Subject: Re: [ppc-mobo] Re: GPL-like hardware design license?
From: "Stephen J. Turnbull" <>
Date: Wed, 6 Oct 1999 09:54:24 +0900 (JST)

>>>>> "Ben" == Ben Tilly <> writes:

    Ben> But as long as a chip fabricating plant costs a billion
    Ben> dollars, sorry.  We cannot make hardware free.

True, if you mean free-as-in-free-beer "hardware = 1G$ fab".  False,
if you mean free-as-in-free-speech chips.

Assuming non-increasing returns to scale, in equilibrium there will be
at most enough entry into the chip-making industry to drive prices
down to the point where profits are sufficient to cover the return to
capital demanded by investors on the $1G fab.  The chip designers can
then fend for themselves without worrying about the fab; in effect,
they simply hire capital at the going rate to build the fab.  (In
fact, this is the way the capital market seems to work for investments 
of that size; for small businesses, of course, it's quite different.)

The rub in practice comes from the fact that there do seem to be
increasing returns to scale, and furthermore the Japanese and Korean
decision-makers are apparently not profit-maximizing, but include
strong components of market share and gross revenues in their
objectives.  All of these factors tend to produce over-investment
compared to the static equilibrium; the logic is apparently that by
getting in big you stand a chance of chasing some others out while
decreasing your costs, and thus winning big.  Since not all can win
big, this results in wasted resources which cannot be covered by the
returns to chip manufacture.

In this environment, firms will seize on any competitive advantage,
and these advantages cannot be decomposed from the large investment in 
the fab plant because the firms must recoup their (on-average)
overinvestment in the fabs, and do this by collecting monopoly rents
on the chip design itself.

However, the overinvestment is primarily due to technological factors
(namely, increasing returns to scale in chip manufacture).  If the
market for chips should increase by an order of magnitude, and
technology should evolve in a way such that the minimum efficient size 
of plants does not increase over current plant sizes, then in the
region relevant to decision making ("the margin") there will be
non-increasing returns to scale.  (The Japanese and Koreans are
already learning the error of their big-at-the-expense-of-profit ways.)

At that point, the plant-building and chip manufacturing decision can
be decoupled from the design, they can be assigned separate profit
centers, it is possible that firms specializing in chip manufacture
can emerge, and "free-as-in-free-speech" hardware design has a
fighting chance of emerging as viable in the market.

I don't argue that this is likely; but even in hindsight I would
argue that Linux is unlikely.  "Hardware GPL" is not implausible,
especially if one picks industries were the technology is not so
unfavorable as chip fab currently is.

University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
What are those two straight lines for?  "Free software rules."