Transcript NFRS Seminar - Harvard Business School

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
S
P
Systems Integration
E
R
R
Top 10 Public
Companies in
US Computer
Industry
Area reflects
Market Value
in Constant
US $
Slide 5
Applications Layer
Y
Middleware Layer
Operating Systems
U H E
N P C
IBM
D CVC
S
Hardware
Y
S
XRC
AMP
Components
TI
Intel
© Carliss Y. Baldwin and Kim B. Clark, 2005
Andy’s Movie
Stack View in 1995
S
Services
P
E
R
R
Y
IBM
First Data
D CVC
U H E
N P C
S
Y
S
XRC
Systems Integration
EDS
AMP
TI
Oracle
Intel
I
Applications Layer
Middleware Layer
Top 10 Public
Companies in
US Computer
Industry
Area reflects
Market Value
in Constant
US $
Slide 6
CA
B MSFT
M
Operating Systems
Hardware: Printers
HP
Hardware: Servers
Hardware: Routers
IBM
Cisco
Components
Intel
Micron
© Carliss Y. Baldwin and Kim B. Clark, 2005
Andy’s Movie—the Sequel
Stack View in 2002
S
Services
P
E
R
R
Y
IBM
D CVC
U H E
N P C
S
Y
S
XRC
AMP
TI
Intel
First Data
First Data
Systems Integration ADP EDS
Systems Integration
Services
I
Applications Layer
Applications
Layer
IBM
Middleware
Layer
Top 10 Public
Companies in
US Computer
Industry
Area reflects
Market Value
in Constant
US $
Slide 7
Oracle
CA
Oracle
B MSFT
M
Operating Systems
Middleware
Layer Printers
Hardware:
MSFT
HP
Hardware: Servers
Operating Hardware:
Systems Routers
IBM
Cisco
Components
Hardware: Printers
HP
Hardware: PCs
Hardware: Servers
Hardware: Routers
Components
Intel
Micron
Dell
IBM
Cisco
Intel
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
$ b illio n
120
60
RISC
40
20
77
Slide 9
3678
7370
M ic ros oft
7373
95
7374
92
7372 ex M ic ros oft
HTML; XML(?)
89
7377
Unix and C; Linux
86
7371
83
3674 ex Intel
80
IB M
AD Rs
74
3571
71
3570 ex IB M
68
3572
65
3575
0
62
3576
59
3577
56
3670
53
Intel
50
3672
Internet Protocols
(end-to-end
principle)
© 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
2
3
4
5
6
7
8
9
10
11
1
SLT architecture and standard circuits
2
Erich Bloch - August 1961
12
13
14
15
3
New Processor Line Architectural Ground Rules
4
SPREAD Task Group - 12/28/61
16
17
18
19
20
5
New Processor Line control, product and programming standards
6
Corporate Processor Control Group (CPC) - 4/1/62
7
SLT Transistors
8
SLT Modules
9
SLT Cards
10
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
SLT Boards and Automatic Wiring
11
Processor 1 - Endicott, New York
12
Processor 2 - Hursley, England
13
Processor 3 - Poughkeepsie, New York
14
Processor 4 - Poughkeepsie, New York
15
Processor 5 - Poughkeepsie, New York
16
Main memories, Corporate Memory Group (1)
17
Internal memories, CMG
18
Read-only memories for control, CMG
19
"Binary-addressed" Random Access Files
20
Corporate File Group (2)
21
Tape devices running at 5000+ char/sec
22
Corporate Tape Group (3)
23
Time-multiplex system for switching I/O devices
24
DSD Technical Development Group
25
Techniques to measure processor performance, system
26
throughput and software efficiency, Group Staff
27
A unified Input/output Control Structure (IOCS)
28
System Software for Configuration I (4)
29
System Software for Configuration II (4)
30
System Software for Configuration III (4)
31
FORTRAN and COBOL compilers
32
A unified programming language
33
Announcement and Marketing
34
Production, Testing and Integration
35
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
2
3
4
5
6
7
8
9
10
11
1
SLT architecture and standard circuits
2
Erich Bloch - August 1961
12
13
14
15
3
New Processor Line Architectural Ground Rules
4
SPREAD Task Group - 12/28/61
16
17
18
19
20
5
New Processor Line control, product and programming standards
6
Corporate Processor Control Group (CPC) - 4/1/62
7
SLT Transistors
8
SLT Modules
9
SLT Cards
10
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
SLT Boards and Automatic Wiring
11
Processor 1 - Endicott, New York
12
Processor 2 - Hursley, England
13
Processor 3 - Poughkeepsie, New York
14
Processor 4 - Poughkeepsie, New York
15
Processor 5 - Poughkeepsie, New York
16
Main memories, Corporate Memory Group (1)
17
Internal memories, CMG
18
Read-only memories for control, CMG
19
"Binary-addressed" Random Access Files
20
Corporate File Group (2)
21
Tape devices running at 5000+ char/sec
22
Corporate Tape Group (3)
23
Time-multiplex system for switching I/O devices
24
DSD Technical Development Group
25
Techniques to measure processor performance, system
26
throughput and software efficiency, Group Staff
27
A unified Input/output Control Structure (IOCS)
28
System Software for Configuration I (4)
29
System Software for Configuration II (4)
30
System Software for Configuration III (4)
31
FORTRAN and COBOL compilers
32
A unified programming language
33
Announcement and Marketing
34
Production, Testing and Integration
35
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
Category Definition
1960
1970
1980
3570
Computer and Office Equipment
5
2
9
3571
Electronic Computers
1
8
29
3572
Computer Storage Devices
1
6
36 *
3575
Computer Terminals
2
5
23 *
3576
Computer Communication Equipment
1
1
10 *
3577
Computer Peripheral Devices, n.e.c.
3
5
12 *
3670
Electronic Components and Accessories
11
7
11 *
3672
Printed Circuit Boards
2
19
39 *
3674
Semiconductors and Related Devices
8
4
10 *
3678
Electronic Connectors
5
15
16 *
7370
Computer Programming, Data Processing,
and Other Services
26 *
Computer Programming Services
1
0
9
7371
2
12 *
7372
Prepackaged Software
0
7
13 *
7373
Computer Integrated Systems Design
1
3
16
7374
Computer Processing, Data Preparation
and Processing
0
0
41
5
29 *
10
108
7 *
298
* Firms in these subindustries make modules of larger computer systems.
Firms making modules =
34
95
244
88%
82%
7377
Computer Leasing
Percent of total =
Slide 16
83%
© 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)
–
–
–
–
–
–
DEC PDP-11 minimalist strategy (exclude and invite)
+ Intel 8088 chip
+ DOS system software
+ IBM manufacturing
+ Lotus 1-2-3
A modular design architecture with a mass market
Visions of $$$ dance in your heads!
Slide 18
© 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
 Mobile
computers (Whitney-Weinstein)
– Power management favors more integral designs

Semiconductors (Strojwas)
– IDMs vs. Fabless-Foundry pairs
– Competing types of option potential
– Both forms seem viable for now
Slide 24
© Carliss Y. Baldwin and Kim B. Clark, 2005
Stack View of the Semiconductor
Industry (Strojwas, 2005)
Top 10 Firms:
1994 and 2004
Slide 25
© 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 26
© 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 27
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 28
© 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 29
© 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 (1985): 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 30
© Carliss Y. Baldwin and Kim B. Clark, 2005
Sources of option value 3
10,000,000
Key
Pfister
TPC-C v.3
TPC-C v.5
Spec CPU 95
Spec CPU 2000
1,000,000
tpm C a n d S pe c C In t
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 31
© 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 should be a (process) module
– Stage in the value chain, node in the network
Slide 32
© 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
– Use value
– Option value = Potential to improve
 Modular
Designs with Option Potential
– Create hurricane-type forces
– Will change their economic “space”
– Unmanageable and dangerous (unless you understand
them)
Slide 33
© 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 34
© 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 35
© 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 36
© 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 37
“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 selfconfirming beliefs (Game theory)
Slide 38
© 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 39
© Carliss Y. Baldwin and Kim B. Clark, 2005
Three Stable Patterns
(not quite Equilibria…)
 “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 40
© 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
2
3
4
5
6
1
Intel 8088
2
Instruction Set
7
8
9
10
11
12
13
3
IBM PC Technical Reference Manual
4
for Hardware and Software —published
14
15
5
Microsoft DOS
6
Application Programmer Interfaces (APIs)
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
7
8
PC Hardware Components
9
10
11
12
13
Intel 8088 Microprocessor
14
15
16
17
Japanese DRAMs
18
19
20
Outsourced Floppy Disk Drives
21
22
Tape drives
23
24
Time-multiplex system for switching I/O devices
25
Techniques to measure processor performance, system
26
throughput and software efficiency, Group Staff
27
IBM BIOS
28
Microsoft DOS
29
Microsoft Basic
30
Lotus 1-2-3
31
Word Perfect
32
Other Applications
33
Announcement and Marketing
34
Production, Testing and Integration
35
Shipment, Delivery and Installation
Slide 42
© 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 43
© Carliss Y. Baldwin and Kim B. Clark, 2005
1990—IBM PC is the standard, but IBM
makes no money
1
2
3
4
5
6
1
Intel 8088
2
Instruction Set
7
8
9
10
11
12
13
3
IBM PC Technical Reference Manual
4
for Hardware and Software —published
14
15
5
Microsoft DOS
6
Application Programmer Interfaces (APIs)
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
7
8
PC Hardware Components
9
10
11
12
13
Intel 8088 Microprocessor
14
15
16
17
Japanese DRAMs
18
19
20
Outsourced Floppy Disk Drives
21
22
Tape drives
23
24
Time-multiplex system for switching I/O devices
25
26
Magazines rate components' quality and compatibility
27
Clones
28
Microsoft DOS
29
Microsoft Basic
30
Lotus 1-2-3
31
Word Perfect
32
Other Applications
33
Announcement and Marketing
34
Clones
35
Production, Testing and Integration
Shipment, Delivery and Installation
Slide 44
© Carliss Y. Baldwin and Kim B. Clark, 2005
Footprint Competition—Apollo
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
19
3
O
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
Key:
Floating Point Accelerator
4
x=
Outsourced
5
O
T=
Outsourced—Commodity
7
O
8
9
transfer of material or information from column
task to row task;
Memory chips DRAMs, ROM
6
transaction: sale of good by column owner to row
owner;
Storage—Disk Drives
O=
Outsourced
D= downstream or complementary task blocks;
O
10
11
outsourced task blocks;
Storage—Tape Drive
highly interdependent task blocks with many iterations
Outsourced
and high within-block mundane transaction costs;
O
Printed circuit boards
Apollo's footprint (tasks performed inhouse).
Outsourced—Commodity
12
13
O
Display Monitor
14
Outsourced
15
O
Keyboard, Cabinet, Fans
16
Outsourced
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Aegis proprietary
18
Inhouse
19
Design
20
20
Outsourced—Motorola 680x0
2
17
18
Processor chip—CPU
1 O
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Operating System
OS
DOMAIN proprietary
21
Network
Network Architecture
22
23
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Hardware Design
24
DN series = 3-4 boards incl.
25
Hardware
IO and Display controllers,
26
27
Power supply
T
T
T
T
T
T
T
T
T
T
T
T
T
T
T
T
28
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Board Assembly
x
x
x
x
x
System Assembly
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Component Test
Keeps
Design
Control
29
30
31
32
33
34
35
36
37
Kits
Board stuff and Solder
Test Boards
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
40
x
x
x
x
x
x
x
x
x
x
41
x
x
x
x
x
x
x
x
x
x
42
x
x
x
x
x
x
x
x
x
x
38
39
Slide 45
Purchase Components
x
Inhouse
x
Manux
facturing
x
x
x
x
x
D
Many Software Applications
D
D
T
Many OEMs
T
T
D
D
D
© Carliss Y. Baldwin and Kim B. Clark, 2005
Then Sun came along…
Apollo Computer
Aegis proprietary
Inhouse
Operating System
Design
OS
DOMAIN proprietary
Network
And did even less!
Network Architecture
Hardware Design
DN series = 3-4 boards incl.
Hardware
IO and Display controllers,
Power supply
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Board Assembly
x
x
x
x
x
System Assembly
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Component Test
Kits
Board stuff and Solder
Test Boards
System Test
Quality Assurance
Consolidate and Ship
Purchase Components
x
Inhouse
x
Manux
facturing
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Inhouse
Customize Unix
Proprietary MMU
x
x
x
x
x
Design
Internal bus
x
T
How?
T
x
T
T
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Board Assembly
x
x
x
x
System Assembly
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Component Test
Kits
Board stuff and Solder
Test Boards
System Test
Quality Assurance
Consolidate and Ship
Slide 46
Single Board Layout
x
Purchase Components
x
O
T
Manux
O
facturing
T
x
x
x
x
x
© Carliss Y. Baldwin and Kim B. Clark, 2005
Then Sun came along…
Apollo Computer
Aegis proprietary
Inhouse
Operating System
Design
OS
DOMAIN proprietary
Network
And did even less!
Network Architecture
Hardware Design
DN series = 3-4 boards incl.
Hardware
IO and Display controllers,
Power supply
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Board Assembly
x
x
x
x
x
System Assembly
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Component Test
Kits
Board stuff and Solder
Test Boards
System Test
Quality Assurance
Consolidate and Ship
Purchase Components
x
Inhouse
x
Manux
facturing
x
x
x
Design Architecture for
performance
x
x
x
x
x
x
x
x
x
x
x
x
Inhouse
Customize Unix
Proprietary MMU
x
x
x
x
x
Design
Internal bus
x
T
T
x
T
T
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Board Assembly
x
x
x
x
System Assembly
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
Component Test
Kits
Board stuff and Solder
Test Boards
Public Standards for
outsourcing
Slide 47
Single Board Layout
x
System Test
Quality Assurance
Consolidate and Ship
Purchase Components
x
O
T
Manux
O
facturing
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
Invested Capital Ratios (Annualized)
Net Working Capital/ Sales (%)
29%
15%
Low is good
Ending Net PPE / Sales (%)
24%
13%
Low is good
Invested Capital/Sales (%)
57%
31%
Low is good
0%
6%
High is good
2%
20%
High is good
Profitability
Net Income/Sales
ROIC
ROIC (excl Cash, Annualized)
Sun used its ROIC advantage to drive
Apollo out of the market
Slide 48
© 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
footprint in production, logistics and distribution
costs
– Negative Net Working Capital
– Direct sales, no dealers
 Result
Slide 49
= Higher ROIC
© Carliss Y. Baldwin and Kim B. Clark, 2005
Higher ROIC always wins!
1997
Compaq
Computer
Dell
Computer
Invested Capital Ratios (Annualized)
Net Working Capital/ Sales (%)
-2%
-5%
Low is good
Ending Net PPE / Sales (%)
8%
3%
Low is good
Invested Capital/Sales (%)
8%
-2%
Low is good
8%
7%
High is good
Profitability
Net Income/Sales
ROIC
ROIC (excl Cash, Annualized)
101%
-287%
!!!
Dell started cutting prices; Compaq struggled, but in
the end had to exit.
Like Apollo, they were acquired by HP!
Slide 50
© 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 51
© 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 52
© 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 53
© Carliss Y. Baldwin and Kim B. Clark, 2005
Designs as an object of economic
analysis—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
 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 54
© 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
$ b illio n
120
60
RISC
40
20
77
Slide 55
3678
7370
M ic ros oft
7373
95
7374
92
7372 ex M ic ros oft
HTML; XML(?)
89
7377
Unix and C; Linux
86
7371
83
3674 ex Intel
80
IB M
AD Rs
74
3571
71
3570 ex IB M
68
3572
65
3575
0
62
3576
59
3577
56
3670
53
Intel
50
3672
Internet Protocols
(end-to-end
principle)
© Carliss Y. Baldwin and Kim B. Clark, 2005
The Bright Side of the Option-rich
Modular Designs
Slide 56
© 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
$ b il l io n
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 57
© 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
$ b il l io n
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 58
© Carliss Y. Baldwin and Kim B. Clark, 2005
Ultimate unmanageability
4500
4000
3500
3000
$ b il l io n
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 59
© Carliss Y. Baldwin and Kim B. Clark, 2005
A reason—if we need one—to try
to turn our stories into science…
Slide 60
© Carliss Y. Baldwin and Kim B. Clark, 2005
Thank you!
Slide 61
© Carliss Y. Baldwin and Kim B. Clark, 2005