jensen Huang, ~1997] [00:07:01]
"I wasn't all that worried because what reviewed in my mind, all the characteristics that Apple is looking for in a supplier — technology, manufacturing, and customer trust — we thought our customers trusted us more than Intel's customers trusted Intel." — Morris Chang [on the Intel competition for Apple's foundry business, 2011] [01:40:02]
Disclaimer: Orignal content owned by or sourced from third parties. It does not represent the views of 'Nuggets' platform or it's team. AI is used extensively across this platform including for summaries. Accuracy is not guaranteed, there can be mistakes. Any info or content on this platform is not a financial, legal, or investment advice. Do your own research. Refer for complete disclosures:- Terms of Use · Full Disclaimer
"There is a tide in the affairs of men which, taken at the flood, leads on to fortune." — Morris Chang [quoting Shakespeare on committing to 28-nanometer, via his autobiography] [00:46:29]
"Sitting in Hsinchu, being in the foundry business, I actually see a lot of things before they actually happen." — Morris Chang [on the strategic intelligence advantage of the pure-play foundry model] [01:55:24]
"It wasn't bet-the-company to me. I didn't think I would lose." — Morris Chang [on the multi-billion-dollar commitment to serve Apple's 20-nanometer demand] [01:36:12]
"Intel just does not know how to be a foundry." — Tim Cook [to Morris Chang, ~March 2011, explaining Apple's decision against Intel] [01:43:55]
"Since TSMC was founded in 1987, the world's semiconductor market has grown from $26 billion to $527 billion last year. They rode a ridiculous tailwind." — Ben Gilbert [post-game analysis] [02:45:51]
Speakers & Credentials
Morris Chang — Founder and former Chairman & CEO of TSMC (Taiwan Semiconductor Manufacturing Company). Age 93 at time of recording. Spent formative career at Texas Instruments, rising to Head of Worldwide Semiconductors. Previously President of ITRI (Industrial Technology Research Institute) in Taiwan, from which TSMC was spun out in 1987. Returned as CEO in 2009 after a brief succession experiment. Published Volume 2 of his autobiography in early 2025 (in Traditional Chinese; unpublished in English at time of recording). Widely regarded as the architect of the modern semiconductor supply chain and the pure-play foundry model. Former Board Director of Goldman Sachs.
Ben Gilbert — Co-host of Acquired. Co-founder and Managing Director at Pioneer Square Labs (PSL), a Seattle-based startup studio and venture fund.
David Rosenthal — Co-host of Acquired. Angel investor and former early-stage operator; previously at Vevo and other early-stage companies.
Research contributors credited: Karina Bao (memoir translation, funded by Tyler Cowen/Emergent Ventures), Aart de Geus (Synopsys co-founder), Rene Haas (ARM CEO), Sir Peter Bonfield (TSMC board member, former BT CEO), Wally Rhines (former Mentor Graphics CEO), Jon Bathgate & Brinton Johns (NZS Capital), Jon (Asianometry YouTube), Tim Culpan (Culpium Substack), Arvind Navaratnam (Worldly Partners).
Executive Summary
TSMC is the world's dominant pure-play semiconductor foundry, founded in 1987 by Morris Chang in Hsinchu, Taiwan, and as of 2025 one of only two trillion-dollar companies headquartered outside the West Coast of the United States — the other being Saudi Aramco.
The company's core strategic thesis — articulated in Morris Chang's original pitch deck to the Taiwanese government — was to be a dedicated pure-play foundry that would never compete with its customers in chip design, enabling a new class of "fabless" chipmakers to exist at all.
TSMC surpassed $1 billion in revenue by 1995, just eight years after founding, serving initially as a manufacturer of "excess capacity" for integrated device manufacturers (IDMs) before the fabless revolution it anticipated took hold.
The episode centers on three pivotal relationships: NVIDIA (beginning 1997), revealing how Morris personally salvaged a near-bankruptcy customer relationship and, in 2009, settled a yield dispute worth more than $100 million; Apple (beginning 2010–2011), detailing how TSMC secured the iPhone chip relationship through a diplomatic dinner with Jeff Williams and Terry Gou, navigated a detour to 20-nanometer node development, absorbed a Samsung shock when Apple briefly placed 16-nanometer orders with Samsung, and ultimately recovered all of Apple's business; and Qualcomm, illustrating how TSMC used competitive intelligence (Qualcomm's exit from IBM Semiconductor) to anticipate IBM's requests and decline a damaging co-development partnership.
The learning curve — co-developed by Morris Chang, Bill Bain (founder of Bain & Company), and BCG's Bruce Henderson at Texas Instruments in the early 1970s — is the fundamental operating philosophy underpinning TSMC's capital deployment, pricing strategy, and market timing.
TSMC's structural moat rests on a self-reinforcing flywheel: highest volume → deepest cost curve → lowest prices → more customers → more volume, compounded by an 8%-of-revenue R&D commitment (set by Morris upon returning as CEO in 2009) and CapEx scaling that tripled to nearly $6 billion in 2010 alone.
The semiconductor market grew from $26 billion (1987) to $527 billion (2024), providing a near-unprecedented industry tailwind; TSMC has ridden this to a market cap in excess of $1 trillion.
The Hsinchu Science Park ecosystem — with Synopsys, Cadence, ARM, Qualcomm, MediaTek, and two PhD-granting universities all co-located — represents what Ben and David call the single most successful government-funded industrial initiative in history, and is essentially irreplicable on any short timeline.
Morris Chang's management philosophy — no layoffs during downturns, subjective performance reviews cannot be the basis for dismissal, long-term customer trust above short-term margin capture — is as central to TSMC's success as any technology decision.
The episode is drawn directly from Morris Chang's newly completed Volume 2 autobiography, making this conversation a primary historical source for events (Apple negotiations, 40nm NVIDIA crisis, IBM refusal) previously only known secondhand.
Morris Chang spent approximately 25 years at Texas Instruments (TI), rising to Head of Worldwide Semiconductors — the largest division of a company that also had defense, materials, geophysics, and controls businesses. [00:39:07]
At TI, Morris was a direct contemporary of Jack Kilby (inventor of the integrated circuit) and Bob Noyce (co-inventor, later co-founder of Intel).
His R&D budget at TI was capped at 4.8% of revenue despite his repeated requests to raise it to 5.5% — a constraint he would later eliminate at TSMC by setting an unconditional 8% of revenue rule. [00:39:54]
In the early 1970s, Morris was one of the only TI executives to oppose performance-based layoff methodology when the company faced a downturn, foreshadowing the management philosophy he would institutionalize at TSMC decades later. [00:19:35]
Morris absorbed the learning curve / experience curve theory by becoming TI's designated partner for BCG's Bruce Henderson and Bill Bain (who would later found Bain & Company). Bill Bain worked three days a week at TI for roughly two years, sitting in an office adjacent to Morris's, developing curve-based pricing strategy specific to the semiconductor industry. This two-year partnership shaped Morris's entire strategic thinking. [02:07:41]
ERA 2: The Origin Story — ITRI and the Birth of TSMC (1983–1987)
Before TSMC, Morris served as President of ITRI (Industrial Technology Research Institute), the Taiwanese government's industrial R&D body, which operated Fab 1 (which TSMC would later utilize). [02:16:30]
Two formative encounters in his final months at General Instrument (a prior role) seeded Morris's conviction that the fabless model was coming: (1)Gordie Campbell (widely credited as the first fabless entrepreneur) came asking for a $50 million investment to start a chip company, only to return three weeks later saying he needed only $5 million because he decided not to build a fab. [02:16:35] (2) The founder of Atmel approached General Instrument to use its spare fabs for his new fabless company. [02:18:48]
In 1987, Morris founded TSMC at age 56 with backing from the Taiwanese government, Philips Electronics (which contributed process technology in exchange for equity), and private Taiwanese investors. The founding thesis — "dedicated pure-play foundry that does not compete with customers" — was explicitly written into his original pitch deck, seen today at the TSMC Museum of Innovation at Hsinchu. [02:22:15]
Initial customers were primarily IDMs (Integrated Device Manufacturers) offloading their worst, least strategic manufacturing orders — secondary-source production for older nodes. Morris knew this was the price of building volume on the learning curve before the fabless wave arrived. [02:16:11]
ERA 3: The Fabless Revolution and First Major Customers (1987–2004)
Growth in the early years was constrained by the slow emergence of the fabless ecosystem. Morris knew the business case but spent the first five to seven years largely dependent on IDM excess-capacity orders. [02:16:42]
In 1995, TSMC crossed $1 billion in annual revenue. [00:06:00]
In 1997, a letter from a four-year-old company called NVIDIA — near bankruptcy, ~50–60 employees — reached Morris's Hsinchu office after being ignored by TSMC's San Jose office. Morris, personally irritated by the sales team's negligence toward a potential future customer, called Jensen Huang from the stationery's phone number during a US trip. Jensen, in the middle of an argument with his staff, shouted "Quiet — Morris Chang is calling me." [00:07:01] Morris was immediately impressed by Jensen's articulateness and bold claim that this new chip would not only save NVIDIA but make it a major TSMC customer within 2–3 years (requiring $50M+ annual revenue). The prediction proved correct. Within 2–3 years, NVIDIA became one of TSMC's top-five customers. [00:09:19]
The IBM-Qualcomm intelligence read: In the mid-to-late 1990s, Qualcomm's operations VP, who had long given TSMC minimal business while relying primarily on IBM Semiconductor, suddenly switched to TSMC en masse (~1997–1998). Morris immediately understood this signal: IBM was losing its major customer. [01:58:49] Sure enough, IBM shortly approached TSMC to co-develop 0.13-micron (130nm) technology, positioning itself as the "senior partner" in the arrangement (requiring TSMC engineers to work at IBM facilities). Morris declined without hesitation, recognizing that accepting would subordinate TSMC's proprietary process development capability and cultural identity to IBM's institutional superiority complex. [02:00:10] IBM then turned to UMC, which accepted — and deeply regretted it years later. [02:01:43]
The dot-com recession (Q1 2001 through Q3 2003 — approximately three years) caused TSMC revenue to fall sharply; customers who had placed capacity deposits chose not to take their wafers. TSMC "confiscated" no deposits (Morris had authorized the term "confiscate" as a deterrent to over-ordering, knowing it would never actually be used) and instead allowed customers to apply deposits toward future demand. [01:28:44]
ERA 4: The Succession Crisis and Return of Morris Chang (2005–2009)
In 2005, Morris stepped back as CEO while retaining the chairmanship, handing the CEO role to a designated successor (Rick Tsai) as a deliberate succession development experiment. [00:10:15]
Rick Tsai split TSMC's operations into two divisions: advanced technology (led by Mark Liu) and mainstream technology (led by C.C. Wei), a structural decision Morris considered inappropriate for a foundry business with a shared customer base. [00:52:45] (An identical divisional experiment had been tried by American president Don Brooks in 1996; Morris had resisted it, brought in McKinsey — at a cost of a couple of million dollars over two months — and McKinsey recommended the functional structure, citing Boeing as a comparable.) [00:56:11]
The 40-nanometer yield crisis developed under Rick Tsai's tenure: a manufacturing/quality problem caused NVIDIA — then the largest customer on that node — to bear the brunt of defective output. TSMC's quality director argued TSMC was not at fault; the CEO offered NVIDIA nothing in compensation. [00:11:23]
The 2008 financial crisis triggered Rick Tsai's knee-jerk response: using performance review rankings to lay off approximately 600–700 employees — without calling the special board meeting that Morris had explicitly requested be held before any such action. [00:15:38] Morris viewed the terminations as punishing employees for subjective performance ratings, a methodology he considered unfair, non-credible, and counterproductive given semiconductor industry talent economics (separation costs ~0.5 years of salary; retraining ~0.5 years; net benefit negative if rehires needed within a year). [00:19:40]
The laid-off employees demonstrated in front of Morris's private home, requiring 50–60 police officers to manage the crowd. On the second round of protests, 25 employees camped overnight in a nearby park. [00:21:22] Sophie Chang (Morris's wife), without being asked, went to a neighborhood market at 6 AM, bought Chinese breakfast — fried bread, buns, soybean milk — for 25–30 people, and distributed it to the protestors. They thanked her and agreed not to march to the presidential palace that day. [00:23:44]
In 2009, Morris retook the CEO position, offered to rehire all laid-off employees who wished to return, moved Rick Tsai to head of New Businesses (solar and LED — businesses that ultimately failed: China subsidized solar into commoditization; LED was blocked by entrenched patent holders), and began addressing several simultaneous crises: the NVIDIA dispute, pricing that was dropping faster than manufacturing cost reductions, and the need for a unified business development organization. [00:12:45]
ERA 5: The 28-Nanometer Bet and the Smartphone Supercycle (2009–2013)
One of Morris's first structural decisions upon returning was to lock TSMC's R&D budget at 8% of revenue unconditionally — ending the annual budget negotiation that had previously stressed the R&D director and allowed arbitrary cuts during downturns. [00:41:21] The R&D director later described this as the single most important management decision of his career.
After resolving the 40nm crisis (see below), Morris's R&D team told him: "28-nanometer is going to be the sweet spot." Morris invoked Shakespeare — "There is a tide in the affairs of men which, taken at the flood, leads on to fortune" — and committed fully, tripling annual CapEx from approximately $2–2.5 billion to nearly $6 billion in 2010. [00:43:04]
The TSMC board (majority independent directors from the US and UK) strongly objected to the tripled CapEx, calling the general counsel before the board meeting even occurred. Morris listened, then told them: "You need to let me go ahead with this one." They accepted. [00:49:52]
The 28nm bet coincided with the smartphone era — specifically Apple's iPhone and the rise of Qualcomm's mobile modem and application processor business — creating demand that validated and exceeded the investment thesis. [00:51:12]
The NVIDIA $100M+ settlement (2009): Within 4–5 weeks of retaking the CEO role, Morris spent nearly half his working hours analyzing the NVIDIA dispute. He assembled all available intelligence — including visibility into what NVIDIA's own downstream customers were demanding from NVIDIA — to construct a number he believed was fair to both parties. He emailed Jensen: "I'll be at your home at 6 PM. Let's have pizza and salad. At 8 PM sharp we'll discuss business." [00:32:05] Over dinner with Jensen and his wife Lori, Morris checked his watch at 8:00 PM and moved to Jensen's home office. He presented an offer exceeding $100 million, made it non-negotiable with a 48-hour expiry (vs. the alternative of going to arbitration, which Jensen had previously requested but which Morris wanted to avoid), and Jensen accepted within two days. [00:35:23] The relationship has since generated many billions of additional business for TSMC.
In reorganizing the business development function, Morris offered the combined BD/marketing leadership role to Mark Liu (who managed 10,000 employees) and explained it as a career-broadening move. Mark declined: "You want me to take a job with only 60–70 people?" Morris then offered it to C.C. Wei, who accepted enthusiastically. [01:05:29] C.C. Wei is today TSMC's Chairman and CEO.
ERA 6: Winning Apple and the 20nm Detour (2010–2014)
Morris spent months "strategizing" on how to win Apple's business before the opportunity arrived unexpectedly through family connections. His wife Sophie mentioned that her second cousin Terry Gou (founder of Foxconn / Hon Hai) was coming to dinner and bringing "a vice-president from Apple." [01:07:20]
The Apple visitor was Jeff Williams, Apple's Chief Operating Officer (not merely a VP). Dinner conversation was approximately 80% Jeff, 20% Morris. Jeff made clear Apple wanted TSMC to foundry its chip wafers. He mentioned TSMC could expect 40% gross margins — a figure Morris did not respond to, since TSMC was already at 45% and pushing toward 50%. [01:14:27]
The unexpected twist: when Morris raised 28-nanometer, Apple said no — they wanted 20-nanometer, a half-step node that TSMC had not planned to develop. Twenty-nanometer was a detour from the planned 28→16nm roadmap, and TSMC's R&D at that time lacked the capacity to develop two nodes simultaneously. Pursuing 20nm meant delaying 16nm. [01:18:11]
After the $6 billion CapEx commitment to 28nm, funding 20nm for Apple required additional capital. Morris consulted Goldman Sachs (with whom he had cultivated a relationship since TSMC's ADR listing; he served on Goldman's board), evaluated options (cutting dividend, issuing new equity, or borrowing), and chose borrowing via corporate bonds, while only accepting half of Apple's originally requested volume — because one-third of TSMC's shareholders were meaningfully dependent on dividends, and eliminating them (as Jeff Williams suggested) would have triggered a stock selloff. [01:24:57]
When C.C. Wei told Apple's purchasing team that TSMC would take only half the demand, they said: "You must be crazy." When Morris conveyed the same decision to Jeff Williams in Cupertino, Jeff accepted it. The only pushback: Jeff suggested eliminating the dividend. Morris refused. [01:31:34] The conversation ended there.
The Samsung shock: While TSMC built out 20nm infrastructure for Apple, Samsung — having been excluded from the 20nm business — sprinted ahead on 16-nanometer (which TSMC was forced to delay due to the 20nm detour). Apple placed its first 16nm orders with Samsung. [01:52:06] Morris immediately emailed Jeff Williams: "We invested in all this equipment and were counting on you for 16nm — and now you're buying from Samsung." Jeff replied: "Don't worry. I'll be in Hsinchu next week." Jeff explained that as soon as TSMC was ready with its 16nm, Apple would buy all its needs from TSMC. [01:54:17] TSMC developed its 16nm approximately six months later, and recovered the majority of Apple's 16nm business.
On the question of value capture (gross margin split): Morris noted TSMC today earns ~55–57% gross margins, while customers like Apple and NVIDIA earn 70–80% gross margins. Morris was direct: "I don't have the privilege of sorting this out anymore — C.C. does." [01:49:32] He acknowledged there is no formula; it is personal, relationship-based, and constrained by market realities.
ERA 7: The Ecosystem Matures, Moore's Law Continues (2014–Retirement and Beyond)
The 28nm node confirmed TSMC's foundry leadership position; subsequent nodes (16nm, 7nm, 5nm, 3nm/N3E/N3P, 2nm in production ramp as of the interview's recording) each extended the gap. [01:15:55]
TSMC's gross margins eventually crossed 50% after Morris's retirement — propelled by technology leadership and COVID-era demand — achieving the goal Morris had pursued for years. [01:15:23]
The 2nm fab construction at Hsinchu (visible to Ben and David during their visit) is proceeding in four phases; Phase 1 was complete at the time of recording, with small-volume production runs beginning in H2 2025. [02:32:30] Each phase covers the equivalent of many football fields.
Morris noted that TSMC's ADR (American Depositary Receipt) currently trades at a 20% premium to the underlying Taiwan-listed shares, with board permission required to convert ordinary shares to ADR — a structural feature preventing arbitrage. [01:22:29]
3. The Acquired Playbook & Themes
Power Context (Hamilton Helmer's 7 Powers)
Scale Economies — TSMC's primary and most powerful moat. The learning curve means that the highest-volume foundry achieves the lowest unit cost per wafer, enabling lower prices, which attracts more customers, driving more volume. At current fab construction costs of ~$20 billion per facility (projected to reach $40–80–100 billion in future nodes), the number of players capable of building and sustaining a competitive leading-edge fab approaches one. David and Ben argue this is a natural monopoly structure baked into the economics of the industry. [02:02:18]
Process Power — TSMC's 8%-of-revenue R&D commitment has generated proprietary manufacturing processes — including CoWoS (Chip-on-Wafer-on-Substrate) advanced packaging for AI chips, and analogous mobile chip packaging technologies — that are not available elsewhere and create substantial switching costs. The refusal to co-develop with IBM (which would have transferred process IP and subordinated TSMC's engineering culture) was an explicit defense of this power. [00:41:21]
Counter-Positioning — The pure-play foundry model is inherently counter-positioned against IDMs like Intel. Intel's business model requires designing, marketing, and manufacturing its own chips; entering the foundry business at scale would require Intel to serve customers who are its direct product competitors. This "poisoned gift" structure prevented Intel from replicating TSMC for decades. As Morris notes: "Intel just does not know how to be a foundry" — and Tim Cook's explanation was that Intel lacks the customer-service culture, not merely the technology. [02:22:15]
Cornered Resource — The Hsinchu Science Park ecosystem — with co-located Synopsys, Cadence, ARM, Qualcomm, MediaTek, two semiconductor PhD programs, and ASML (extreme ultraviolet lithography equipment) proximity — represents a geographically concentrated talent and knowledge resource that cannot be airlifted or rapidly recreated. Ben and David describe this as the most successful government-funded industrial initiative in history. Arizona and other international fabs will not replicate this density. [02:27:43]
Switching Costs — As TSMC's proprietary packaging technologies (CoWoS for AI chips; mobile equivalents) become required components of leading-edge chip products, the cost to customers of dual-sourcing increases. Morris's anecdote about requiring customer deposits — with the threat of confiscation for non-performance — is an early-stage example of contractual switching costs built into the business model from year one. [01:25:51]
Network Effects — Secondary; the more fabless chip designers standardize on TSMC's process design kits (PDKs), design libraries, and EDA tool integrations (jointly developed with Synopsys and Cadence), the more new chip designers default to TSMC as the manufacturable platform, creating a soft network effect among the fabless design community. [02:29:10]
Branding — Significant at the B2B level. TSMC's brand among fabless customers is synonymous with "we will not compete with you." Every customer who has experienced Intel's arrogance (as described by Morris — "none of them liked Intel") or Samsung's conflict of interest (Samsung competes with Apple in the consumer market while manufacturing Apple chips) understands the TSMC brand premium. [02:24:02]
Value Creation vs. Value Capture
TSMC creates extraordinary value for the entire technology ecosystem — enabling the iPhone, AI chips, automotive electronics, IoT, and every category of modern computing — while capturing a structurally constrained fraction of that value. Morris's own admission is revealing: TSMC earns 55–57% gross margins while customers like Apple and NVIDIA earn 70–80%. The chip designers capture more of the value they help create because (a) their IP (software, architecture, brand) is more defensible than manufacturing; (b) TSMC historically operated near cost-plus pricing to maximize volume and learning curve descent; and (c) customers have negotiating leverage, particularly large ones like Apple who can credibly threaten to dual-source. [01:49:15]
However, at the trillion-dollar market cap level, TSMC has captured enormous absolute value — suggesting that while the margin percentage is lower than its customers', the underlying revenue scale (and capital asset base) makes the total value capture extremely large in dollar terms. The 8% R&D rule and continuous CapEx investment represent value re-invested into future competitive position rather than extracted as margin. [02:14:18]
The Bull and Bear Case
Bull Case:
Moore's Law demand (spiritual, if not technical) is undefeated: the world continues to demand 2x compute capacity every two years, creating persistent tailwind. [02:44:16]
Leading-edge fab construction costs are reaching levels ($20B+, heading to $40–100B) that effectively foreclose new entrants; TSMC's scale advantage compounds with each node. [02:41:56]
Proprietary packaging (CoWoS) and process IP create defensible switching costs that go beyond raw lithography capability. [02:31:15]
The Hsinchu ecosystem is a decades-in-the-making resource that geopolitical competitors (including CHIPS Act-funded US fabs) will struggle to replicate meaningfully at the leading edge. [02:30:48]
Every AI compute demand increase (NVIDIA GPUs, Google TPUs, Apple Neural Engines, Amazon Trainium/Inferentia) flows through TSMC's fabs. [02:32:12]
Management culture — no-layoff philosophy during downturns, trust-first customer relationships — reduces talent attrition and customer churn in ways that compound over decades. [00:16:51]
Bear Case:
Geopolitical concentration risk in Taiwan; TSMC's trillion-dollar market cap is physically instantiated in facilities that sit 100 miles from mainland China, and no international fab can replicate leading-edge capability in the near term. [02:27:50]
Arizona and other international fabs are not leading-edge; they represent diluted versions of TSMC's competitive position and drain engineering talent from Taiwan's ecosystem. [02:30:15]
Samsung remains a competitive threat, as evidenced by the 16nm episode — TSMC's detours or missteps create windows for Samsung to leapfrog nodes. [01:51:39]
Intel Foundry Services, though currently behind, has multi-billion-dollar government subsidies (CHIPS Act) and is pursuing an IDM2.0 strategy; the threat is structural even if current execution is poor. [01:37:28]
The transition from planar transistors to 3D packaging and chiplets may reduce the direct comparability of process node labels, creating potential for new competitive frameworks to emerge where TSMC's current advantages are less decisive. [02:45:25]
As AI training compute shifts to increasingly custom silicon (Google, Amazon, Meta all building in-house), TSMC becomes more dependent on fewer very large customers, increasing customer concentration risk. [02:30:08]
1. The Learning Curve / Experience CurveOrigin: BCG's Bruce Henderson and Bill Bain, refined with Morris Chang at Texas Instruments in the early 1970s. [02:07:06]
Application: As cumulative production volume doubles, the cost per unit falls by a predictable percentage. The strategic implication — first explained by Henderson and Bain, absorbed by Morris — is that the market leader who achieves the highest volume wins the cost race, which compounds into a pricing advantage, which crowds out competitors. At TSMC, this became the operating thesis for pricing new nodes below current cost (sometimes unprofitably) to acquire volume and descend the curve faster than any competitor, knowing that at mature scale, margins would be recovered. The 28nm CapEx tripling and the willingness to take only half of Apple's 20nm demand both reflect learning curve calculus — get enough volume to matter on the curve, but not so much that the company's solvency is threatened if the forecast is wrong. [02:02:18]
2. The "Tide in the Affairs of Men" — Strategic Timing (Shakespeare)Origin: Shakespeare's Julius Caesar, quoted by Morris in his autobiography and in this interview. [00:46:29]
Application: Morris explicitly invokes this to describe how he identified 28-nanometer and later 7-nanometer as "sweet spot" nodes — moments where the intersection of technology readiness, manufacturing capability, and market demand created a brief window for decisive, outsized capital commitment. The framework is anti-hedging: waiting for certainty means missing the tide. The learning curve then locks in the position taken at the flood.
3. The Pure-Play Foundry / Non-Compete ArchitectureOrigin: Morris Chang's original business plan, 1987. [02:22:15]
Application: TSMC's founding thesis — do not design chips, do not sell end-products, do not compete with any customer at any point in the value chain — eliminated the central reason chip designers were reluctant to outsource manufacturing (fear that their IP would be appropriated by a vertically integrated competitor). This structural trust asymmetry between TSMC and IDMs like Intel is, in David and Ben's analysis, the single most important reason TSMC became the dominant foundry. Tim Cook's explanation ("Intel just does not know how to be a foundry") is interpreted by Morris as primarily a customer-service and culture argument — not a technology one. [01:43:55]
4. Deposit + Confiscation as Commitment DeviceOrigin: Morris Chang, early TSMC history (1990s). [01:26:12]
Application: Because TSMC bore all the CapEx risk of capacity expansion based on customer forecasts, customers had an incentive to over-order (upside: guaranteed capacity; downside: none, since they could simply not take the wafers). Morris introduced required deposits — and authorized salespeople to tell customers that TSMC would "confiscate" the deposit if wafers were not taken. The word "confiscate" was deliberately aggressive and unprecedented in the industry; it reset the psychological contract with customers who had previously treated TSMC's capacity as an option rather than a commitment. In practice, TSMC never confiscated; they allowed deposits to be applied to future demand. The credible threat was sufficient. [01:27:00]
5. The Anti-Performance-Review Layoff PrincipleOrigin: Morris Chang, Texas Instruments, early 1970s, refined at TSMC. [00:18:42]
Application: Performance reviews are inherently subjective (each is given by one supervisor; aggregating them to identify the "700 worst performers" is not a rigorous process). Laying off based on performance reviews is not credible to employees, damages morale company-wide, and in a Moore's Law industry — where talent needs are cyclical and the downturn recovery often arrives within a year — typically destroys more value than it saves (separation cost + retraining cost typically exceeds the savings if re-hiring occurs within 12 months). Morris's formula: no layoffs; use furloughs, probation, or transfers; preserve the workforce for the inevitable recovery. [00:19:40]
6. The "Plant Seeds Early" Institutional Relationship StrategyOrigin: Morris Chang; illustrated by the Goldman Sachs and customer relationship stories. [01:21:11]
Application: Morris cultivated Goldman Sachs as an investment banking partner early in TSMC's life — arranging TSMC's ADR listing through them — not because he needed them immediately, but because he "knew that one of these days we will probably need top-level investment bank advice." When the Apple 20nm financing question arose ($10B+ of additional CapEx), Goldman's advice was immediately available. The same logic applied to customer relationships: Morris was "always on a plane" visiting current top-15 customers and prospecting for the next top-15. The Jensen relationship — pizza and salad dinners over years before any crisis — created the social capital that enabled a $100M+ settlement at a single dinner. [01:04:46]
7. The Kissinger Principle — Small Teams, High LeverageOrigin: Morris Chang, invoking Henry Kissinger as National Security Advisor. [01:04:04]
Application: When Morris offered Mark Liu the combined business development/marketing role (60–70 people) rather than keeping him in operational leadership (10,000 people), he framed it using the Kissinger analogy: Kissinger had a couple hundred staff as NSA, while the Secretary of State had thousands worldwide — yet Kissinger had more power. Strategic leverage is not proportional to headcount. Mark Liu rejected this framing; C.C. Wei accepted it. The result: C.C. Wei became Chairman and CEO of TSMC, with business development experience as a key part of his executive formation. [01:05:29]
8. The Functional vs. Divisional Structure DebateOrigin: Morris Chang, tested against McKinsey's analysis (1996), confirmed twice (1996, 2005–2009). [00:56:11]
Application: The foundry business has a fundamentally shared customer base across all fabs and all nodes. Dividing it by fab (as Don Brooks proposed) or by node generation (as Rick Tsai implemented) creates redundant customer touchpoints, confuses strategic marketing priorities, and destroys the economies of unified operations. The functional structure (one operations organization, one business development organization) maintains strategic coherence. Boeing was McKinsey's cited comparable — not divided by airplane model, only by commercial vs. government. A foundry divided by fab would be like Boeing creating a separate division for each aircraft variant. [00:57:08]
6. Anecdotes & Lore
1. The Breakfast That Prevented a Presidential Protest
During the 2008–2009 employee protest crisis, approximately 25 laid-off workers camped overnight in a small park near Morris's home after two rounds of demonstrations. At 6 AM the next morning, Sophie Chang — without being asked — took a bodyguard to the neighborhood market, bought Chinese fried bread (youtiao), buns, and soybean milk for 25–30 people, and distributed breakfast to the protestors in the park. They thanked her and announced they would not march to the presidential mansion that day. [00:23:44] This act of human decency — unrehearsed, unstrategic, and entirely Sophie's initiative — is credited by Morris as a partial catalyst for his decision to retake the CEO role.
2. "Quiet — Morris Chang Is Calling Me"
In 1997, Morris received Jensen Huang's letter in Hsinchu. During a US trip the following week, Morris called Jensen cold from the stationery's phone number. Jensen was in the middle of a heated argument with his staff. The moment Morris said "this is Morris Chang," Jensen shouted to his people: "Quiet. Morris Chang is calling me." [00:07:01] Within days, they met in person. Morris was struck by Jensen's articulateness, frank admission of NVIDIA's financial difficulties, and his bold prediction that a single chip would not only save the company but make it one of TSMC's top customers. Both predictions came true within three years.
3. The Pizza-and-Salad Diplomacy
Morris resolved the >$100 million 40nm yield dispute with NVIDIA through a carefully choreographed home dinner. He emailed Jensen with a specific schedule: arrive 6:30 PM, eat pizza and salad, move to the study at 8:00 PM sharp for business. [00:32:05] Jensen's wife Lori made the salad; the pizza may have been homemade or delivered. At 8:00 PM, Morris checked his watch and moved to the study. He presented one number — more than $100 million — gave a 48-hour non-negotiable deadline, and Jensen accepted within two days. [00:35:23] David noted: "You probably don't want to go to arbitration with your best customer."
4. Terry Gou's Surprise Dinner Invitation
The Apple relationship began not through formal business development but through a second-cousin connection between Morris's wife Sophie and Terry Gou, the founder of Foxconn (Hon Hai). [01:07:20] Sophie greeted Morris at the door after a board meeting and mentioned Terry was "coming to dinner and bringing a vice-president from Apple." Sophie didn't understand the business significance. Morris suspected the visitor would not be "just an ordinary vice-president." He was right: Jeff Williams was Apple's COO. Dinner was home-cooked Chinese food. Jeff Williams dispensed with small talk almost immediately and began his pitch for TSMC to foundry Apple's chip wafers.
5. The McKinsey Confirmation That Didn't Convince
In 1996, American president Don Brooks wanted to restructure TSMC along divisional lines. Morris disagreed but, rather than simply overruling him, offered: "Why don't we get McKinsey?" After two months and roughly two million dollars, McKinsey recommended the functional structure, citing Boeing as an example. [00:56:13] Brooks then said: "Tell me one big company that's functionalized." McKinsey immediately answered: Boeing. Brooks pointed out Boeing has commercial and government divisions. Morris's retort: "They don't have a 707 division and a 747 division. If we divide by fab, it would be like Boeing dividing by airplane model." [00:57:19] Brooks was not fully persuaded. Morris allowed Rick Tsai to repeat the same experiment nine years later in 2005 — and then unwound it in 2009.
6. The IBM Co-Development Refusal
After Qualcomm defected from IBM Semiconductor to TSMC (~1997–1998), Morris immediately predicted that IBM would approach TSMC for a co-development partnership on 0.13-micron technology, positioning itself as the "senior partner." [01:58:49] When that approach came in 1999, Morris declined without deliberation. His reasoning: co-development with IBM would require TSMC engineers to work at IBM facilities in a different culture, subordinate TSMC's process IP development to IBM's institutional hierarchy, and potentially destroy TSMC's ability to independently develop future generations. [02:00:10] IBM was "quite angry" — it considered TSMC a small Taiwan company beneath its dignity. IBM then approached UMC, which accepted — and UMC "regretted seriously their acceptance a few years later."
7. The Samsung Shock
After TSMC invested tens of billions of dollars to serve Apple's 20nm demand (despite only taking half the volume), Samsung — excluded from the 20nm node — sprinted ahead on 16nm. Apple placed its first 16nm orders with Samsung. [01:52:06] Morris's description: "I got a shock. It was a real shock." He emailed Jeff Williams immediately: "We invested so much…we were counting on you to take the 16 from us." Jeff replied by email that he would be in Hsinchu the following week to explain. [01:54:17] In person, Jeff assured Morris that as soon as TSMC's 16nm was ready, Apple would buy all its needs from TSMC — and that is precisely what happened, approximately six months later.
8. The Gordie Campbell Fabless Epiphany
Before founding TSMC, while still at General Instrument, Morris was approached by Gordie Campbell — a serial entrepreneur with a strong reputation — asking for a $50 million investment and offering only a mental business plan. [02:16:35] Three weeks later, no written plan had arrived. When Morris called, Campbell explained he no longer needed $50 million — only $5 million — because he had decided not to build a fab. This single insight — that a chip company could exist without owning manufacturing — was the foundational signal that convinced Morris the fabless model was coming. [02:18:21] Almost simultaneously, Atmel's founder approached General Instrument to use its spare fab capacity, providing a second data point.
7. References & Recommendations
Books & Publications
Morris Chang's Autobiography (Volume 2) — Published in Traditional Chinese in early 2025 after a 26-year gap from Volume 1. Contains the Apple, NVIDIA, and IBM stories discussed in this episode. Not yet published in English. Translation work funded by Tyler Cowen via Emergent Ventures; translator Karina Bao. [00:01:52]
Shakespeare, Julius Caesar — "There is a tide in the affairs of men..." — Quoted by Morris in his autobiography and in the interview to describe his decision to commit to 28nm and 7nm nodes. [00:46:29]
Michael Porter's Competitive Strategy (and related books) — Morris mentions having all of Porter's books (~700 pages each). Porter served on TSMC's board. Morris considers Porter's original competitive strategy memo (reportedly ~20 pages) to be among the best business writing ever. [02:04:58]
Bruce Henderson / BCG — Experience Curve — BCG's foundational framework for cost-based competitive strategy. Morris worked directly with Henderson and Bill Bain at TI in the early 1970s to apply the experience curve to semiconductor pricing. [02:04:16]
Competing Companies & Industry Peers
Intel — Primary counter-example throughout the episode. Cited for: arrogance toward Taiwan-based customers ("none of them liked Intel"), inability to serve as a foundry, competition for Apple's iPhone chip manufacturing (2011), and historical dominance in the x86 era that paradoxically prevented it from transitioning to the foundry model. [01:43:55]
Samsung — Cited for: winning Apple's first 16nm orders (a temporary TSMC setback due to the 20nm detour); serving as Apple's mobile processor manufacturer before TSMC; representing the ongoing foundry competitive threat at leading-edge nodes. [01:51:39]
UMC (United Microelectronics Corporation) — TSMC's Taiwanese peer and competitor. Accepted the IBM co-development partnership that TSMC declined in 1999; "regretted seriously" the decision years later. By 1999, UMC was already smaller than TSMC. [02:01:43]
IBM Semiconductor — Major Qualcomm foundry partner until ~1997–1998 when Qualcomm switched to TSMC, triggering IBM's decline as an independent foundry player and its eventual exit from the semiconductor manufacturing business. [01:58:52]
MediaTek — Taiwan-based fabless chip designer; Hsinchu Science Park co-resident; the former TSMC CEO (Rick Tsai) is now Vice-Chairman and CEO of MediaTek. [00:28:58]
Foxconn / Hon Hai — Terry Gou's manufacturing conglomerate; key Apple supply chain partner; intermediary through which Apple's Jeff Williams was introduced to Morris Chang. [01:07:55]
ARM Holdings — Ben and David argue that the rise of ARM architecture was a necessary enabling condition for TSMC's success. ARM, TSMC, Synopsys, Cadence, and ASML were all founded within a few years of each other in the mid-to-late 1980s. [02:24:50]
Synopsys & Cadence — EDA tool companies co-located in Hsinchu Science Park; tightly integrated with TSMC's process design kits (PDKs). [02:27:43]
ASML — Extreme ultraviolet (EUV) lithography equipment manufacturer; co-founded approximately the same era as TSMC. [02:27:34]
Qualcomm — Major TSMC customer (top 5) from the late 1990s onward; key player in the 40nm node crisis; anchor customer for both 28nm and subsequent nodes. [00:13:34]
Key Operators, Investors & Advisors
Jensen Huang — NVIDIA founder and CEO. Facilitated this interview by asking Morris Chang on Ben and David's behalf. His 1997 letter to Morris inaugurated TSMC's most strategically important customer relationship. [00:01:13]
Jeff Williams — Apple COO; primary counterpart for Apple's TSMC foundry negotiation from 2010 onward. Personally negotiated pricing in Cupertino and accepted TSMC's half-volume offer. [01:11:36]
Tim Cook — Apple CEO; met Morris Chang in March 2011 when Jeff Williams was unavailable. Told Morris: "Intel just does not know how to be a foundry." [01:43:05]
Paul Otellini — Intel CEO who approached Apple at "the highest level" in early 2011 to compete for iPhone chip foundry business. [01:41:31]
Terry Gou — Founder of Foxconn / Hon Hai; second cousin of Sophie Chang (Morris's wife); introduced Jeff Williams to Morris. [01:07:20]
C.C. Wei — Current TSMC Chairman and CEO; appointed by Morris as head of unified Business Development in 2009. [01:05:38]
Mark Liu — Former TSMC Co-CEO and Chairman; declined Morris's offer to lead Business Development; headed the advanced technology operations division. [01:03:28]
Rick Tsai — TSMC CEO from 2005–2009; implemented the controversial divisional org structure and the performance-review layoff during the 2008 crisis. [01:05:07]
Don Brooks — First American president of TSMC (~1996); wanted divisional org structure. [00:54:08]
Irwin Jacobs — Founder of Qualcomm; founded slightly before TSMC (early-to-mid 1980s). [01:56:32]
Gordie Campbell — Widely credited as the first fabless chip entrepreneur; his conversation with Morris was the primary signal that fabless was coming. [02:16:46]
Bruce Henderson — Founder of BCG; presented the experience curve theory to TI's CEO Mark Shepherd in ~1970. [02:04:16]
Bill Bain — BCG partner assigned to TI for ~2 years to develop semiconductor learning curve applications; subsequently founded Bain & Company. [02:07:06]
Karina Bao — Translator of Morris Chang's memoir Volume 2 into English. [00:02:18]
Tyler Cowen — Economist at George Mason University; funded the memoir translation via Emergent Ventures. [00:02:24]
Aart de Geus — Co-founder and Executive Chairman of Synopsys; provided prep guidance for the interview. [02:50:38]
Rene Haas — CEO of ARM Holdings; provided prep guidance. [02:50:59]
Sir Peter Bonfield — Current TSMC board member; provided prep input. [02:51:06]
Wally Rhines — Former CEO of Mentor Graphics; TI contemporary of Morris Chang. [02:51:13]
Arvind Navaratnam — Worldly Partners; wrote an analytical research report on TSMC. [02:52:42]
Jon (Asianometry) — YouTube channel on semiconductors; hosted Ben and David in Taipei. [02:52:01]
Tim Culpan — Former Bloomberg journalist; Culpium Substack; provided topic guidance. [02:52:34]
Geopolitical & Macro Events
Taiwan Semiconductor Science Parks (Hsinchu) — Government-initiated industrial parks housing TSMC and its entire ecosystem. [02:27:43]
CHIPS Act (US) — Referenced implicitly through discussion of Arizona TSMC fabs and US government pressure. [02:30:15]
China's subsidization of solar — Morris cites China's aggressive solar industry subsidization as the reason TSMC's solar business failed. [00:27:43]
The 2008–2009 Global Financial Crisis — Triggered TSMC's internal succession crisis and Morris's return. [00:17:13]
The Dot-Com Recession (2001–2003) — Caused a ~3-year TSMC revenue depression. [01:28:44]
8. The Bottomline (by AI)
TSMC's story is ultimately about the compounding power of two interlocking commitments made before they were obviously correct: the structural commitment to never compete with customers (enabling trust at an industry level that Intel could never replicate), and the strategic commitment to the learning curve (accepting present losses to own the volume that drives future cost advantages). Morris Chang made both commitments in 1987 based on pattern recognition from Texas Instruments, not because the outcome was foreseeable — TSMC's first decade was built largely on other people's discarded orders. The lesson for founders and investors is that durable, trillion-dollar positions are almost always built by players who choose a lane that existing incumbents cannot credibly enter (counter-positioning), then invest relentlessly in the volume flywheel before the market arrives to validate the thesis. In the AI era, the same logic applies to every infrastructure layer: whoever accumulates the most volume, process IP, and ecosystem lock-in before the supercycle peaks will find the economics compounding in their favor for decades — and the players who tried to hedge by competing on every front will find themselves unable to earn the trust that a pure-play position commands.
Jul 16, 2026
Dr. Robert Wachter | A Giant Leap: How AI Is Transforming Healthcare... | 14 Jul 2026 | Talks at Google
"don't get me wrong US healthcare delivers miracles every day particularly when it comes to cutting edge and intensive care... but the health care system itself is a headache wrapped in red tape inside the nightmare that France Kofka himse…