AAcquired
← All episodes
Morris Chang26 January 2025

TSMC Founder Morris Chang

5Frameworks
14Insights

Frameworks in this episode

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Hot Take· 1

Hot Take26:30

The Ironic Failure of TSMC's Solar and LED Diversification

Around 2009, TSMC bet on solar and LED as 'new businesses' outside its core semiconductor manufacturing — both failed, while the core business became a $600B+ market.

  • China's subsidies collapsed global solar cell pricing, killing TSMC's solar ambitions
  • LED patents were tightly controlled by a few incumbent companies, blocking entry
  • The executive placed in charge of 'new businesses' eventually left to become vice chairman/CEO of MediaTek
#diversification#strategy mistakes#solar#LED

Explainer· 1

Explainer2:24:00

Hosts' Analysis: How ARM Enabled the Fabless Revolution

In their post-interview analysis, Ben and David argue that ARM's rise as a licensable CPU architecture was a key enabler of TSMC's pure-play model, since an x86-dominated world would likely have stayed vertically integrated.

  • Without ARM as a viable alternative to x86, there may have been no viable path for a standalone foundry industry to emerge
  • Apple's shift to its own ARM-based silicon meant Intel didn't automatically get to manufacture iPhone chips
  • TSMC, ARM, Synopsys, Cadence, and ASML were all founded within a few years of each other in the mid-to-late 1980s

“if arm hadn't become such a viable CPU architecture platform and apple hadn't standardized you know their apple silicon on arm probably Intel would be…”

David Rosenthal · 2:26:00
#ARM#fabless industry#semiconductor ecosystem

Story· 8

Story03:30

Jensen Huang's Letter and TSMC's First Call to Nvidia

In 1997, a young, nearly-bankrupt Nvidia sent TSMC founder Morris Chang a letter after being ignored by TSMC's San Jose office, kicking off a partnership that would become one of TSMC's largest.

  • Nvidia was 4 years old with ~50-60 employees when it wrote to Chang directly
  • TSMC was already a billion-dollar company but Chang insisted no future customer be ignored
  • Chang called Jensen Huang without advance notice and visited Nvidia the next day
  • Huang predicted the chip TSMC would fab would make Nvidia a major TSMC customer — it did
#Nvidia#TSMC history#customer relationships
Story09:30

The 40nm Yield Crisis Strains the Nvidia Relationship

A manufacturing and quality problem on TSMC's 40nm node in 2009 caused major delays and losses for top customers like Nvidia, ultimately prompting Chang to retake the CEO role.

  • The prior CEO's quality team insisted TSMC wasn't at fault, so Nvidia was offered nothing
  • Nvidia was one of the largest customers affected by the yield problem
  • Chang retook the CEO position in 2009 partly to resolve this and other priority issues

“it was a you problem and everybody was suffering from it and of course Nvidia at that time was the perhaps the biggest customer of…”

Morris Chang · 11:00
#yield problems#Nvidia#crisis management
Story21:00

700 Layoffs Spark Protests at Chang's Home

When TSMC's prior CEO laid off 600-700 employees under the guise of performance management, protesters gathered outside Chang's home for months, escalating until he retook the CEO role.

  • 50-60 police officers were dispatched to manage over 100 protesters at Chang's home
  • Protesters later camped overnight in a nearby park
  • Chang's wife brought them breakfast, and they called off a planned march on the presidential palace

“the laid-off employees as I said there were 700 of them six or 700 of them uh came to my home uh to demonstrate and…”

Morris Chang · 21:00
#layoffs#company culture#crisis
Story32:00

Pizza, Salad, and a $100M Handshake With Jensen

Chang resolved the long-running Nvidia dispute over a casual family dinner at Huang's home, delivering a nine-figure offer with a 48-hour deadline.

  • Chang scheduled the offer discussion for after a normal pizza-and-salad family dinner
  • The offer was over $100 million with a strict 48-hour acceptance window
  • Huang accepted within two days, and the partnership has generated many billions since

“I'm coming to Silicon Valley next week on this stage I will be at your home at 6:00 let's have just Salah and pizza”

Morris Chang · 32:00
#Nvidia#negotiation#customer relationships
Story1:06:30

A Mysterious Dinner Guest: Terry Gou Brings Apple's COO

TSMC's path to Apple began with an unannounced dinner visit from Chang's wife's cousin, Foxconn founder Terry Gou, who brought Apple COO Jeff Williams to pitch TSMC directly.

  • Terry Gou is a second cousin of Chang's wife Sophie and was Foxconn's chairman
  • Jeff Williams, Apple's COO (not just 'a vice president'), came to pitch wafer capacity
  • Williams did roughly 80% of the talking and proposed an initial 40% gross margin, which Chang didn't answer that night

“uh uh Terry G called in the afternoon and said he was coming to dinner”

Morris Chang · 1:07:00
#Apple#Foxconn#Terry Gou#origin story
Story1:51:00

The Shock of Apple's First 16nm Order Going to Samsung

After investing tens of billions on the 20nm detour for Apple, Chang was blindsided to learn Apple placed its first 16nm orders with Samsung — until Jeff Williams personally reassured him TSMC would get the bulk of the business once ready.

  • TSMC had converted most of its 20nm-related equipment expecting to carry Apple's 16nm volume too
  • Samsung skipped 20nm and went straight to 16nm, undercutting TSMC's timing
  • Jeff Williams flew to San Jose to reassure Chang, and TSMC ultimately won most of Apple's 16nm business

“I got a shock I mean it was a real shock when I heard that Apple had placed the first orders of 16 with Samson”

Morris Chang · 1:52:00
#Apple#Samsung#competition
Story1:59:00

Declining IBM and Watching UMC Regret Saying Yes

When IBM, losing Qualcomm's foundry business, asked TSMC to co-develop its next process node as a junior partner, Chang refused instantly — IBM went to rival UMC instead, who came to regret accepting the deal.

  • Chang inferred IBM's semiconductor business was struggling once Qualcomm started shifting volume to TSMC
  • IBM wanted TSMC engineers to work under IBM's process, risking TSMC's technological independence
  • UMC accepted the co-development deal with IBM and reportedly regretted it seriously

“IBM in fact was uh quite uh Angry you know I mean they thought we we still a small Taiwan uh backward place”

Morris Chang · 2:01:00

“so they uh immediately went to uh uh UMC and UMC accepted them only to regret seriously their acceptance”

Morris Chang · 2:01:30
#IBM#UMC#Qualcomm#competitive intelligence
Story2:05:00

Bill Bain's 'World Imperative' and the Learning Curve's Origins

Chang recounts how BCG's Bruce Henderson and a young Bill Bain embedded at Texas Instruments to apply the experience curve to semiconductors — until Bain abruptly left to found his own firm, Bain & Company.

  • Bill Bain worked out of an office next to Chang's at TI for roughly two years, refining learning curve applications together
  • Bain reported back to Bruce Henderson weekly before eventually deciding to leave BCG
  • Bain told Chang he was leaving to start his own firm because of a 'world imperative'

“you are the first one I tell this to outside the Boston Consulting Group I am leaving B concerning group to start my own consulting…”

Morris Chang · 2:11:30

“but there is the world imperative that's the first time I heard that term”

Morris Chang · 2:12:00
#learning curve#Bain & Company#BCG#Texas Instruments

Q&A· 1

Q&A1:14:00

Apple's Opening Pitch: 40% Margins and a Surprise Node Request

Apple's COO Jeff Williams proposed a 40% gross margin (below TSMC's existing 45%) and, surprisingly, asked for a 20nm node rather than TSMC's planned 28nm-to-16nm path.

  • TSMC's margin was already at 45% and targeting 50%, above Williams' proposed 40%
  • Apple's request for 20nm was a 'detour' TSMC hadn't planned to develop
  • TSMC lacked the R&D resources at the time to pursue two nodes (20nm and 16nm) simultaneously

“now that was a surprise to me and Frank it was also a disappointment because the most law progression after 28 was going to be…”

Morris Chang · 1:18:00
#Apple#pricing#technology roadmap

Tool· 1

Tool1:23:00

Financing a Bet-the-Company Apple Order: Bonds Over Dividends

To fund the capex needed for Apple's order without diluting shareholders, TSMC chose to issue corporate bonds rather than cut its dividend or sell new stock — and only committed to half of Apple's requested volume.

  • TSMC's financial planning, aided by a longstanding Goldman Sachs relationship, ruled out cutting the dividend, which a third of shareholders cared deeply about
  • TSMC agreed to supply only half of what Apple initially requested
  • Apple's COO suggested eliminating the dividend, which Chang rejected outright

“we decided not to cut dividend we decided not to sell new stock we decided to just borrow”

Morris Chang · 1:25:00
#capital allocation#Apple#corporate finance

Takeaway· 2

Takeaway2:27:00

Why Hsinchu Science Park Can't Be Replicated

Visiting TSMC's home ecosystem in person, the hosts describe Hsinchu Science Park as a tightly integrated cluster of suppliers, customers, and universities that took decades of government-backed investment to build and can't be quickly copied elsewhere.

  • Cadence, Synopsys, ARM, Qualcomm, and MediaTek all have offices within the same science park as TSMC
  • Two universities sit inside the park, feeding graduates directly into the ecosystem
  • The hosts call it possibly the most successful government-funded industrial policy initiative ever

“even if you wanted to you couldn't airlift tsmc and this capability out of Taiwan and recreate it somewhere else”

Ben Gilbert · 2:27:30

“this has got to be the single most successful government funded industry initiative of all time”

David Rosenthal · 2:30:00
#Hsinchu Science Park#industrial policy#ecosystem effects
Takeaway2:43:00

Moore's Law as a Self-Reinforcing Demand Engine

The hosts close by reflecting that the entire semiconductor industry, and TSMC's rise, rests on the world's persistent demand for roughly double the compute every two years — a demand-side force that has outlasted the original technical definition of Moore's Law.

  • TSMC's addressable market grew from $26B in 1987 to $527B last year
  • The demand side of computing has proven durable even as the technical pace of transistor scaling has slowed
  • Continuous global demand for more compute creates market incentives that keep driving the supply side forward

“but like spiritually the world demands roughly 2x the computing power that it had two years ago every two years and that has continued for…”

David Rosenthal · 2:44:00
#Moore's Law#semiconductor market growth#demand trends