AI summary by TwinMind
The Early Years at TSMC
Asianometry
16m 37s
TSMC's Founding and Initial Investment (1987)
- TSMC opened February 1987 with factory and staff from Taiwan's Industrial Technology Research Institute
- Taiwanese government invested ~$100 million at start
- Founder Dr. Morris Chang: government never invested another dollar after initial funding
- New unproven business model: only semiconductor company solely dedicated to manufacturing
- No existing market for foundry services at founding
Historical Context and Precedents
- Custom chip concept not original to TSMC
- Fairchild Semiconductor 1960s: micro mosaic logic chip with ~150 logic gates
- Aluminum interconnects customized at end of fabbing process
- Used computer for placement and routing (casually invented electronic design automation)
- LSI Logic founded by Fairchild team members for custom chips
- Jensen Huang worked at LSI Logic before founding Nvidia
- March 1976: Morris Chang proposed foundry idea to Texas Instruments
- Suggested TI aggregate outside customer demand to strengthen manufacturing capability
- TI rejected idea - made no sense to open fabs to potential competitors
- TI went opposite direction: started making consumer electronics (competing with own customers)
Critical Technological Foundation
- Morris Chang: foundry concept impossible without Carver Mead and Lynn Conway's revolutionary work
- Their contributions: abstraction hierarchies, systems/subsystems approach, common block reuse
- Enabled splitting of chip design from manufacturing
Early Leadership Structure
- Morris Chang: Chairman (not CEO initially)
- Asian companies: chairman more active role, corporate affairs representative, not daily operations
- James E. Dykes: First CEO (March 1987)
- Background: General Electric VP, managed all GE semiconductor work
- Integrated RCA factories after GE 1985 acquisition
- After 2 years closing RCA factories, was "put out to pasture"
- Taiwan government wanted foreign investor and executive
- Initially refused Chang's offer, convinced after in-person meeting in North Carolina
- Oversaw construction of first actual fab (Fab 2)
- Completed contract, returned to US
- Died Florida 2013, age 75, remembered for kindness and love of hamburgers
Early Survival Strategy (1987-1991)
- No fabless industry existed yet
- Most people didn't care about "newfangled company" in "backwards Taiwan"
- Survived by providing extra capacity in trailing edge products
- First customer likely Intel (Morris Chang knew Robert Noyce, Gordon Moore)
Intel Partnership Development (1988)
- Andy Grove visited Taiwan to see Fab 1 and 3-micrometer node
- Chang showed good yields on 3-micron technology (2.5 generations behind Intel/TI)
- Grove impressed, said "maybe Intel can use you"
- Taiwan fabs had superior yields since RCA technology transfer days
- Intel accreditation phase: found 200 issues, TSMC resolved one by one
- Certified to produce processors and Intel chipsets
- Other American IDMs struck deals for popular products lacking capacity
- TI-TSMC 256k dynamic RAM production deal 1988 (leading edge was 1 megabit)
Fab 1 Operations and Innovation
- Capacity: ~1,200 150mm wafers/month
- Emoji sticker system for equipment status:
- Smiling: machine working fine, actively used
- Sleeping: working fine, not currently used (not making money)
- Hurt face: down for repair, needs attention
- Skeptical (raised eyebrow): not right, engineer should investigate
- Goal: all machines showing happy sticker
- Psychological gaming to address non-smiley stickers immediately
Fab 2 Development (1990)
- Dedicated 1990
- 500 nanometer node
- Capacity: 3,000-4,000 150mm wafers/month
- Fab 1 closed 2001 (only fab TSMC ever closed)
- Technically correct claim of never closing fab since they didn't build Fab 1
Manufacturing Secret: Cycle Time Focus
- Cycle time: duration for wafer lot (25+ wafers) through entire fabrication (600-1000 steps)
- 1989 Dykes interview: plan to outmanufacture Japanese
- Target: 20-25 days input to output
- Much faster than Japanese (promised 40 days, actually ran 55 days)
- Japanese weakness: cycle time
- Peter Thiel concept: every great business built around hidden secret
- TSMC's secret: cycle time focus equal to yield focus
- Shorter cycle time = faster wafer processing = faster yield learning
- 1998 Morris Chang handwrote top priorities: "short cycle time from wafer start to finish" (#4 priority)
Competitive Advantages vs Japanese (Early 1990s)
- Japanese had capacity and technology for foundry services
- Example: Xilinx fabbed first products at Seiko Corporation
- Japanese demanded concessions:
- Seiko got exclusivity in Japan market for Xilinx products
- Wanted product information about chips being fabbed
- TSMC position (Dykes): don't compete with customers, no incentive to steal designs
- Unlike TI or Japanese, TSMC enshrined non-compete value in company charter
Leadership Transition: Don Brooks Era (1991-1997)
- Don Brooks: third president, replaced Klaus Weimer (who went to Chartered Semiconductor Singapore)
- Background: 22 years at TI, associate of Morris Chang
- Called Chang "one of top semiconductor managers in 70s and 80s"
- Praised Asian talent pool: "trained loyal workforce and financial means"
- Taiwan described as "homogeneous, like US 1950s, team effort"
- Social pressure: if one person doesn't pull weight, jeopardizes others' bonuses/profit sharing
Workforce Economics and Quality
- Lower base salaries allowed 2x more engineers per task vs West
- Later: engineers/workers in profit sharing earned TSMC stock worth hundreds of thousands annually
- Low wages not primary advantage (China fabs didn't have superior economics)
- Quality, retention, education critical factors
- Talent pool: two top Taiwan universities nearby, graduates fed into TSMC
- 1987: 32,000 of 130,000 foreigners getting advanced US degrees were from Taiwan
- Taiwanese familiar with West, spoke passable English
Industry Transformation (1991+)
- Rise of fabless industry
- Semiconductor fabs increasingly expensive: new line cost $200 million
- Squeezed startups and middle-class semiconductor makers ($100M-$1B revenue)
- Adding new line risked overcapacity
- Foundry model: always able to tap capacity interested in customer success
- Prestigious customers (Intel, TI) brought Taiwanese design houses to TSMC
- Design houses: small companies, little capital, plenty of chip ideas (video games to computers)
- Fabless industry growth (Taiwan, then US) fueled TSMC better than second-source supplier role
Morris Chang's Initial Skepticism
- Brooks and Dykes compelled by foundry idea from beginning
- Morris Chang most negative despite recruiting them
- 1991 Brooks joining: Chang said building Fab 2 but "wouldn't anticipate Fab 3, don't get your hopes up"
- Brooks: "when he speaks of being visionary, I don't think he had any idea what he was getting started"
- Chang still working at Industrial Technology Research Institute until 1994
- Also chairman of Vanguard Semiconductor (another foundry) until 2003
- Saw TSMC only once monthly, not involved in day-to-day operations
Brooks' Vision and Results
- Saw incredible yields at fabs and economic power of scaling
- American semiconductor industry experience: understood customer pains
- Ignored Chang's pessimism, got to work
- TSMC grew 54% annually during Brooks' 6-year presidency
- Chang's conservative nature: "always told me business was going to crap out, you're over investing"
- "But he didn't get in my way when it came down to doing it"
- Chang surprised by success in 95-96, seized opportunity to become president for rewards
Brooks' Departure and Aftermath
- Wanted to return to US, originally supposed to work at US office
- Chang changed mind, Brooks resigned
- Joined TSMC rival UMC, "ruffled feathers" with old colleagues
- Died Texas 2013, age 74
- Recalled departure with "wistful bitterness"
Philosophical Reflection
- Morris Chang read Ecclesiastes 9:11 at Harvard
- "Race is not to the swift or battle to the strong... but time and chance happened to them all"
- Business histories make TSMC seem like overnight success
- Reality: Chang founded TSMC on just one idea among many
- Taiwan already becoming successful semiconductor power
- Success relied on mix of hard-working Taiwanese, American-educated Taiwanese, foreign semiconductor veterans
- Climbed value chain, carved out brand new industry niche
- "Time and chance happens to us all - TSMC came at right time, had right chance, made most of it"
