AI summary by TwinMind
How Apple Changed TSMC
Asianometry
12m 5s
Apple's Semiconductor Strategy Origins
- Apple = American company, products assembled China/India, supply chain heart in Taiwan
- 5 top electronic suppliers in Taiwan, including 2 closest: Foxconn and TSMC
- Foxconn = one of several assembly partners (alongside Pegatron, Wistrom)
- TSMC = sole supplier of device semiconductors, most critical component
- 2008: Apple acquired PA Semi for $276M to design system-on-chips
- 2010: Steve Jobs unveiled A4 chip in first iPad, first custom semiconductor
Samsung Foundry Relationship & Departure
- First A4 chip fabbed at Samsung foundry
- Samsung provided microprocessors for iPhone, 3G, 3GS + screens
- Samsung's mobile phone entry chilled relationship
- Apple sued Samsung for IP theft
- Apple hated Samsung using component profits to compete in mobile phones
- Sought supply chain diversification
TSMC Partnership Formation
- TSMC business principle: never compete with customers
- Sept 2014: iPhone 6s A8 processor on TSMC 20nm process
- Parties talking since 2010 (Jeff Williams & Morris Chang dinner in Taiwan)
- Took several years to finalize deal
- Exception: A9 dual-sourced from TSMC + Samsung
- Media alleged Samsung chips had worse battery life
- Apple claimed real-life differences negligible
- Single sourcing for every Apple processor since
TSMC's Extraordinary Accommodations for Apple
Financial Investment
- TSMC invested $9B, deployed 6,000 people in Tainan for Apple-specific fab
- Comparison: SMIC raised $7.6B in 2020 Shanghai IPO
- Samsung announced 10-year $116B foundry plan
- TSMC spends $16-17B capex in 2020 alone
R&D Strategy Alignment
- TSMC crafted entire R&D/tech rollout to match Apple's annual iPhone cycle
- Previously: TSMC rolled out new nodes when ready
- Apple needs new, better, high-yielding process annually
- TSMC cannot afford to miss annual cycle = "TSMC equivalent of Moore's Law"
Half-Step Process Strategy
- Intel's tick-tock strategy failed (stuck at 14nm for 4 years)
- TSMC created incremental 10-20% performance/power improvement yearly
- vs. ambitious jumps every 2 years
- Example: A12 on N7 (7nm), A13 on N7P (half-step improvement)
- Note: nanometer measurements = marketing terms, not physical dimensions
EUV Technology Adoption
- TSMC progressively adopted EUV technology step-by-step
- Different approach from Intel/Samsung
- First foundry to ship full EUV chips at high yield (2020)
- Reliability/predictability more important than risky new tech
N3 Node Strategy
- N3 (next after N5) won't use Gate All Around (GAA) technique
- Reason: Apple cannot afford risk of technique failing annual iPhone cycle
- N5 used for A14, N5P for A15 (next year)
Business Impact & Revenue
- Apple = TSMC's single biggest customer
- 2019: Apple represented 23% of TSMC annual revenue (~$8.2B)
- Huawei = second largest customer ($5B revenue)
- 2014: Major revenue jump visible when partnership began
- 2020: TSMC celebrated 1B chips shipped on 7nm process (2 years old)
Premium Pricing Model
- Apple one of few companies wanting/affording cutting-edge processes
- TSMC 5nm wafer costs $17,000 each
- Each chip costs $20-30 with good yield
- For $1,000 phone, $20-30 chip cost not significant
- Traditional customers (MediaTek, Qualcomm) sell chips directly, cost-sensitive
- Samsung's cheaper prices attractive to cost-conscious customers
Current Market Position
- Huawei no longer able to book TSMC orders
- Apple now unquestionably TSMC's most important customer
- Apple booked TSMC's entire N5 capacity
- Relationship secure and mutually beneficial
- No legitimate substitute for TSMC's capabilities
- TSMC needs Apple's revenue for cutting-edge R&D investment
Why Apple Won't Buy TSMC
Financial Constraints
- TSMC market cap: $389B (9th largest company globally)
- Similar size to Tesla, Visa, Nvidia
- Apple cash on hand: $190B
- Apple debt: $100B (taken against cash for stock buybacks)
Business Model Destruction
- TSMC's value = neutral third party that doesn't compete with customers
- Apple ownership would destroy this value proposition
- Would eliminate $400B+ investment value
Geopolitical Barriers
- TSMC = Taiwan's flagship company (Taiwan in company name)
- Taiwanese government would block transaction (even from US ally)
- China tried buying TSMC for years, failed
- TSMC represents 50% of Taiwan's entire R&D spending
- Sale would cause political uproar, heads would roll
- Control not for sale at any price
