RoI-Based Investment Model for the Semiconductor Industry RIBIM v0.01 Kenneth Flamm, 2003 Presentation to EMW

Size: px
Start display at page:

Download "RoI-Based Investment Model for the Semiconductor Industry RIBIM v0.01 Kenneth Flamm, 2003 Presentation to EMW"

Transcription

1 RoI-Based Investment Model for the Semiconductor Industry RIBIM v0.01 Kenneth Flamm, 2003 Presentation to EMW Kenneth Flamm Portland, Oregon May 22, 2003

2 Where is this coming from? Conceptual framework for thinking about impact of tech node acceleration on semiconductor price Conceptual framework for thinking about impact of semiconductor prices on electronic system prices Both of above need to be understood to model/forecast semiconductor demand Relation between product sold and semiconductor price (demand function) needs to be understood to say anything reasonable about semiconductor investment

3 Publications now available on first two topics Flamm, The New Economy in Historical Perspective: Evolution of Digital Electronics Technology forthcoming in the Handbook of the New Economy, Academic Press, 2003 Aizcorbe, Flamm, and Khurshid, The Role of Semiconductor Inputs in IT Hardware Price Decline: Computers vs. Communications, Finance and Economics Discussion Series Paper , (Washington: Board of Governors, Federal Reserve Board), June Additonal working papers now under construction. First two pubs are on disk.

4 Input baseline Semi producer capacity Input description of forecast market conditions (demand curve) Input pricing scenario for major semi equip Structured to link to ISMT IEM model if desired Other options possible, this one works with IEM Demo d DRAM variant with DRAM cost model, forecasts Output forecast of product prices, RoI on incremental fab investment in selected product segments Very early prototype, under continuing development Next versions will generate multi-year investment forecast Potential Uses for RoI-Based Investment Model (RIBIM)

5 Applications Forecast tool Strategic planning tool Customer sales tool Validation / testing of corporate strategy Sensitivity to dependence on market assumptions

6 LEM as Test Case Why LEM? Leading edge exceptionally high price changes Cost change dominated by manufacturing tech change Not so for other semi products

7 Not Everything is Leading Edge: Very Different Price Declines! Nominal $$ in 1999 CADR Computers Communicati MOS Micro % 21.2% MPU % 2.5% MPR % 4.0% MCU+DSP % 14.8% Dig Memory % 11.7% MOS Logic % 30.3% Thyristors & Rect % 0.9% Small Sig Trans % 2.4% Pwr Trans % 4.2% Diod & Other Discrete % 1.7% OptoElectronics % 6.0% Dig BiPolar % 0.6% Analog % 20.8%

8 Price Indexes for Semiconductors 1992= MOS MPU MOS Memory MOS MPR Other MOS Logic Thyristors & Rectifiers MOS MCU Power Transistors Small Signal Transistors Optoelectronics Diode & All Other Discrete Digital Bipolar DRAM Analog Source: Flamm, The New Economy in Historical Perspective: Evolution of Digital Electronics Technology in Handbook of the New Economy, (Academic Press), forthcoming.

9 Beyond Moore : More Generalized Functionality Indexes Growth in Quality per Chip, CAGR MOS Micro, of which 91= % 76.4% Microprocessors 91= % 176.2% Microperipherals 91= % 27.6% Microcontrollers 91= % 23.0% MOS Memory, of which 91= % 52.5% DRAM 91= % 69.7% SRAM 91= % 30.1% Flash Memory 91= % 24.1% Other MOS Logic 91= % 13.9% Analog 91= % 13.5%

10 US Semi Mfg var prof margin invest/sales

11 Overview of Model Structure Initial Capacity Demand Curve Investment & Production Costs by Vintage New Invest RoI on New Capex Stop? Equilibrium

12 Caveats Model is roughed out version of preliminary software implementation Not pretty, not elegant Bare functionality implemented All aspects of package to be improved Much greater functionality to be added Kenneth Flamm, 2003

13 Agenda Lock and load Step through model structure Exercises: Change price elasticity Change A&T cost Manually upgrade fabs Add new fabs Increase litho tool cost

14 Demo goes here Run Demo

15 Roadmap: Developing Economic Logic for Fab Investment Given existing capacity, assumed functionality/chip, estimate price/functionality, price/chip at year t from functionality demand curve for year t Calculate variable cost/chip by vintage Take capacity vintage offline if P<VC, recalculate price (capacity not fully utilized) If P>VC and all capacity utilized, consider new investment in greenfield or upgrade Calculate RoI for greenfield and upgrade Select investment with higher RoI, N year out window for calculations If RoI > hurdle rate, add/upgrade a fab Stop when RoI on next investment <= hurdle rate Iterate this procedure for N years

16 Next Step: How much more fab investment before hit hurdle rate for new projects Manual logic to be implemented in automated VBA macro under construction Uses Excel solver Full simulation will have possible additional capacity investments in out years Next generation tool to be finished over summer Release 0.2 of RIBIM to be available from Flamm in Fall 03 Structured so will link to ISMT IEM if user wishes Stay tuned

17 Conclusion Have shown credible RoI calculation methodology Roadmap for improvement and incorporation into broader planning model Decent studies of demand critical to success in projecting industry growth trajectory B&E module of IEM, ISMT transistor density, die size estimates/models have proven value as useful tools for cost side