140 likes | 598 Views
Outsourcing and Offshoring in the Semiconductor Industry. David A. Hodges Robert C. Leachman Competitive Semiconductor Manufacturing Program UC Berkeley Sloan Industry Centers Annual Conference Atlanta, GA April 19-21, 2004.
E N D
Outsourcing and Offshoring in the Semiconductor Industry David A. Hodges Robert C. Leachman Competitive Semiconductor Manufacturing Program UC Berkeley Sloan Industry Centers Annual Conference Atlanta, GA April 19-21, 2004
U.S. Integrated Device Manufacturers(e.g. Texas Inst., Motorola, Intel, …) Labor-intensive chip assembly work mostly off-shored since the 1960s • Initially, plants served just one company • More recently, independent assemblers and testing firms are serving multiple customers • IBM automated in the 1960s • Automation of assembly and testing now spreading industry-wide and world-wide
U.S. IDMs, 1960-1990 • Capital-intensive wafer fabs were off-shored selectively: important aid market access • Cost of direct labor not a significant factor • US ownership, international professional staff • Hazards: weak infrastructure, long supply lines, business and political climate • Early examples: Texas Instruments (Japan), Analog Devices (Ireland), Intel (Israel)
U.S. IDMs, 1960-1990 • Skills-intensive process development and product design mostly remained in the US • Firms sought advantages from proprietary technologies • Few skilled professionals available abroad • Some exceptions: Chip design centers in England (TI), Israel (Intel); typically devoted to specific products for worldwide markets • Sales, marketing, customer support efforts carried on world-wide
Changing business models:IDMs forced to become specialists • Intel, AMD: microprocessors • Samsung, NEC, Micron, Infineon: memory • Texas Inst., STM: chips for cell phones • These are standard products, MM units; same designs purchased by many competing original equipment manufacturers (OEMs) • Above categories represent about ½ of total worldwide semiconductor production • What about the other half?
Factors leading to “foundries” • Competitive modern wafer fabs cost $2-4B • employ ~ 1000 people (total for 7 x 24 operation) • Annual revenues > ½ fab cost for profitability • Worldwide standardization of mfg. process • Innovative design firms require only a fraction of one fab’s capacity • Vastly different management skills: design vs. fab • IDMs rarely succeed in serving fabless firms • Foundries were established to serve this need • Leadership of Morris Chang!
Fabless-foundry business model • Fabless firms define, design, & market chips • small investment, quick response • $300-500K revenue/employee • ~50,000 well-paid U.S. jobs; ~13,000 ROW • Asian foundries fabricate chips for many firms • huge investments; fixed costs ~75% of total • ~15,000 factory jobs, well-paid by local scales • highly automated for tight process control • short production cycle • timely intro of new technology generations • excellent customer service • some niche specialists with old technology
Outsourcing, Offshoring? • Fabless design centered in the U.S. • MS, PhD grads of top U.S. universities • U.S. is #1 (78% of ‘03 revenues) • Taiwan is #2 (11% of ’03 revenues) • Equivalent design skills very rare elsewhere • Most silicon foundries are in Asia • Many process development jobs in Asia • Many grads of top US universities • Weak U.S. domestic investment (except Intel)
“Food chain” for semic. industry • Semiconductor production equipment & raw materials are supplied mainly from U.S., Japan, and Europe • U.S. leads in key areas: • MS & PhD education • Computer-aided design for semiconductors • University-industry cooperation • Climate for innovation • Market for advanced technology • Government support is strongest in Asia
Factors influencing location for manufacturing investments • Trophy value of semiconductor fabs (Think about the steel industry in the 1960s) • Trophy sought by gov’ts worldwide: tax incentives! • China is the current leader in incentives • Most capital comes from outside PRC • Fading concerns about investment risks • Weaker controls on U.S. equipment export • Commodity status of manufacturing technology • Return of expatriates; spread of higher education • Protected IP less important than know-how • Improving infrastructure in China, other nations
Chinese competition for foundry business • Semiconductor Manufacturing Int’l Corp. (SMIC) • largest, most advanced Chinese foundry • founded in 2002; 3 8” fabs in Shanghai • purchased Motorola’s 8” Tianjin facility • 12” fab in Beijing under construction • 3/17/04: $1.8B IPO in HK & NY; -12% as of 4/6/04 • U.S. filed WTO complaint re: China’s lower VAT for locally designed or manufactured semiconductors • China remains far behind in chip design capability • China establishes unique domestic standard for cellular telephony; Chinese partners required
2003 Foundry revenue leaders • 1. TSMC (Taiwan) $5.9 billion + 26% • 2. UMC (Taiwan) 2.7 + 27% • 3. Chartered (Singapore) .73 + 49% • 4. IBM (U.S.-IDM) .56 - 27% • 5. NEC (Japan-IDM) .43 + 33% • 6. SMIC (China) .37 +630% • 7. Hynix (Korea-IDM) .34 + 39% • 8. DongbuAnam (Korea) .33 + 27% • 9. Jazz (U.S. ex-Rockwell) .19 + 16% • 10. HHNEC (China) .17 + 13% • 11. SSMC (Singapore) .16 + 82% • 12. X Fab (E. Germany) .13 + 27%
Survival strategies of U.S. IDMs • Intel: heavy investments; try new markets • Texas Inst: limit investments + use foundries • IBM: partnered with Chartered, Infineon • AMD: more German incentives in Dresden • Micron: innovation; more cost reductions • Motorola: divesting semiconductor business • National: product focus; use foundries • Analog Devices: limit investments + foundries
Conclusions for semiconductor industry • It’s a fully globalized industry • Microprocessors: Intel unchallenged • Memory is a commodity; Samsung leads by far • IDM business model is dead for other products • U.S. leads in innovative chip design • U.S. unchallenged in design software • design & software skills are bound to spread! • Asia leads in foundry manufacturing • U.S. is not a serious competitor; poor ROI • TSMC, UMC are likely to remain leaders • Overcapacity looms; SMIC payoff is uncertain