<<

Proprietary & Confidential

The Future of Industry – A Foundry's Perspective –

Dr. F.C. Tseng Vice Chairman TSMC January 24, 2008

© 2007 TSMC, Ltd Proprietary & Confidential Outline

z Semiconductor Market Outlook z Challenges and Solutions

„ Economic „ Technology z Summary

2

© 2007 TSMC, Ltd Proprietary Semiconductor Market Will Continue to & Confidential Grow

Year Growth

60% ’02 +1%

40% ’03 +18% 21% 16% ’04 +28% 20% ~8% 5% ’05 +7% Growth 0% ’06 +9%

-20% ’07 +4% ’08E +7~9% -40% ‘70 ‘75 ‘80 ‘85 ‘90 ‘95 ‘00 ‘05 ‘10 ‘15 Year Source: SIA

3

© 2007 TSMC, Ltd Proprietary Steady Market Expansion & Confidential

Increasing semiconductor penetration in electronics

Data Processing 20.9% 27.2% 28.8%

Communications 13.5% 16.4% 18.8%

Consumer 12.6% 16.3% 19.5% Electronics

‘01 ‘05 ‘10E Year

Source: WSTS, IC Insights, TSMC estimates

4

© 2007 TSMC, Ltd Increasing Semiconductor Content in Proprietary & Confidential Systems – Functionalities and performance

LCD Mobile Automotive Headset

? ? ? 1.3x$95B $1,025B $77B 1.8x $61B 10x $33B $98B LC LC FH u LC Co Luxur In 4G 2. 3G HD 5G te D D D m llig D p y a e c n t

Source: Gartner, iSuppli, Strategy Analytics t

5

© 2007 TSMC, Ltd Proprietary New Applications as Growth Drivers & Confidential

100

1 Consumer PC Electronics End User (3C Convergence) 1/100

Units / Minicomputer

1/10000 Mainframe

‘70 ‘80 ‘90 ‘00 ‘10 Year

6

© 2007 TSMC, Ltd Demand from High-end and Low-end Proprietary & Confidential Markets

More Functionality 9 Voice 9 Video 9 Voice 9 Messenger Functionality 9 SMS 9 GPS 9 Voice 9 WAP 9 Wi-Fi 9 Camera 9 PDA 9 MP3 ‘90 ‘00‘ ‘05 Year

9 Voice

9 Voice Lower

Price Price

7

© 2007 TSMC, Ltd Proprietary In Search of A PC to Serve “The Next & Confidential Billion” z The low-cost PCs for the developing nations:

INTEL One Laptop Per Child “ClassMate” “Eee” “XO”

8

© 2007 TSMC, Ltd Proprietary Growth Opportunities are Global & Confidential

Product Focus Population (millions) Feature-rich & connected Advanced 800

/ e r tu e a c Emerging e n 1,500 F a m / m u e r tr ic fo c r r e P e p P S Developing (Base of the Pyramid) 4,000 Basic functions

Source: “The fortune at the bottom of the pyramid” by C. K. Prahalad, and IMF

9

© 2007 TSMC, Ltd Proprietary Plenty of Opportunity Ahead & Confidential

265

255 US Semi / Person = ~$250 245

11 65

10 55

9 45

8 35 Semi $ / Person 7 25 China & India Semi / Person = ~$20 6 15

World Population (Billion) 5 5

4 -5 ‘80 ‘85 ‘90 ‘95 ‘00 ‘05 ‘10 World Population Worldwide average semiconductor $ per person

Source: US Census Bureau; IC-Insights; TSMC estimates

10

© 2007 TSMC, Ltd Proprietary & Confidential Outline

z Semiconductor Market Outlook z Challenges and Solutions

„ Economic „ Technology z Summary

11

© 2007 TSMC, Ltd Proprietary & Confidential Challenges

Economic Technology

1. Huge CapEx 4. Nanometer Manufacturing 2. ROI Risk – Process 5. Nanometer Design 3. ROI Risk – Product 6. Design Complexity

12

© 2007 TSMC, Ltd Proprietary 1. Huge CapEx & Confidential

z Capital expenditure for constructing a new fab is rapidly increasing

„ Major factor for financing and future profit

15

$8B ? (GigaFab with 10 100K WPM)

$3B 5 (30K WPM) $0.4B $1B (20K WPM) Capital cost of fab ($B) (20K WPM) 0 6-inch 8-inch 12-inch 18-inch

13

© 2007 TSMC, Ltd Solution – Foundry Based Business Proprietary Model & Confidential

z Many IDMs are changing to either fab-lite of fabless z A wide variety of consolidation and collaboration are inevitable

PastPast PresentPresent FutureFuture Design Product IP Fabless IP Design Focused Vendor Fabless Product Vendor Service IDM Foundry Few Manufacturer IDM Foundries Getting Few Assembly Distribution Fab-lite Assembly IDMs & Test & Test OEM

Vertically IDMs in Transition, Few IDMs & Foundries, Integrated Fabless Getting Bigger Many Fabless

14

© 2007 TSMC, Ltd 2. ROI Risk – Process Proprietary & Confidential z Incremental challenges in developing next generation processes z Process technology development costs are continuously increasing 2.0 ?

1.5 $1.4B $1.1B $0.9B

lopment cost ($B) 1.0 e $0.7B $0.5B 0.5

0

Technology dev 0.13um 90nm 65nm 45nm 32nm 22nm Development Cost Ramp-up Cost 15

© 2007 TSMC, Ltd Proprietary 2. ROI Risk – Process (Cont’d) & Confidential z Yield ramp-up requires significant time and investment

In-house manufacturing Foundry has the economy of scale with limited products 100% 0.13um 90% 90nm 0.25µm 65nm 80% 0.18µm 0.13µm 90nm 45nm 70% 65nm 32nm 25%25% Q-Q-to-Qto-Q

Normalized Yield 60% ImprovementImprovement

50% 12Year ’98 ’99 ’00 ’01 ’02 ’03 ’04 ’05 ’06 ’07

16

© 2007 TSMC, Ltd Proprietary Solution I – Technology Alliance & Confidential

z Collaboration is required to overcame ever – increasing financial as well as technical challenges IMEC

Matsushita

17

© 2007 TSMC, Ltd Solution II – Collaboration among Proprietary & Confidential Foundry, IP/EDA Vendors, And Design Service Suppliers z Standardization of IP & EDA tools

IDM

IP & Design Service EDA

Foundries

18

© 2007 TSMC, Ltd Proprietary 3. ROI Risks - Product & Confidential z Design complexity and cost increase rapidly z Short time to market

40M 40 80 $70M

30 27M 60 $48M

20 40 16M Gate Count (M) Design Cost ($M) 10 9M $18M 20 $9.2M

0.13um 90nm 65nm 45nm Gate Count Design Cost

19

© 2007 TSMC, Ltd Proprietary Solution - Product & Confidential

z Optimal system partition z Foundry

„ Design Infrastructure

‹ SPICE

‹ PDK

‹ Foundation IP „ Prototyping

‹ Cybershuttle

‹ MLM

Minimize the product risk and NRE

20

© 2007 TSMC, Ltd Proprietary 4. Nanometer Manufacturing & Confidential

z New materials and device structure

„ High-K gate dielectric „ Metal gate „ 3D FINFET „ Low-K Interconnect z New EDA solutions for technology modeling and advanced lithography

21

© 2007 TSMC, Ltd Solution – Technology Modeling / Proprietary Advanced Lithography & Confidential

z Secure accurate SPICE modeling STI Systematic WPE effect Strained silicon effect Accuracy OPC Random Corner model effect Statistical model z Modeling in new material, 3D device and equipment / topography z Polarization, OPC, double exposure and mask 3D effects

Collaborate with EDA vendors to achieve seamless interfacing hierarchy

22

© 2007 TSMC, Ltd Proprietary 5. Design for Nanometer & Confidential

z Increasing systematic / parameter / random yield loss due to process variation

„ Physical patterning effects, open and short, etc. „ Chemical and mechanical impact, planarity, antenna effect, and via opens, etc. „ Timing, signal integrity and voltage drop z Increasing leakage (quantity, source, variation)

23

© 2007 TSMC, Ltd Solution – Accurate DFM Proprietary & Confidential z Model & simulation-based approach for physical and electrical DFM that represents Manufacturing accurately

„ Physical DFM ‹ Identify hot spot, fix it and improve geometric yield.

„ Electrical DFM ‹ Identify electrical performance deviation, correct it and improve paramedic yield

Yield

CAA/LPC Statistic Design

STA CMP

Time Set-up standard interface between phase and design which is design and layout phase

24

© 2007 TSMC, Ltd Proprietary Solution – DFM Ecosystem & Confidential

z Foundry to set up DFM ecosystem and open license DUF to reduce cost z Standard interface between the design infrastructure and manufacturing

Example:

TSMC/GUC + EDA Alliance IP/Lib Alliance TSMC DCA Alliance

TSMC DFM Compliance Initiative

25

© 2007 TSMC, Ltd Solution – Low Power Proprietary & Confidential z Aggressively develop integrated low power solutions for dynamic and leakage power reduction „ Low power process ‹ Advanced processes with ELK/XLK die-electric: lower voltage, smaller geometry and capacitance for dynamic power reduction ‹ HK/MG and gate CD bias for leakage reduction „ Low power IP ‹ Full set of low power foundation IP, dual power SRAM „ Low power design Reference Flow ‹ Silicon proven design methodologies for TSMC IP and processes ‹ Voltage scaling (DVFS, AVS), power gating with data retention ‹ Low power design automation enabling Develop vertically integrated solution from system-level to layout and to process

26

© 2007 TSMC, Ltd Proprietary 6. Design Complexity & Confidential z Increased design complexity causes longer time-to- market and requires significant effort for verification and software design

Communication Device 4G Mobile Multimedia Device Full HD

Moore’s Law HD 3G

Complexity Current Design Productivity D1

2G

VGA

‘95 ‘00 ‘05 ‘10 ‘15 ‘20 Year 27

© 2007 TSMC, Ltd Main Stream vs. Proprietary ESL Co-development Environment & Confidential

GPrime Platform DDireirectct VectVectoror LocaLocall StStataticic SDRSDRAAMM Memory Interrupt Memory Memory Controller CPCPUU Memory Interrupt Memory Memory Controller AcAccecessss ControlControllerlerControlControllerlerControlControllerler (4(4--CCH)H) APAPBB BrBrididggee Electronic AAMMBBAA M Muultlti-i-lalayyeerr A AHHBB Main Stream AMAMBBAA A APPBB System

DAI KB DAI PWM GPIO KB WDT I2C Timer SSI RTC UART (I2S) PWM GPIO MOUSE WDT I2C Timer SSI RTC UART Level (ESL) (I2S) MOUSE

z Architecture Exploration z Software Profiling z Hardware Profiling ICE z Power Estimation

Network

28

© 2007 TSMC, Ltd Proprietary Solution – ESL & Confidential

z New ESL design technology to address design complexity, and to provide massively parallel heterogeneous MPSOC design z SIP and 3D packaging can integrate large capacity memory and analog circuit with short time to market and low cost z Develop product with chip and PC board at same time to improve success rate

Collaborate with EDA vendors

29

© 2007 TSMC, Ltd Proprietary Summary & Confidential

z Semiconductor market growth will continue but moderate z Future growth opportunities will be global but bifurcate z High ROI risks in design, fab, and technology could be alleviated with the integrated z Close collaboration is required between EDA vendors, foundry and IC companies

30

© 2007 TSMC, Ltd Proprietary & Confidential

Thank You

www.tsmc.com

© 2007 TSMC, Ltd