Hi Geoff Nunn; Regarding what happened in the super
computer industry.
I was working at SCS in 1986, which made a mini-super
computer called the SCS-40. It did about 40million 64-bit
floating point operations per second, peak. It cost about
$1MM. The ratio commonly known as bang/buck was
therefore about 40flops/$-sec. (Flops = floating point
operations.)
Cray sold a super computer that did about 10flops/$-sec.
Their machines cost about $10MM each, and ran at a
rate of about 100million 64-bit floating operations per
second. Our machines were software compatible with
theirs.
Management figured we could steal market share from
them cause our machines gave 4 times as much bang/
buck. This is a large enough amount that it gets people
interested.
Most engineers don't bother to look much at the industry
they are in, instead they just keep their noses to the
keyboard and design what they are told.
But something happened that forced us all to look at
the market we were in.
The company brought in Unix workstations (made by
Intergraph), to design the successor to the SCS-40.
The workstations cost about $25,000 each, and could
do around 10million floating operations per second.
They used a processor from Fairchild.
Naturally we computed the bang/buck, and it turned
out that the workstations we were using gave about
400flops/$-sec. When we looked at how much the
computer we were designing cost, and how fast it
would perform, we knew we were doomed. The
integration wave that started with the Intel 4004
was going to sweep through our section of the
market in a year or two, many years after it had
swept through the low end computer market.
Sure there was some elasticity in demand. But not
enough to counteract a 10x decrease in price. The
PC industry is in the same fix. Sure there is some
elasticity in price, but nowhere near enough to save
the company's butts that are in the business.
So give me the figure. How many more computers
can you sell given that the price drops 10x?
Everybody I know who might possibly want a
computer already has it. So maybe 2 or 3x as
many? That leaves the industry with a 3 to 5x
reduction in revenue. Dead meat. And it doesn't
take a degree in accounting, marketting, or
communications to figure that out. The
consequences are obvious to those who
understand the technology, and have any sort
of clue how markets work. But if you don't
understand the technology, then the fall in
ASPs is not obvious. Mumbling about "value
added" isn't going to help either. Walmart
runs on mighty thin margins, and so will the
PC industry once it goes commodity.
I really don't want to have to argue against the
quaint notion that the PC industry can expect
people to kindly purchase more expensive machines
than they really need, thereby keeping the box
makers in business. This has never happened in
any industry before, and I don't see it happening
over the next few years. People just aren't that
stupid, and an investor certainly doesn't want to
rely on that sort of logic.
People are still going to be in denial about ASPs.
The guys who said PC prices couldn't go lower
than $600 are already off by a factor of 2. The guys
who will say that $300 is the lowest possible price
will be shown to be wrong later. Look for a price
about $50 more than a low-end TV set, and watch
for monitor prices to drop drastically when industry
gears up for digital TV.
-- Carl |
