From: Tim May <tcmay@got.net>
Date: Sat, 10 Jan 1998 04:18:46 +0800
To: Ryan Anderson <ryan@michonline.com>
Subject: Re: In God We Antitrust, from the Netly News
In-Reply-To: <v03102800b0dc148aaea5@[207.167.93.63]>
Message-ID: <v03102800b0dc2d3e93db@[207.167.93.63]>
MIME-Version: 1.0
Content-Type: text/plain



At 11:01 AM -0800 1/9/98, Ryan Anderson wrote:
>On Fri, 9 Jan 1998, Tim May wrote:

>> Intel's failed competitors (Cyrix, AMD, Motorola, Sun, SGI/MIPS,
>> Intergraph) can be counted on to run crying to Mother Government, crying
>> that Intel is too successful.
>
>Failed? AMD is farfrom a failed competitor.  Intel is nowhere near being a
>monopoly.  In this industry,Intel creates a chip, then AMD,Cyrix, etc take
>the published specs on that chip and duplicate the work.  They same some

Charitably, I'll assume you just don't follow the industry very closely. I
don't mean in terms of claims and press announcements, I mean in terms of
the "ground truth" of what is real.

Intel conservatively is now two full iterations ahead of AMD/Nexgen and/or
National/Cyrix. The AMD K6 may not be quite the dog the K5 was, but AMD is
woefully unable to *make* the part!

(Manufacturing is more essential than most people realize. I could
elaborate on this for pages and pages, but this is well-trod ground in,
say, the Silicon Investors Forum and other such groups.)

There have been reliable reports, from several kinds of sources, that AMD's
Fab 25 in Austin is yielding only a handful of workingn (at speed) K6s per
8-inch wafer, versus well over 100 working (at full rated speed,
importantly) Pentium IIs (and variamts) devices per 8-inch wafer. Intel is
running at a nearly perfect yield rate (most die are functional, a very
nice position to be in, and a very hard one to arrange).

Intel also has about an order of magnitude more clean room space capable of
making the Pentia (and Merced and Gunnison, etc.) than AMD has in Fab 25.

("Fab 25," by AMD's naming convention, means the fab opened in the25th year
of business, not their 25th fab.)

>effort from having to make opcode decisions, but then they don't get first
>crack at the market.  AMD has a *Very* good chip in the K6, receiving much
>attention as being a serious competitor to Intel's chips.

Well, look at their profits, Ryan! Go to Yahoo and look at both their
earnings reports over the last, say, 5 years. Also look at their stock
patterns.

AMD is now trading at $18. Five years ago it was trading at the same price.
In fact, it's been a narrow range between about 20 and 30 for most of that
time, briefly blipping up to 40 before dropping back to the level it was
half a decade ago. In fact, it's where it was in 1983, 15 years ago. (Check
the charts.)

Meanwhile, Intel has moved from $15 to $72 (today's price) in 5 years, and
from something like $2 (or less, as the charts don't go back to '83 for
Intel), up a factor of 30 or more times.

Market caps are similarly skewed. Intel's market value is $120 billion,
AMD's is $2.5 billion, a factor of 50 times lower. (They were within a
factor of 3 of each other 15 or so years ago.)

The problem AMD faces is not the adequacy of its design, "borrowed" from
Intel, but its inability to compete in manufacturing costs.

Even as AMD struggles to get yields up, and struggles to invest in the next
generation of production equipment, Intel is building several new $2
billion fabs, all equipped with the latest equipment.

(Intel continued to book orders for production equipment through the last
mini-downturn in '95-'96, ordering equipment from Applied Materials, Lam
Research, and so on. Guess what? When AMD and other small fry decided it
was time to order, pronto, they found that the Applied Materials, etc.
production was already committed to go to Intel! Jerry Sanders, top guy at
AMD, cried "foul," but the fact was that Intel's enormous financial and
market position strength allowed it to plan ahead and order the equipment
and manufacturing space they'd need.)

I have no doubt that some companies will design in the K6, and even the
Cyrix version. I would if I were them, if only to put some pricing pressure
on Intel.

But the market share of Intel, at something like 90% (I won't dissect the
various segments, but they range from 80% to 95%), and the huge costs to
compete in the _next_ generation, really makes it almost impossible for
small fry like AMD and Cyrix to do anything significant. Even if their
yields were significantly higher than Intel's, which is technically
impossible, their lack of capacity limits the inroads they can make.

And my point is a larger, longer-range one. I haven't said I expect Justice
Department action this year, or even next.

But if Intel's Merced displaces workstation chips (the DEC deal, Sun to
work on a competitive Solaris for Merced, the H-P/Intel alliance, and
several other major deals), and if AMD and Cyrix are unable to make a dent
in Intel's market share for PC chips, and if the motherboard integration
continues (with Intel supplying motherboards that competitors can't readily
match the peformance and pricing of), then I expect a Democratic Justice
Department to move on Intel.

Of course, those who feel AMD and Cyrix are about to knock Intel out of its
present position have the best of all ways to vote their beliefs: by buying
AMD (or National/Cyrix) stock at the current low prices.

I wish you luck, really.

--Tim May



The Feds have shown their hand: they want a ban on domestic cryptography
---------:---------:---------:---------:---------:---------:---------:----
Timothy C. May              | Crypto Anarchy: encryption, digital money,
ComSec 3DES:   408-728-0152 | anonymous networks, digital pseudonyms, zero
W.A.S.T.E.: Corralitos, CA  | knowledge, reputations, information markets,
Higher Power: 2^2,976,221   | black markets, collapse of governments.
"National borders aren't even speed bumps on the information superhighway."