Forget Moore's Law

[R. A. Hettinga]

One would just as soon forget punditry, one would think. Or at least
forgotten pundits...

Cheers,
RAH
------

http://www.redherring.com/insider/2003/02/moore021003.html

Red
Herring

Essay 
Forget Moore's Law 
Because it's unhealthy. Because it has
become our obsession. Because it is dangerous--a runaway train, roaring
down a path to disaster. 
By Michael S. Malone 
February 10, 2003 

Forget
Moore's law, but not because it's merely a commitment by the semiconductor
industry to drive silicon gate technology forward 67 percent a year.


Forget Moore's law, but not because it isn't true. On the contrary,
Moore's law may be the truest "truth" about human events over the last
half- century. In fact, while our understanding of the cosmos, particle
physics, and brain chemistry gets revised almost by the month, Moore's
law--so fragile, so on the razor's edge of knowledge, so at the mercy of
human weakness--clicks on with the precision of an atomic clock. Indeed,
there are days when Moore's law seems the only thing we can still believe
in. 

Forget Moore's law because it is unhealthy. Because it has become our
obsession. Because high tech has become fixated on it at the expense of
everything else--especially business strategy. It is precisely this
fixation, at the cost of other considerations like profit, product, and
market, that led to the dot-com bubble and bust. 

Forget Moore's law
because there are more important things to worry about--like restoring the
lost vitality of the electronics industry. The only people who ought to be
obsessing about Moore's law are the folks working in the semiconductor
industry, and Gordon Moore himself has suggested that even in the chip
business his law hasn't always been a helpful fixation. Lately, some
disturbing new trends support his case. 

But most of all, forget Moore's
law because it has become dangerous. It is a runaway train, roaring down a
path to disaster, picking up speed at every turn, and we are now going
faster than human beings can endure. If we don't figure out how to get off
this train soon, we may destroy an industry. 

* * * * 

An extraordinary
announcement was made a couple of months ago, one that may mark a turning
point in the high-tech story. It may be the single most important news item
in the digital world since the implosions of Cisco Systems and Yahoo almost
two years ago. 

It was a statement by Eric Schmidt, CEO of Google. His
words were both simple and devastating: when asked how the 64-bit Itanium,
the new megaprocessor from Intel and Hewlett-Packard, would affect Google,
Mr. Schmidt replied that it wouldn't. Google had no intention of buying the
superchip. Rather, he said, the company intends to build its future servers
with smaller, cheaper processors. 

Most who know Google dismissed Mr.
Schmidt's words as yet another example of the company's notoriously
militant not-invented-here mentality. The company has always assembled its
own motherboards. Hell, Google would probably grow its own silicon crystals
if it could. Meanwhile, computer-industry analysts saw Mr. Schmidt's
remarks as a potential death knell for Sun Microsystems, Cisco, and,
ultimately, for PCs as well. If one of the world's largest consumers of
routers has decided to step off the express train of ever-larger CPUs, the
implications could be devastating. 

But few people realized that Mr.
Schmidt's announcement was even more far-reaching than that. In essence,
what he said was that Google, the hottest young company in technology, had
committed the ultimate apostasy: it had declared its independence from
Moore's law. 

Not completely, of course, because that would be impossible,
but a selective independence. Often forgotten is that Moore's law has three
variables: price, density, and performance, each of which contributes to
the 100 percent improvement the law promises every couple years. But in
practice, technology leaders have always stuck with the density track,
adding ever more computing power to their products for the same price.
Thus, each new generation of Pentiums and Athlons inevitably leads to new
generations of PCs, routers, and game machines. Nobody gets excited that
8088 prices have now fallen to, say, 50 cents, making cardboard box cameras
12 cents cheaper--after all, that's 1974 technology. 

But Mr. Schmidt's
announcement changes all that. "We aren't interested in getting maximum
power for a high price," he says. "What we're looking for is maximum
functionality and that's a whole different thing." Each of Google's
thousands of motherboards (a computer's main circuit board) are designed
for the quick switching of components. Even the power supply is held on
with Velcro straps: if it burns out, it can be replaced quickly. Recently,
when the expensive top-end disk drives used by the motherboards proved
inadequate, Google tossed out thousands and replaced them with cheaper,
better models. 

"It's all about what works," says one Google technician.


* * * * 

One person who understands the implications of Mr. Schmidt's
announcement sits in the Hobee's restaurant in Sunnyvale, California,
stabbing at his tuna salad. 

Probably no one over the last decade has been
a greater beneficiary--or been more at the mercy--of Moore's law than Marc
Andreessen. As the young cofounder of Netscape Communications, he was the
Internet's first superstar, the poster boy of the Web's genesis. Netscape's
1995 IPO made Mr. Andreessen rich and set off the dot-com boom. 

But then
Netscape, the epitome of a Moore's law company, ran into the marketing meat
grinder of Microsoft and was left on the slaughterhouse floor. Shortly
after that, Netscape was snapped up by that other marketing-over-innovation
giant, America Online. Mr. Andreessen put in his time with Case & Company
in northern Virginia, but by 1999 was back in the Valley, this time at a
new startup called LoudCloud, a Web-based managed services company. 

After
an initial flurry of attention, LoudCloud disappeared from the public eye,
a victim of both less-than-Netscape growth and the dot-com bust. Few even
noticed when the company sold its services business to Electronic Data
Systems last August and changed its name to Opsware. 

But Mr. Andreessen
doesn't mind lying low or that the working folks in the booths around him
don't recognize him. After the roller coaster of the '90s, he too has
stepped off the ride. These days he is moving quietly and thinking small.


Mr. Andreessen's epiphany about Moore's law came at the same time as Mr.
Schmidt's. "When I read the Google announcement," he says, "I understood
exactly where Mr. Schmidt was going. This is a fundamental, even
revolutionary, change in the IT world, and most people don't even realize
it yet. It's going to be disastrous for a lot of the big companies out
there." 

In selling the service side of LoudCloud and stripping down to,
in Mr. Andreessen's words, "a software design house building the glue to
hold all these thousands of low-cost servers together," Opsware has
reconfigured itself for a post-Moore's law world. Its software is designed
to automate everything that a server farm with endless racks of disposable
CPUs would need, including deployment, scaling, changing, and reallocation
of operations. 

"The rules of this business are changing fast," Mr.
Andreessen says, vehemently poking at his tuna salad. "When we come out of
this downturn, high tech is going to look entirely different." 

* * * *


A well-known rule about doing business in technology is that nobody ever
bet against Moore's law and won. It is a rule proven true a thousand times
each year. Over the last 40 years, a score of brave and brilliant, but
foolhardy, companies have tried to get ahead of the law: Trilogy Systems
and MicroUnity with their superchips and IBM with X-ray lithography. All
have failed, often spectacularly, losing billions of dollars in the
process. 

Less obvious, and less interesting, are the tens of thousands of
tech companies that trail the law and finally opt out of the chase
altogether. Theirs is the world of weary industrial parks, forgotten trade
shows, and unmotivated employees waiting for their stock options to vest.
Their fate is the long, slow--and often quite comfortable--slide to
inconsequence and oblivion. 

But few have pointed out the fate of those
companies that choose to live by Moore's law alone. These are the creatures
of the booms and bubbles. We saw thousands of them four years ago--all of
them are dead now. 

After all, what was the dot-com bubble but an army of
would-be entrepreneurs who saw a new industry emerging-- e-commerce--and
founded companies with the sole strategy of intersecting that industry in
12 to 18 months at the point where Moore's law said it would be. Everything
else--products, customers, revenue, and profits--was secondary to hitting
that moving target, thus dominating the chosen market niche. 

Abetting
this single-minded chase were the venture capitalists, themselves fixated
on Moore's law because of the simplicity of its arithmetic compared to the
messy, unpredictable human side of business. And egging on the whole
delusion were the stock markets and their millions of investors, willing to
award the most impossible price/earnings ratios to those startups aiming at
the most distant point of the law's trajectory. 

It was all a delusion, we
now know, just so much hocus-pocus made momentarily legitimate by its
proximity to the unassailable law. Moore's law did hold for the
Internet--it still does--but the market's narrow-minded obsession over it,
and thus its ultimate abuse, only served to lead us astray and into
disaster and recession. 

The dot coms weren't alone. The same thing
happened in telecommunications, where a messianic belief in
"bandwidth"--the automatism of Moore's law in a different guise--led to one
of the most expensive low-return build-outs in business history. And it is
about to happen again in biotechnology, where variants of Moore's law are
already popping up in such fields as bioinformatics and biochips. 

* * * *


There is a once-famous, but now nearly forgotten, statement made in the
early '60s by an executive at Texas Instruments. The executive said that
when the electronics age was over and all the profits and losses were added
up, the bottom line would be red. 

The electronics age is nowhere near its
end. The Semiconductor Industry Association's Technology Roadmap--yet
another product of Mr. Moore's genius--shows few technological obstacles to
the pace of change in semiconductors (i.e., Moore's law) until at least
2010. If the past is precedent, only a fool would bet against the chip
industry maintaining its pace well into the middle of this century. 

Even
now, at trade conferences, Advanced Micro Devices, IBM, and Intel calmly
discuss the advent of terahertz chips, three-gate transistors, and
90-nanometer design features. Meanwhile, researchers at the University of
Wisconsin at Madison have announced single-atom memory storage. 

Yet even
as the pace of newly announced technological wonders continues unabated,
there is a growing sense among many high-tech thinkers that there are too
many pieces on the chessboard of Moore's law, that the pace of business and
society can no longer keep up with the physics. 

The dot-com bubble, in
which thousands of new companies, well-versed in Moore's law, saw where
they needed to go but failed to make it, taught us an important lesson
about the growing disconnect between the pace of change and our ability to
cope with it. But the warning has been there for a quarter-century. 

After
all, corporate IT may moan about its legacy problems, but the electronics
industry as a whole has the longest legacy trail of any industry in human
history. Few people, even in technology, realize that the original 4- and
8-bit microprocessor architectures of the early '70s are still being built
by the billions each year for use in everything from cameras to toys. We
may be focusing on Pentium 4s, but its great-great-great-great-great
grandfather, the 8008, is, quantitatively speaking, a greater part of our
daily lives. 

What's more, we haven't even spun out all of the potential
applications of the 8080, Z80, or 6502, much less the 80386, DEC Alpha, or
the first PowerPC. And that suggests that in its relentless pursuit of the
power curve of Moore's law, at the expense of the much less interesting
price curve, the entire electronics industry may be unknowingly trapping
itself in a giant and attenuated version of the dot-com bubble--living
years into the technological future, only to be painfully snapped back to
the human and economic present. 

That's what haunts Mr. Schmidt, even as
he currently leads the hottest company in technology. It is also why he has
privately circulated his thoughts on this very subject. After looking at
precedents in the electrical and railroad industries in the early 20th
century, the commoditization of ever-larger portions of the computer world,
the lagging world of applications, the slowing demand for upgrades of
existing technologies, the shortage of apparent new "killer apps," and the
soaring costs of innovation, Mr. Schmidt soberly concludes: "Info-tech is
absolutely indispensable, but not that great a business." 

He isn't alone.
Industry watcher Donald Luskin noted earlier this year that even Intel is
finding itself being slowly crushed by Moore's law. He pointed out that
just to keep its revenue level, Intel must convince its customers to double
their power every 18 months or to stick with its current offerings and find
twice as many customers. 

That was a lot simpler five years ago, when the
economy was strong, much of the market was still untapped, and wafer fabs,
which double in price every four years (jokingly called "Moore's second
law"), were a lot cheaper. 

"That's why Intel's revenue growth just
imploded, even as they ship record volume," said Mr. Luskin. "In this deep
recession, Intel just can't keep up with the law named after its founder."
Mr. Luskin wasn't talking about ever-more-powerful Pentiums--Intel can do
that--but ever-hungrier customers. Even Intel can't manufacture them. 

As
Mr. Schmidt points out in his notes, with Intel's research and development
costs doubling every 18 months (apparently R&D follows Moore's law as
well), in another 20 years the company's R&D costs will be $31 trillion
annually. Something must give long, long before then. 

But give the last
word to Mr. Moore himself, who once said, "Obviously, you can't just keep
doubling every couple years. After a while the numbers just become absurd.
You'd have the semiconductor industry alone bigger than the entire GDP of
the world." 

* * * * 

Even as we emerge from the devastation of this last
tech bust, we may well find ourselves facing an even greater and more
dangerous challenge. Some of high tech's brightest minds are already
looking for answers, but until now they've only begun to ask the right
questions: How do we deal with a world in which the old financial models no
longer work? What new business model do we put in their place? Is there one
that still offers that chance for entrepreneurship and explosive growth
that has made high tech the most thrilling of all industries? How do we
disprove--or at least postpone--that cruel prediction about the fate of
tech made many years ago by that Texas Instruments executive? 

We cannot
escape the rule of Moore's law, nor should we want to. It has brought us
untold blessings, and it will bring us many more in the years to come. But
even as we celebrate Mr. Moore's brilliant law, we must also free ourselves
of its tyranny. Before it's too late, we somehow must put Moore's law in
its place. We must learn to work beyond it, or at least in spite of it.


Somehow, even as we dance to its beat, we must learn to forget it.


Michael S. Malone is a former columnist for the New York Times, former
editor of Forbes ASAP magazine and the author of The Microprocessor: A
Biography (Springer-Verlag, 1995). 


-- 
-----------------
R. A. Hettinga <mailto: rah@ibuc.com>
The Internet Bearer Underwriting Corporation <http://www.ibuc.com/>
44 Farquhar Street, Boston, MA 02131 USA
"... however it may deserve respect for its usefulness and antiquity,
[predicting the end of the world] has not been found agreeable to
experience." -- Edward Gibbon, 'Decline and Fall of the Roman Empire'