CANDE Panel on EDA R&D
Rajesh Gupta
UC San Diego
3 questions, 3 answers
Has government (e.g., DARPA) changed what it wants
from universities? Why has the government funding
dropped? Has it moved to other areas?
A1: Shift is pretty fundamental
•
•
3X increase in submissions
over 5 years
Pathetic funding rates:
–
–
–
–
–
–
–
•
~5% ; ~6% requested $$
Theory 05: 11% $70K/yr
CCF 04: 5%
Cybertrust 04: 8%
IDM 04: 3%, FY05: 0%
IIS 04: 5%
CNS 04: 10-15%
DARPA
–
CISE provides 86% of federal obligations for basic CS
EDA is a surefire way to
get your proposal rejected.
research.
Sources: Various, Stefan Savage
Remember the IT Boom?
FY00
ITR $ awarded FY00-FY02 from NSF/CCR division
FY01
FY02
$45,000,000
$8.8M $0
$2.9M
$40,000,000
$35,000,000
US Dollars
$30,000,000
$25,000,000
$20,000,000
$188M
$15,000,000
$10,000,000
$5,000,000
$0
COMM
CSA
DA
DSC
EHS
GSG
CCR Programs
Source: Sankar Basu, NSF
SEL
SPS
TC
COMM
CSA
DA
DSC
EHS
GSG
SEL
SPS
TC
TOC
TOC
Communications Research
Computer Systems Architechture
Design Automation for Micro and Na
Distributed Systems and Compilers
Embedded and Hybrid Systems
Graphics and Symbollic
Software Engineering and Language
Signal Processing Systems
Trusted Computing
Theory of Computing
Do we need big science centers or individual projects,
or a combination?
A2: Ill-posed. What is the role of
unsolicited research?
• Agencies routinely engage in divining future:
– DARPA disease without the DARPA bucks
– picking up technology winners by committee or
worse.
• It has never worked, it never will
– General Magic, MCC, FGCP, …
• But we now know what does work
Golden Age of US Semi Industry Recovery
Year
1985
1986
1987
1988
1989
1990
1991
1992
#consortia
registered
50
17
27
33
34
45
59
59
Is the government view that industry should pay for
its own research?
Yes and No
A3: GOP still works
Good old politics
– Typically, industry claims to invest x% of sales to R&D (10-K)
• SEMI: 6%, Telecom semi: 15% (QCOM: 20%)
– About 1.0 to 1.5% of R&D is on long term research (SRC,
MARCO)
• Looking at the semiconductors research
– The ITRS related R&D is $900M WW
• Equally split between US and ROW
• US portion about equally split between Gov and Industry
– ($200M from US Gov, about 5x the SRC budget)
• EDA tends to pile along Semi
– The last serious attempt for 1994 SRC Gap study that showed a
gap of about $500M that led to MARCO FCRP
US Appropriations: FUD Works
Closing: EDA R&D Boat Rocks in a
Rule Changing World
• Goals of technology policy are changing
– Pre 9/11:
• We need technology advantage for a superior military
• Collateral commercial developments are a good thing
– Post 9/11:
• We have technology advantage. We need deployment.
• A deep paranoia of the foreigners among our midst
• AWOL Advocacy of EDA R&D
– Our NAE members, industry leaders need to be proactive
– Educate, enlighten the policy makers and their mentors
about what is at stake with EDA R&D
– Semi’s are 6% of the manufacturing GDP
– Precompetitive advantage has moved from equipment to
methods
But, ROW has Changed
Slides I will never get to
Research Funding:
Global Picture
Rajesh Gupta
UC San Diego
Research Spending Has Flatlined
US S&T Appropriations
• Highly decentralized
– Individual subcommittees that have funding oversight
over individual agencies.
– The chief educators in this process
•
•
•
•
Congressional Research Service, LoC, CBO, GAO, NRC
OTA was closed in 1995
EOP actors: OSTP, CEA, NSC, OMB
Long term planning is difficult to achieve since most honchos
are political appointees
– Let us look at NSF, DARPA, Consortia (SRC,
SEMATECH)
NSF: A Historical Perspective
• At the end of WWII, Pres Roosevelt asked Dr Vannevar
Bush to lead review of S&T research in the US in an
attempt to consolidate the science lead nation had built
up as a part of the war effort
– As Director of Office of Scientific Research and Development, he
oversaw work of 6000 scientists involved in the war effort
• Out came the report: “Science: The Endless Frontier”
submitted to President Truman
– Formed the basis of “University, Industry, Government” compact
on research and education
– One of the most persuasively written policy document in the
nation’s history
• “Scientific progress is one essential key to our security as a nation,
to our better health, to more jobs, to a higher standard of living, and
to our cultural progress.”
• Led to the emergence of “Research University”
NSF est 1950
• Three policy pillars
– Federal support of basic scientific research
– Role of research universities
– Federal support of education of young people in science and
engineering
• “Research University”
– “The publicly and privately supported colleges, universities, and
research institutes are the centers of basic research. They are
the wellsprings of knowledge and understanding.”
• Currently at $5.47B (-105M)
– $4.22B for research, $841M EHR, $174 Maj Equip
• CISE and ENG at about $600M each
• But, it is no longer business as usual…
Thought Leaders Today
• For the first time in nation’s history, this compact is up for
elimination
– A nation at war has other priorities
• Partly driven by “neocon” dislike for centrally organized
anything
– A (minority) thought that never really bought the endless frontier.
Cf: @Cato & Hoover; Donald Kennedy on “Riding through the
Endless Frontier –Right past the students”
• Partly by a palpable political sense that science is getting
in the way of policy (and ethics)
– Not really laughable concerns. Cf: Bill Joy
DARPA: Mission Oriented
DARPA
• FY 2005: $2.97B
– $1.3B is basic research, $1.6B is applied research
• Materials and Electronics is about $0.5B
– Zero for design technology, embedded stuff
• FY 2006: $3.08B
– $1.4B basic, $1.5B applied
• All computing related money to Cognitive Computing
($200M)
• electronics at $241M
– Network centric warfare is a growth component.
State Participation
• Goes long back: 19th century – land grant
universities that focused on agriculture and
technology
• Feds took dramatic lead during WWII
– Vannever Bush vision
• IT boom enabled states to get back into the
action
– Even as feds withdrew to 0.9% of GDP from 1.5% in
1965
– Almost all of Federal R&D is mission-oriented
• About 5b of 68b is general R&D (NSF, NIH)
Consortia
• EDA players can not afford stand-alone
R&D efforts
• Precompetitive R&D is perhaps the oldest
outsourced activity
– A new global perspective to this outsourcing
– Plus Sarbanes+Oxley 3a-8 prevents deferred
recognition of R&D costs
• “call options” on R&D outcomes no longer as easy
to exercise
Consortia Evolution
• NCRA 1984 allowed R&D consortia to emerge
– (beyond the reach of anti-trust legislation)
– Consortia evolving from small budget R&D outfits to large budget
development “joint-ventures”
• MCC (1982) in response to Japanese Fifth Generation Computer
Program in 1981, Alvey program in UK
– MCC failed because of conflict in expectations and political reality
• Participating companies expected quick(er) returns than internal R&D
• FGCP (1982-1992) also failed for its ill conceived technical vision
– DARPA is now going down the same path “intelligent information
processing”, AI+Cmputer
• SEMATECH
– Improved execution (insist on assignee quality)
• The ROW has caught onto to Consortia even faster
– The European Frameworks are generally $13-15B
Sematech
• 1986: US share of semi market was project to reach
20% by 1993
– DSB and SIA contributed to the overall noise against the
Japanese
• 14 companies accounting for 80% of US semi industry
banded together
– Strange bedfellows bounded by a common threat
– Goal: “To provide the US Semi industry the capability of
achieving world-leadership manufacturing position by the mid
1990s.”
– $100M/year from US, 1% of sales ($1M-$15M)
• $200M/year operating budget
• By 1993, US Semi overtook Japan; by 1996: US: 44%,
Japan: 36%
– Central research facility key to its success.
• Look at Alberto’s EDAtech proposal along these lines…
The Network Centric Warfare and
National Defense
• Remember “General Magic”?
– That magical partnership between Apple, Sony,
Motorola, AT&T with the goal of building “a ubiquitious
communication infrastructure”
– MagicCap, Telescript
– IPO’d in 1995, broadsided by the real McCoy’s only a
year later…while it kept on reinventing itself year after
year as a internet, webservices company before
finally quitting in 9/02