[In Stanford, I heard a talk by Katsuhiro Shimohigashi, a prolific inventor in semiconductor technology, and learnt all about it in an hour; this column, from Business Standard of 30 December 1999, was the result]
Whither research?
Wireless
communication is crucial to warfare. At the beginning of World War II, it
relied on the radio, which used valves. Even today there is a company in Poona,
Cadence, which uses old-fashioned valves to produce wonderful hi-fi amplifiers.
But valves used mica, of which India and Russia were almost the only producers.
There was a danger that the Japanese would take India, and the Allies would be
left without mica. So the US worked on a substitute on a war footing, and invented
the semiconductor. The semiconductor has transformed lifestyles in my lifetime.
First there were
germanium semiconductors; in 1957 germanium was replaced by silicon, which
apart from being more compact gave more reliable performance. Silicon semiconductors
led to transistor radios. By the early 1960s, a number of circuits were being
etched on to a single semiconductor; integrated circuits led to transistor
television sets and electronic calculators. Large-scale integration (LSI) came
in the late 1960s. The chip of that time had 1 kilobyte capacity, and was 10
micron meters (a 100,000th of a meter) in size. It was the building
block of the first personal computers, which came in as a rather expensive and
limited substitute for mainframes. Very large scale integration (VLSI) followed
in the mid-1970s, chips could pack in 16 kB, and formed the basis of the first
laptops; my first laptop, bought in 1987, was just good enough to run Wordstar
and Supercal. But by then, ultra-large scale integration (ULSI) had already
arrived; 1mm ULSI chips could pack in 256kB. Today,
256MB, 0.2mm chips are used in cordless telephones. Even
smaller, 1GB, 0.1mm chips are being used to record 70 minutes
of hi-fi music on CDs. Car navigation systems use the same compact chips. What
is going to happen next is that the application of chips to products of single
use – computer, television, telephone etc – will cease, and products will come
to embody a number of functions. Sony already makes a laptop which is also a DVD
camera; similarly, we will get compuvision and telecam.
Can this dizzy
ascent go on? Whatever happens to the technology, the growth in demand is
already slowing down. The market for semiconductors expanded from $1 billion
thirty years ago to $50 billion in 1990 and $100 billion in 1994. In 1999,
however, it reached only $141 billion. The market is no longer doubling every
four years; it did not expand even 50 per cent in the last 5 years. And as
market growth slows down, the market structure is also changing. Already, production
is getting concentrated in big firms; by the early years of the century, chips
manufacture will become an oligopoly. Equipment production is still rising
rapidly. But here too, larger firms are consolidating.
What will these
trends mean for research and development? I heard Katsuhiro Shimohigashi,
Hitachi’s Manager of Semiconductor Technology Development, speculate on this.
According to him, the past trend, towards rising density and falling price per
bit, must continue. But in addition, multipurpose products will require more
complex systems to be etched into the chip; and as their performance
requirements go up, their energy requirements will have to be brought down. So
low energy architecture will remain the focus of R&D. But as chips become more
complex, the investments required for manufacture will rise; so the
manufacturing system will become a major object of study – keeping capital
costs down will be paramount.
Fabrication
facilities are getting bigger and fewer; it is getting less possible to learn
from the building of one facility for the construction of the next. Learning
must be telescoped. This means that R&D must work on lowering capital costs
even as plants are being built.
In other words, the
conceptual distinction between research, development and production is breaking
down. Traditionally, while ideas were being explored and uncertainty was high,
research went on on a laboratory scale. Once the ideas were proved, they were
embodied into a bench-scale process to make sure that they would work in
manufacturing; that was development. Once pilot production was established, it
was scaled up for full-scale production.
What is happening
is that research is now being channeled into the investment for full-scale
production: the distinction between development and production is breaking
down. Development and production engineers are being brought together, and
joint development-production teams are working to cut out the development
phase.
And what will
happen to research before development? Its costs are rising and productivity is
falling. If this trend continues, there will be greater pressure for cost
sharing. The share of corporate R&D will fall. Long range research with no
clear foreseeable results will be done in universities and funded by
governments. Research into infrastructure will be done by consortia. Then, as
technologies emerge which can be used in production, industry organizations
will standardize products and processes; at that stage, there may be joint
development alliances between companies. Companies will concentrate their own
research on building up core competence.
What this amounts
to is the advent of just-in-time production in research. Back in the 1950s,
Toyota cut inventory costs in car manufacture dramatically by bringing in
inputs as close to the point of their consumption as possible. R&D
represents an inventory of ideas that precedes production; companies are now
looking to cutting down this inventory, and shortening the time lag between the
original conception and its final embodiment into production.
That is sombre news
for researchers. Shimohigashi took out 286 patents in his research career of 32
years. His peak was in 1979, when he took out 32 patents. In the 1980s, which
were his peak period; his average annual output was 20. In the 1990s it has
been closer to 5. But of the 286 patents, only 20 contributed to Hitachi’s
business – a hit ratio of 7 per cent! In the old era, Shimohigashi would have
been hailed as an effervescent genius. In the coming years, hard-headed
managers will ask: genius to what purpose?
