Transcript view presentation - MIT SDM – System Design and Management

Unmanageable Designs:
What Some Designs Need from
the Economy and How They Get
It
Carliss Y. Baldwin
Harvard Business School
2005 SDM Conference
October 27, 2005
Slide 1
© Carliss Y. Baldwin and Kim B. Clark, 2005
Three Points
 Designs
“need” to become real
– They become real by creating the perception of “value”
 Designs
act as a financial force
– Perception of value = Incentive to invest
– In making the design— “use value”
– In making the design better— “option value”
 Modular
Designs with Option Potential
– Create hurricane-type forces
– Will change their economic “space”
– Unmanageable and dangerous (unless you understand
them)
Slide 2
© Carliss Y. Baldwin and Kim B. Clark, 2005
Metaphor— “Selfish Designs”
 “Selfish”
designs want to become real
 Their tool is human motivation
– A user perceives use-value => willingness to make or
willingness-to-pay
– Designers and producers estimate willingness to pay
– The result is an asset => financial value
 Humans
move mountains for financial value
– Value operates “as a force” in the economy
– Some designs have so much of this force that we work
for them…
Slide 3
© Carliss Y. Baldwin and Kim B. Clark, 2005
Strong value forces can change the
shape of an industry

Andy Grove described a vertical-to-horizontal transition in
the computer industry:
“Vertical Silos”
“Modular Cluster”
Slide 4
© Carliss Y. Baldwin and Kim B. Clark, 2005
Andy’s Movie
Stack View in 1980
Services
Systems Integration
Top 10 Public
Companies in
US Computer
Industry
Area reflects
Market Value
in Constant
US $
Slide 5
Applications Layer
Middleware Layer
Operating Systems
Hardware
IBM
S
P
E
R
R
Y
D CVC
U H E
N P C
S
Y
XRC
S
AMP
Components
TI
Intel
© Carliss Y. Baldwin and Kim B. Clark, 2005
Andy’s Movie
Stack View in 1995
IBM
S
P
E
R
R
Y
D CVC
U H E
N P C
S
Y
XRC
S
AMP
TI
Services
Systems Integration
First Data
EDS
Oracle
Intel
Top 10 Public
Companies in
US Computer
Industry
Area reflects
Market Value
in Constant
US $
Slide 6
I
Applications LayerB MSFT
Middleware LayerM
Operating Systems
Hardware: Printers
Hardware: Servers
Hardware: Routers
Components
CA
HP
IBM
Cisco
Intel
Micron
© Carliss Y. Baldwin and Kim B. Clark, 2005
Andy’s Movie—the Sequel
Stack View in 2002
IBM
S
P
E
R
R
Y
D CVC
U H E
N P C
S
Y
XRC
S
AMP
TI
Intel
Top 10 Public
Companies in
US Computer
Industry
Area reflects
Market Value
in Constant
US $
Slide 7
Services
Services
First First
Data Data
ADP EDS
Systems Integration
Systems Integration
I
CA
Applications LayerB MSFT
Applications
LayerIBM
Middleware
LayerM
Operating Systems
Middleware
LayerPrinters
Hardware:
Hardware: Servers
Operating
SystemsRouters
Hardware:
Components
Hardware: Printers HP
Hardware: PCs Dell
Hardware: Servers IBM
Hardware: Routers Cisco
Components
Intel
Oracle
Oracle
MSFT
HP
IBM
Cisco
Intel
Micron
TI
© Carliss Y. Baldwin and Kim B. Clark, 2005
Turbulence in the Stack
Departures from Top 10:
 Xerox (~ bankrupt)
 DEC (bought)
 Sperry (bought)
 Unisys (marginal)
 AMP (bought)
 Computervision (LBO)
Sic Transit Gloria Mundi
Slide 8
Arrivals to Top 10:
 Microsoft
 Cisco
 Oracle
 Dell
 ADP
 First Data
… Sic Transit
© Carliss Y. Baldwin and Kim B. Clark, 2005
Value Migration: 1950-1996
Significant OptionRich Modular
Design Architectures
180
IBM System/360
160
DEC PDP 11; VAX
140
IBM PC
100
Sun 2; 3; Java VM
80
60
RISC
40
HTML; XML(?)
89
92
95
IBM
ADRs
3571
3572
3575
3576
3570 ex IBM
86
3577
83
3670
80
Intel
77
3672
74
3678
71
3674 ex Intel
68
7370
65
7371
0
62
Microsoft
Unix and C; Linux
59
7373
56
7372 ex Microsoft
53
20
7374
50
7377
Internet Protocols
(end-to-end
principle)
Slide 9
$ billion
120
© Carliss Y. Baldwin and Kim B. Clark, 2005
This was the puzzle Kim Clark
and I began to tackle in 1987
Where was the value shown in the
slide coming from?
Designs, yes, but what part and why?
Slide 10
© Carliss Y. Baldwin and Kim B. Clark, 2005
After studying the designs and
correlating their changes with value
changes
We concluded that modularity
was part of the answer…
Slide 11
© Carliss Y. Baldwin and Kim B. Clark, 2005
IBM System/360
 First
modular computer design architecture
(1962-1967)
– Proof of concept in hardware and application
software
– Proof of option value in market response and
product line evolution
– System software was NOT modularizable
» Fred Brooks, “The Mythical Man Month”
» Limits of modularity
Slide 12
© Carliss Y. Baldwin and Kim B. Clark, 2005
1965—IBM wanted to be the sole source of
all of System/360’s Modules
1
1
2
3
4
5
6
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
35
SLT architecture and standard circuits
Erich Bloch - August 1961
New Processor Line Architectural Ground Rules
SPREAD Task Group - 12/28/61
New Processor Line control, product and programming standards
Corporate Processor Control Group (CPC) - 4/1/62
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
SLT Transistors
SLT Modules
SLT Cards
SLT Boards and Automatic Wiring
Processor 1 - Endicott, New York
Processor 2 - Hursley, England
Processor 3 - Poughkeepsie, New York
Processor 4 - Poughkeepsie, New York
Processor 5 - Poughkeepsie, New York
Main memories, Corporate Memory Group (1)
Internal memories, CMG
Read-only memories for control, CMG
"Binary-addressed" Random Access Files
Corporate File Group (2)
Tape devices running at 5000+ char/sec
Corporate Tape Group (3)
Time-multiplex system for switching I/O devices
DSD Technical Development Group
Techniques to measure processor performance, system
throughput and software efficiency, Group Staff
A unified Input/output Control Structure (IOCS)
System Software for Configuration I (4)
System Software for Configuration II (4)
System Software for Configuration III (4)
FORTRAN and COBOL compilers
A unified programming language
33
34
35
Announcement and Marketing
Production, Testing and Integration
Shipment, Delivery and Installation
Slide 13
© Carliss Y. Baldwin and Kim B. Clark, 2005
IBM System/360
 IBM
did not understand the option value it
had created
 Did not increase its inhouse product R&D
 Result: Many engineers left
– to join “plug-compatible peripheral” companies
 San
Jose labs —> Silicon Valley
“Compelling, surprising, dangerous”
Slide 14
© Carliss Y. Baldwin and Kim B. Clark, 2005
1975—
What actually happened: Entry on modules
1
1
2
3
4
5
6
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
35
SLT architecture and standard circuits
Erich Bloch - August 1961
New Processor Line Architectural Ground Rules
SPREAD Task Group - 12/28/61
New Processor Line control, product and programming standards
Corporate Processor Control Group (CPC) - 4/1/62
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
SLT Transistors
SLT Modules
SLT Cards
SLT Boards and Automatic Wiring
Processor 1 - Endicott, New York
Processor 2 - Hursley, England
Processor 3 - Poughkeepsie, New York
Processor 4 - Poughkeepsie, New York
Processor 5 - Poughkeepsie, New York
Main memories, Corporate Memory Group (1)
Internal memories, CMG
Read-only memories for control, CMG
"Binary-addressed" Random Access Files
Corporate File Group (2)
Tape devices running at 5000+ char/sec
Corporate Tape Group (3)
Time-multiplex system for switching I/O devices
DSD Technical Development Group
Techniques to measure processor performance, system
throughput and software efficiency, Group Staff
A unified Input/output Control Structure (IOCS)
System Software for Configuration I (4)
System Software for Configuration II (4)
System Software for Configuration III (4)
FORTRAN and COBOL compilers
A unified programming language
33
34
35
Announcement and Marketing
Production, Testing and Integration
Shipment, Delivery and Installation
Slide 15
© Carliss Y. Baldwin and Kim B. Clark, 2005
By 1980, 100s of firms made S/360
“plug-compatible” components
Code
3570
3571
3572
3575
3576
3577
3670
3672
3674
3678
7370
7371
7372
7373
7374
7377
Category Definition
Computer and Office Equipment
Electronic Computers
Computer Storage Devices
Computer Terminals
Computer Communication Equipment
Computer Peripheral Devices, n.e.c.
Electronic Components and Accessories
Printed Circuit Boards
Semiconductors and Related Devices
Electronic Connectors
Computer Programming, Data Processing,
and Other Services
Computer Programming Services
Prepackaged Software
Computer Integrated Systems Design
Computer Processing, Data Preparation
and Processing
Computer Leasing
1960
1970
1980
5
1
1
2
1
3
11
2
8
5
2
8
6
5
1
5
7
19
4
15
9
29
36
23
10
12
11
39
10
16
1
0
0
1
9
2
7
3
26 *
12 *
13 *
16
0
0
41
5
10
108
29 *
7 *
298
* Firms in these subindustries make modules of larger computer systems.
Firms making modules =
34
95
Percent of total =
83%
88%
Slide 16
*
*
*
*
*
*
*
*
244
82%
© Carliss Y. Baldwin and Kim B. Clark, 2005
Short History (continued)

Bell and Newell, Computer Structures (1971)
– General principles of modular design for hardware
– Basis of PDP-11 design—another ORMDA

Thompson and Ritchie, Unix and C (1971-1973)
– Modular design of operating system software (contra
Brooks Law)
– Over time, general principles for evolvable software
design (Unix philosophy)

Mead and Conway, Intro to VLSI Systems (1980)
– Principles of modular design for large-scale chips
Slide 17
© Carliss Y. Baldwin and Kim B. Clark, 2005
Short History (continued)

IBM PC (1983)
–
–
–
–
–
–
Slide 18
DEC PDP-11 minimalist strategy (exclude and invite)
+ Intel 8088 chip
+ DOS system software
+ IBM manufacturing
+ Lotus 1-2-3
A mass-market ORMDA
© Carliss Y. Baldwin and Kim B. Clark, 2005
As scientists, we can visualize and
measure modularity in design
— after the fact
DSMs, Design Hierarchies
Methods are tedious, non-automated
Slide 19
© Carliss Y. Baldwin and Kim B. Clark, 2005
Comparison of different software systems with DSM tools
Mozilla just after becoming open source
Coord. Cost = 30,537,703
Change Cost = 17.35%
Linux of similar size
Coord. Cost = 15,814,993
Change Cost = 6.65%
Conway’s Law: Different organizations deliver different architectures
Mozilla just after becoming open source
One Firm,
Tight-knit
Team, RAD
methods
Coord. Cost = 30,537,703
Change Cost = 17.35%
Linux of similar size
Distributed
Open Source
Development
Coord. Cost = 15,814,993
Change Cost = 6.65%
Mozilla Before Redesign
Location
Size
Coordination Cost
Change Cost
Change Cost %
Mozilla After Redesign
Mozilla April 98 Mozilla Dec 98
On left
On right
1684
1508
30,527,703
10,234,903
292.0932
41.8561
17.35
2.78
!!
But modularity is only half the
story—options matter, too
 “Creates”
vs. “Frees up”
 Designs have “option potential”, denoted s
 s varies by system and by module
Modularity in the absence of option potential
is at best breakeven, at worst an expensive
waste of time
Slide 23
© Carliss Y. Baldwin and Kim B. Clark, 2005
Modularity without (enough)
option potential

Auto front-end modules (Fourcade)
– Much experimentation, leading nowhere
 Semiconductors
(Strojwas)
– IDMs vs. Fabless-Foundry pairs
– Competing types of option potential
– Both forms seem viable for now

Mobile computers (Whitney-Weinstein)
– Power management favors more integral designs
Slide 24
© Carliss Y. Baldwin and Kim B. Clark, 2005
What is this elusive property that
gives rise to option value?
Where does it arise?
Can we measure it?
Slide 25
© Carliss Y. Baldwin and Kim B. Clark, 2005
Measuring Option Potential
 Successive,
improving versions are evidence of
option potential being realized over time—after
the fact
 Designers see option potential before the fact
 What do they see?
Global Design Rules v.1
Version 1.0
Version 1.2
Version 1.5
Version 1.8
s = Low
Slide 26
Medium
Zero
High
© Carliss Y. Baldwin and Kim B. Clark, 2005
Major challenge in research and
practice right now
Science may not be able to deliver
tools to measure ex ante option
potential reliably
But ex ante estimates are what’s
needed
Slide 27
© Carliss Y. Baldwin and Kim B. Clark, 2005
Sources of option value 1
 Physics—
– Whitney: VLSI is different, more splittable
– But there is also Moore’s Law
» Dynamics of miniaturization
» Virtuous cycle in mfr. cost and power consumption as
chips get smaller
» Explained by Mead and Conway in 1980
– Option value lies in seamless, asynchronous
upgrading
» Modeled in Design Rules
Slide 28
© Carliss Y. Baldwin and Kim B. Clark, 2005
Sources of option value 2
 Users—new
perceptions => new preferences
– Perceptions of desires emerge through use
– Value of discovery, direct experience, play
– Unexplored preferences = option potential
 Kim
Clark (198x): Evolving design hierarchies
create new preferences w.r.t. each module
 Case studies of vertical disintegration driven by
search/discovery of new preferences
– Frozen food in the UK
– Mortgage banking in the US (IO vs. PO)
Slide 29
© Carliss Y. Baldwin and Kim B. Clark, 2005
Sources of option value 3
10,000,000
1,000,000
Key
Pfister
TPC-C v.3
TPC-C v.5
Spec CPU 95
Spec CPU 2000
tpmC and Spec CInt
100,000
10,000
1,000
100
10
1
Dec-91
Dec-93
Dec-95
Dec-97
Dec-99
Dec-01
Dec-03
Dec-05
System Availability Date
Slide 30
© Carliss Y. Baldwin and Kim B. Clark, 2005
Sources of option value 3
 Pfister’s
Observation (in my language)
– Recombining modules in new ways has more option
value than the modules themselves
 Amdahl’s
Law “Make the frequent case fast”—
– First architecture is not the best
– Fred Brooks: “Build one and throw it away”
– Value of architectural experimentation/optimization
present in all complex systems
 Not
what we model in Design Rules
– Architecture becomes a (process) module
– Stage in the value chain, node in the network
Slide 31
© Carliss Y. Baldwin and Kim B. Clark, 2005
Where we are in the argument:
 Designs
“need” to become real
– They become real by creating the perception of “value”
 Designs
act as a financial force
– Perception of value = Incentive to invest
– In making the design — “use value”
– In making the design better — “option value”
 Modular
Designs with Option Potential
– Create hurricane-type forces
– Will change their economic “space”
– Unmanageable and dangerous (unless you understand
them)
Slide 32
© Carliss Y. Baldwin and Kim B. Clark, 2005
Next questions
 What
do option-rich modular designs do to
the economy?
 How do you manage something inherently
unmanageable?
 Will you always get a modular cluster of
firms?
Slide 33
© Carliss Y. Baldwin and Kim B. Clark, 2005
QUICK Answers

What do option-rich modular designs do in the
economy?
– Attract entry with a promise of lots of $$$

How do you manage something inherently
unmanageable?
– At first, you don’t
– Then, small footprints yield high ROIC
– Then, lead firm M&A

Will you always get a modular cluster of firms?
– Yes, almost certainly
Slide 34
© Carliss Y. Baldwin and Kim B. Clark, 2005
Faced with this value proposition, what
should you do?
One module
or many?
In each
module you
chose, how
many design
searches?
Which
modules are
most
attractive?
Slide 35
© Carliss Y. Baldwin and Kim B. Clark, 2005
Lots of stories
 They
all make sense
 When you see them play out, the moves are
logical and in some cases “inevitable”
 But our strategic advice for managers and
financiers today comes down to:
–
–
–
–
Slide 36
“plunge in,”
“get lucky,”
“watch out for Microsoft,” and
“get bought by HP”
© Carliss Y. Baldwin and Kim B. Clark, 2005
Our research strategy—Look for
 Stable
patterns of behavior involving
several actors operating within a consistent
framework of ex ante incentives and ex post
rewards
– Equilibria of linked games with self-confirming
beliefs (Game theory)
Slide 37
© Carliss Y. Baldwin and Kim B. Clark, 2005
How a “stable pattern” works
 Anticipation
of $$$ (visions of IPOs)
 Lots of investment
 Lots of design searches
 Best designs “win”
 Fast design evolution => innovation
 Lots of real $$$ (an actual IPO)
“Rational expectations equilibrium”
Slide 38
© Carliss Y. Baldwin and Kim B. Clark, 2005
Three Patterns
 “Blind”
Competition
– PCs in the early 1980s
 High
ROIC on a Small Footprint
– Sun vs. Apollo
– Dell vs. Compaq (and HP and …)
 Lead
Firm Competition
– Monopoly—MSFT
– Mergers & Acquisitions—Cisco, Intel …
Slide 39
© Carliss Y. Baldwin and Kim B. Clark, 2005
“Blind” Competition
All Zeros
4.00-6.00
2.00-4.00
0.00-2.00
-2.00-0.00
-4.00--2.00
-6.00--4.00
-8.00--6.00
-10.00--8.00
ESS (1/8, 3/8, 4/8)
All Ventures
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
All Fighters
0.9
1.0
Footprint Competion 1980—
IBM provided few PC Modules
1
1
2
3
4
5
6
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
35
Intel 8088
Instruction Set
IBM PC Technical Reference Manual
for Hardware and Software —published
Microsoft DOS
Application Programmer Interfaces (APIs)
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
PC Hardware Components
Intel 8088 Microprocessor
Japanese DRAMs
Outsourced Floppy Disk Drives
Tape drives
Time-multiplex system for switching I/O devices
Techniques to measure processor performance, system
throughput and software efficiency, Group Staff
IBM BIOS
Microsoft DOS
Microsoft Basic
Lotus 1-2-3
Word Perfect
Other Applications
33
34
35
Announcement and Marketing
Production, Testing and Integration
Shipment, Delivery and Installation
Slide 41
© Carliss Y. Baldwin and Kim B. Clark, 2005
But then …
 Compaq
reverse engineered the BIOS
 Chips and Technologies made “chipsets”
 Taiwanese clones had cheaper/better
manufacturing
 Intel refused to second-source 80386
 Microsoft sabotaged OS/2
Slide 42
© Carliss Y. Baldwin and Kim B. Clark, 2005
1990—IBM PC is the standard, but IBM
makes no money
1
1
2
3
4
5
6
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
35
Intel 8088
Instruction Set
IBM PC Technical Reference Manual
for Hardware and Software —published
Microsoft DOS
Application Programmer Interfaces (APIs)
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
PC Hardware Components
Intel 8088 Microprocessor
Japanese DRAMs
Outsourced Floppy Disk Drives
Tape drives
Time-multiplex system for switching I/O devices
Magazines rate components' quality and compatibility
Clones
Microsoft DOS
Microsoft Basic
Lotus 1-2-3
Word Perfect
Other Applications
33
34
35
Clones
Slide 43
Announcement and Marketing
Production, Testing and Integration
Shipment, Delivery and Installation
© Carliss Y. Baldwin and Kim B. Clark, 2005
Footprint Competition—Apollo
1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
2
3
O
O
x
x
x
x
x
x
x
x
x
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42
Processor chip—CPU
Outsourced—Motorola 680x0
Floating Point Accelerator
Outsourced
O
Memory chips DRAMs, ROM
Outsourced—Commodity
O
Storage—Disk Drives
Outsourced
O
Storage—Tape Drive
Outsourced
O
Printed circuit boards
Outsourced—Commodity
O
Display Monitor
Outsourced
O
Keyboard, Cabinet, Fans
Outsourced
x
x
x
x
x
x
x
x
x
x
x
x
x
Inhouse
Design
x
x
x
x
x
x
x
x
x
x
x
x
x
OS
Network
x
x
x
x
x
x
x
x
x
x
x
x
Key:
x= transfer of material or information from column
task to row task;
T= transaction: sale of good by column owner to row
owner;
O= outsourced task blocks;
D= downstream or complementary task blocks;
highly interdependent task blocks with many iterations
and high within-block mundane transaction costs;
Apollo's footprint (tasks performed inhouse).
Aegis proprietary
Operating System
DOMAIN proprietary
Network Architecture
x
Hardware Design
DN series = 3-4 boards incl.
IO and Display controllers,
Power supply
Purchase Components
Hardware
T
T
T
T
T
T
T
T
T
T
T
T
Keeps
Design
Control
Slide 44
T
T
T
T
Component Test
Kits
Board stuff and Solder
Test Boards
Board Assembly
System Assembly
System Test
Quality Assurance
Consolidate and Ship
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Inhouse
Manufacturing
x
x
x
x
x
x
x
D
Many Software Applications
D
D
x
x
x
x
x
x
x
x
x
x
x
x
T
Many OEMs T
T
D
D
D
© Carliss Y. Baldwin and Kim B. Clark, 2005
Then Sun came along…
Apollo Computer
Aegis proprietary
Operating System
Inhouse
Design
OS
DOMAIN proprietary
Network Architecture
Network
Hardware
Component Test
Kits
Board stuff and Solder
Test Boards
Board Assembly
System Assembly
System Test
Quality Assurance
Consolidate and Ship
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
And did even less!
Hardware Design
DN series = 3-4 boards incl.
IO and Display controllers,
Power supply
Purchase Components
x
x
x
x
x
x
x
x
x
x
Inhouse
Manufacturing
x
x
x
x
x
x
x
x
x
x
How?
Slide 45
x
x
x
x
x
x
x
x
x
x
x
x
x x
T
T T
T
Component Test
Kits
Board stuff and Solder
Test Boards
Board Assembly
System Assembly
System Test
Quality Assurance
Consolidate and Ship
Customize Unix
Proprietary MMU
Internal bus
Single Board Layout
Purchase Components
Inhouse
Design
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
O
T
x
Manufacturing
O
T
x
x
x
x
x
© Carliss Y. Baldwin and Kim B. Clark, 2005
Then Sun came along…
Apollo Computer
Aegis proprietary
Operating System
Inhouse
Design
OS
DOMAIN proprietary
Network Architecture
Network
Hardware
Component Test
Kits
Board stuff and Solder
Test Boards
Board Assembly
System Assembly
System Test
Quality Assurance
Consolidate and Ship
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Inhouse
Manufacturing
x
x
x
x
x
Design Architecture for
performance
Public Standards for
outsourcing
Slide 46
And did even less!
Hardware Design
DN series = 3-4 boards incl.
IO and Display controllers,
Power supply
Purchase Components
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x x
T
T T
T
Component Test
Kits
Board stuff and Solder
Test Boards
Board Assembly
System Assembly
System Test
Quality Assurance
Consolidate and Ship
Customize Unix
Proprietary MMU
Internal bus
Single Board Layout
Purchase Components
Inhouse
Design
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
O
T
x
Manufacturing
O
T
x
x
x
x
x
© Carliss Y. Baldwin and Kim B. Clark, 2005
Result: ROIC advantage to Sun
Average over 16 Quarters:
Apollo
Computer
Sun
Microsystems
29%
24%
57%
15%
13%
31%
Low is good
Low is good
Low is good
Profitability
Net Income/Sales
0%
6%
High is good
ROIC
ROIC (excl Cash, Annualized)
2%
20%
High is good
Invested Capital Ratios (Annualized)
Net Working Capital/ Sales (%)
Ending Net PPE / Sales (%)
Invested Capital/Sales (%)
Sun used its ROIC advantage to drive
Apollo out of the market
Slide 47
© Carliss Y. Baldwin and Kim B. Clark, 2005
Compaq vs. Dell
 Dell
did to Compaq what Sun did to Apollo …
 Dell
created an equally good machine, and
 Used modularity-in-production to reduce its
production, logistics and distribution costs and
increase ROIC
– Negative Net Working Capital
– Direct sales, no dealers
Slide 48
© Carliss Y. Baldwin and Kim B. Clark, 2005
Dell vs. Compaq 1997
1997
Invested Capital Ratios (Annualized)
Net Working Capital/ Sales (%)
Ending Net PPE / Sales (%)
Invested Capital/Sales (%)
Profitability
Net Income/Sales
ROIC
ROIC (excl Cash, Annualized)
Compaq
Computer
Dell
Computer
-2%
8%
8%
-5%
3%
-2%
Low is good
Low is good
Low is good
8%
7%
High is good
101%
-287%
!!!
Dell started cutting prices; Compaq struggled, but in
the end had to exit.
Like Apollo, they were acquired by HP!
Slide 49
© Carliss Y. Baldwin and Kim B. Clark, 2005
Lead Firms vs. Others
 “Blind”
competitors
– don’t know others exist
 “Footprint”
competitors
– Don’t expect to influence others—just compete
 “Lead
firms”
– Must influence the beliefs of their competitors
– FUD — “Fear, uncertainty and doubt”
– Others cannot be blind!
Slide 50
© Carliss Y. Baldwin and Kim B. Clark, 2005
Monopoly or M&A?

Monopoly needs to deter all potential entrants with
threats of price war
– Very fragile equilibrium
– Potentially expensive to create “enough” FUD

M&A Lead Firm does not try to deter all entry in
the design space
– Expects to buy most successful entrants ex post
– More robust equilibrium
– Maybe more advantageous, when you count the cost of
FUD
Slide 51
© Carliss Y. Baldwin and Kim B. Clark, 2005
Industry Outcomes
 “Blind”
competition
– Cluster (entry everywhere in the architecture)
 “Footprint”
competition
– Cluster (small footprint => vertical
disintegration)
 M&A Lead
Firm
– Cluster (lead firm does not deter all entry)
 Monopoly
– One Big Firm
Slide 52
© Carliss Y. Baldwin and Kim B. Clark, 2005
Remember
 Designs
“need” to become real
– They become real by creating the perception of “value”
 Designs
act as a financial force
– Perception of value = Incentive to invest
 A Modular
Designs with Option Potential
– Create hurricane-type forces
– Will change their economic “space”
– Unmanageable—cannot be confined in one firm or a
supply chain
– Dangerous (unless you understand them)
Slide 53
© Carliss Y. Baldwin and Kim B. Clark, 2005
One more story before we close—
Significant OptionRich Modular
Design Architectures
180
IBM System/360
160
DEC PDP 11; VAX
140
IBM PC
100
Sun 2; 3; Java VM
80
60
RISC
40
HTML; XML(?)
89
92
95
IBM
ADRs
3571
3572
3575
3576
3570 ex IBM
86
3577
83
3670
80
Intel
77
3672
74
3678
71
3674 ex Intel
68
7370
65
7371
0
62
Microsoft
Unix and C; Linux
59
7373
56
7372 ex Microsoft
53
20
7374
50
7377
Internet Protocols
(end-to-end
principle)
Slide 54
$ billion
120
© Carliss Y. Baldwin and Kim B. Clark, 2005
The Bright Side of the Option-rich
Modular Designs
Slide 55
© Carliss Y. Baldwin and Kim B. Clark, 2005
But there was The Dark Side …
4500
4000
$ 2.5 trillion
appeared then
disappeared in
the space of four
years!
3500
3000
$ billion
2500
2000
1500
1000
500
0
50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02
Slide 56
© Carliss Y. Baldwin and Kim B. Clark, 2005
Bubble followed by a Crash
4500
A failure,
4000
not of the Internet’s
design architecture,
3500
3000
$ billion
2500
but of the economic
institutions built on
that architecture
2000
1500
1000
500
0
50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02
Slide 57
© Carliss Y. Baldwin and Kim B. Clark, 2005
Ultimate unmanageability
4500
4000
3500
3000
$ billion
2500
2000
1500
1000
500
0
50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02
Slide 58
© Carliss Y. Baldwin and Kim B. Clark, 2005
A reason—if we need one—to try
to turn our stories into science…
Slide 59
© Carliss Y. Baldwin and Kim B. Clark, 2005
Thank you!
Slide 60
© Carliss Y. Baldwin and Kim B. Clark, 2005