0. BACUS Panel Discussion: The $6 Million Mask Set — Truth and Consequences
Klaus-Dieter Rinnen
Chief Analyst and Director
Semiconductor & Electronics Manufacturing
Gartner Dataquest
BACUS Panel Discussion
At SPIE Microlithography 2003
Santa Clara, 24 February 2003

1. Semiconductor Revenue Forecast: Recovery Gains Momentum in 2003
Billions of Dollars
19%
-6% 2%
28% 9% 1%

2. Semiconductor Revenue Forecast: Cycles Are Nothing New
Revenue Growth

3. Burning Questions for the Future of the Industry
Show Me the Money?
Big Three?
The Red Brick Wall?
Silicon to Carbon?
Declining ROI?
Robotics?
Next “Killer Application?”
China: Friend or Foe?
Privacy?
When Is the Upturn?
From High-Tech to Appliance?
Commoditization of Silicon?
Deflation?
Security?
Zero Inventory?
Virtual Integration?
End-Market Saturation
Limits to outsourcing?
Real Time?
Consolidation: Impact?
Terrorism?
Slowing Moore’s Law?

4. Key Long-Term Trends in Semiconductor Industry
Crisis of confidence:
Where is the “killer application”?
Convergence of applications
Serving a “demand-pull” market
The “M” word:
Is the industry maturing?
Is the semiconductor industry CAGR slowing down?
Pervasiveness of semiconductors
Commoditization of silicon
Big three — Consolidation and concentration of capital
Structural changes — Super IDM
Hype cycle — China, new technologies, applications: Friend or Foe?
Moore’s Law — Slowing down? Where is the limit? What is next?
Escalating cost for fab, design, masks: Raising barriers to entry?

5. What Is the Demand-Pull Market?
Buying center is shifting from a high-margin corporate buyer to a low-margin consumer buyer
Lower price point, lower premium for speed
Easy-to-use human functionality most important
Product cycles fast and fickle
Cost pressures
Market buys technology when cost is lower
Money is wasted (lower profit) if technology comes to the market too fast
Industry reacts:
Restructuring and consolidation
Speeding up ITRS to drive down cost and protect margins (BUT, industry passes through cost savings!)

6. Technology Cycles Are Speeding Up
International Technology Road Map for Semiconductors
Changes to Technology Timing, 1997, 1999 and 2001
Semiconductor Process Technology Cycle Compression
Technology
Node
The Technology Gap!?
Technology Leaders’ Road Map
Technology Leaders

7. Impact on Suppliers Time-collapsed development cycles
Reduced prospects for ROI
Delayed ROI because of technology gap
Less time means higher development cost per year and higher overall cost, as speed is achieved with more money
Increased risk to develop product that will never come to market
Decreased product maturity equates to customer dissatisfaction and higher support/manufacturing cost
Spread-out technology adoption makes for slower manufacturing learning and increased cost
Extreme price pressure
Question: Do I really want to serve this market?

8. In the Age of Subwavelength Lithography, Photomasks Are the Enablers
Source: Numerical Technologies

9. ITRS 2001: Leading-Edge SOC Design Cost
Source: International SEMATECH
Note: Costs called out are for 8M gate PDA in 2001

10. Stepper ASPs: Aggregate
Millions of Dollars
Growth

11. Photomask Set Cost in 2002 Thousands of Dollars 0.5 mm 18 Layers
Note: Set cost is for 2002, best estimate.

12. Photomask Set Cost for Second Year of Production
Thousands of Dollars
0.5 mm
18 Layers
0.35 mm
20 Layers
0.25 mm
22 Layers
0.18 mm
28 Layers
0.13 mm
32 layers
Note: Set cost is for 2002, best estimate.

13. Photomask Set Cost for Second Year of Production, Extended
Thousands of Dollars
0.5 m
18 L
0.35 m
20 L
0.25 m
22 L
0.18 m
28 L
0.13 m
32 L
90 nm
65 nm
45 nm
30 nm
Note: Set cost is for 2002, best estimate.

14. Photomask Set Cost for Second Year of Production, Extended
Thousands of Dollars
0.5 m
18 L
0.35 m
20 L
0.25 m
22 L
0.18 m
28 L
0.13 m
32 L
90 nm
65 nm
45 nm
30 nm
Note: Set cost is for 2002, best estimate.

15. Semiconductor vs. Photomask Revenue
Semiconductor Revenue
(Billions of Dollars)
Photomask Revenue (Percentage of Semiconductor Revenue) 12

16. Summary Points Hype about photomask cost is off the charts
Cost of a mask set is dependent on number of layers, distribution of critical dimension layers, photomask technologies involved (PSM, OPC) and maturity of manufacturing tools. If you want a set for 32nm design today, $6 million will not be enough, if it could be done
For mainstream production, the “$6 million mask set” will not happen, as it cannot happen. Gartner Dataquest projects photomask set costs to remain below $6 million, even at 32nm in the second year of production
Wanted! New business models for photomask and ASIC maker
There are myriad possibilities (in design, material, equipment process, usage) to reduce photomask cost. However, where are the margins for the photomask maker to secure its own and the industry’s future? This is not a nonprofit business.
Photomask expenditure as a percentage of semiconductor revenue is not declining but also not appreciating, despite all the hype about photomask prices. Maybe photomask prices are too low to ensure options for an economically viable future?