March 17, 2003

Moore's Law Comes to Telecom

Why are WorldCom's physical assets worth only 1/4 what they paid for them? Some people blame incredibly stupid management. I have always ascribed the write-down of the market value of telecom assets to vastly greater progress in telecommunications software control than had been expected--if you can push ten times as many bits through a fiber as you had thought you would be able to, that kinda cramps your ability to make money off of the scarcity of fiber. Kevin Werbach and Scott Rafer seem to have a somewhat different take, but I'm not quite sure what it is.


New York Times: "WorldCom's hard assets, including its network, are now worth almost 75 percent less than what they had cost." UPDATE: Scott Rafer's take on the same announcement. He thinks we're ascribing different causes to the writedown, but I think we agree. Worldcom's problem is that it didn't build its business (and capex plans) with Moore's Law in mind. [Werblog]

Posted by DeLong at March 17, 2003 07:16 PM | TrackBack

Comments

It's not simply a matter of the scarcity or abundance of "fiber". It's (as Rafer notes) the difference between "smart" and "dumb" networks. MCI (and, indeed, all the telecoms) have invested heavily in the former (for which demand is stagnant) while demand for the latter remains robust.

Just today, the Austin American Statesman carried an article about the rising popularity of voice-over-IP.

The observation that "smart" networks (access to which one presumably hopes will command premium prices) will inexorably be trounced by "dumb" ones (the internet, where bandwidth carries commodity pricing) dates back to David Eisenberg (1977)

http://www.isen.com/stupid.html

and has been reiterated recently by Searls and Weinberger

http://www.worldofends.com/

Actually, this tension predates even the internet itself, as is engaginly retold by Paul Kunz's "Bringing the Web to America"

http://www-project.slac.stanford.edu/interlab99/program/kunz-web-to-america.htm

(see the part about the role of the German and French PTT's).

Anyway, whenever a company (or a whole industry) bets heavily on the wrong horse, there's bound to be much wailing in the grandstands.

Posted by: Jacques Distler on March 17, 2003 09:19 PM

Ususal disclaimers apply--I am not an expert in telecoms, and while this isn't exact, I think it is right, in principle.

This isn't just a software issue--the hardware improved tremendously ina couple of ways. Capacity used to be governed by the amount of cable. The advent of multiplexing made it possible to increase the capacity of fibre tremendously.

If you were Wolrdcom, you layed a tremendous amount of line 1) assuming that the .com thing was real and 2) because you needed the capacity. Then the DWDM people multiply, in 2 years, your capacity by 16 or 128. So you buy the multiplexers---and the market collapses. No more irrational than the rest of the boom, and all that capacity is useless.

The DWDM was a hardware advance--and one that happened very, very, very fast--it wasn't just Moore's Law.

I know Moore's Law is used as a proxy for all really fast technical advances, but it really isn't the only driver. Storage cost ($/MB) decreased much faster than Moore's law, for entirely different technical reasons. Fiber capacity increased, again for very different reasons.

Put another way, why would the physical size of a transistor impact the size of a magnetic bit, or the efficiency of a laser?

B

Posted by: Brennan Peterson on March 17, 2003 09:25 PM

Off topic, but relevant in the coming years: (think Biological Revolution, not Information Revolution).

The Whole Biochip: Your Genome for Less than $1

Just as the microprocessor enabled the cost of computation to drop from $75 to less than a billionth of a cent for a million computations between 1970 and 2002, so too will the cost of sequencing an individual's genotype drop through the floor over the next twenty years.

With the completion of the Human Genome Project just three years behind us, the cost has already fallen from approximately $3 Billion to sequence an individual's genome to several hundred thousand dollars today. Recent claims have been made that a $1000 genotyping system could be possible in the next 5-10 years.

Looking ahead further it is possible to imagine sub-$1 genotyping systems as the market for such services grows extremely large (i.e. all of humanity and beyond), driving competitive pressures up and prices down. This type of cost decrease will need to happen in proteomics as well for the Whole Biochip to emerge.

I wonder when genotyping will become all the rage as a Holiday gift? 2014?

Posted by: Zack Lynch on March 17, 2003 09:57 PM

I think Brennan and Zack are right. This is about much more than chip-imaging. There is a whole technological revolution out there, and it's getting faster, faster.

The economist used to go on and on about how the internet was bringing down transport and communication costs. They were thinking about e-commerce and B2B. But perhaps the really big impact of the internet is going to be in information storage, access and sharing. Think distibuted computing SETI style. Think biological as Zack says. Think weblogs, communities, Brad's comments section. The cost of upgrading your own human capital has come down a lot. (Should the corporate value of the business schools get a WorldCom style write-down? Shouldn't they go for the 'dark cable' model, not the 'smart kind').

Doesn't this have implications for eg the German 'reform' process. The whole R&D situation has just changed dramatically. The EU is still thinking large scale, state-subsidised research. Isn't the key to change in Germany about motivating a new generation of young people with different values. Isn't it time they woke up to the fact that 'sharing' is about more than MP3. Wouldn't they do better subsidising high speed internet for the young people, and actively promoting on-line collaborative communities?

On the economic consequences of all this, didn't Brad get it right in his Jacksons Hole piece a couple of years ago: it all depends on relative elasticities and growth in market scale.

Posted by: Edward Hugh on March 17, 2003 10:33 PM

Yes, Dense Wavelength Division Multiplexing (DWDM) has had a tremendous impact on capacity (more than I realized).

But it's a trend that Worldcom and co. should have seen coming (as it was introduced in the mid '90s, and has steadily gained in capacity since then).

But (excuse me if I have misunderstood) it does not devalue existing fiber optics networks the way Moore's law devalues existing microprocessors. You can always retrofit your network with the latest DWDM multiplexors, and the same old fiber will have much greater capacity.

But the existing fiber is already a sunken cost. Whether you decide to pony up for new DWDM multiplexors (surely cheaper than laying more fiber), depends on how much you hope to be able to charge for the increased bandwidth of your network.

That's where the "smart" versus "dumb" (or "premium" versus "commodity") elment comes in.

Posted by: Jacques Distler on March 17, 2003 11:14 PM

"Put another way, why would the physical size of a transistor impact the size of a magnetic bit, or the efficiency of a laser?"

I imagine that there are general scientific abilities which affect all three. Just being able to view and manipulate matter at a particular smallness of scale is required before one can even contemplate research to improve any of those. In other words... there's something about a 200MB hard drive that corresponds to a 486 25Mhz processor and there's something about a 200GB hard drive that corresponds to a P4 2Ghz.

Posted by: snsterling on March 18, 2003 01:38 AM

Fiber isn't exactly scarce. There's tons of "dark fiber" all over the USA. Lester Thurow did a good lecture on this very subject, you might like it:

http://web.mit.edu/mitworld/content/economics/thurow.html

Posted by: Charles on March 18, 2003 02:01 AM

"The EU is still thinking large scale, state-subsidised research. Isn't the key to change in Germany about motivating a new generation of young people with different values. Isn't it time they woke up to the fact that 'sharing' is about more than MP3. Wouldn't they do better subsidising high speed internet for the young people, and actively promoting on-line collaborative communities? "

I am in complete agreement with Brad De Long's theory. Almost certainly, Moore's Law was ignored in World Com's business plan. The old copper lines were of use for many years. This definitely is not the case for the fiber laid just a few years ago. The Germans and the other Old Europeans are foolish if they continue down the Socialist path. However, My ancestors are too far gone to wake up to reality.

Posted by: David Thomson on March 18, 2003 03:07 AM

DWDM doesn't appear to be a good explanation; the timing seems wrong. I have a post at http://mydoxy.blogspot.com

Posted by: matt wilbert on March 18, 2003 05:43 AM

Don't forget, WorldCom was the SOURCE of the myth about bandwidth demand doubling every hour, or some such nonsense. How could they not assume that even with DWDM and other technologies there would still be demand for more cable?

Similarly we also know that WorldCom's accounting was fantastically corrupt, hiding the very losses from overcapacity that would have demonstrated the folly of laying more cable. But much of their stratosperic market valuation was based on (their) estimates of future capacity demand, coupled with their large installed network. So if the market thinks cable is profitable (because you lied) and cable demand will ever rise (because you lied), what's the best way to increase the stock valuation? Lay more cable.

Result: dumb money, due to problems largely of their own creation.

Posted by: Ethan on March 18, 2003 08:05 AM

...or, to use the rather morbid cliche from the period,

"the fools - they drank their own Kool-aid."

Posted by: Ethan on March 18, 2003 08:11 AM

I'd like to elaborate on the "smart networks" explanation for what drove the telecoms to madly overbuild capacity.

Say I have a client who wants to send data from Houston to Cleveland. If I want to sell him "smart" network services, my data network has to extend to both places (and thousands of others too).

On a dumb network, that's unnecesary. His data could be handed off to different network providers 3 or 4 times on its journey (that's what the "inter" in "internet" is all about). My network doesn't need my network to be ubiquitous; I can peer with others who serve locations that I don't.

But if I and my competitors all want to sell "smart" network services, we're all going to try to lay fiber to everywhere, leading to ridiculous overcapacity.

A similar problem afflicts the airlines too ...

Posted by: Jacques Distler on March 18, 2003 08:30 AM

"Put another way, why would the physical size of a transistor impact the size of a magnetic bit, or the efficiency of a laser?"

I think some of you misunderstand what DeLong means. He's not actually referring to "Moore's Law" in the shrinking-transistor-size sense, but rather he's engaging in the increasingly common practice of using the label as a shorthand way to describe any kind of technological advance in which some desired quantity (e.g. transistor count, network capacity, bandwidth, etc.) doubles per unit cost every so many months.

This usage is actually quite a stretch, and I wrote an article a while back at Ars Technica that was intended to provide a corrective. However, I doubt it'll have much of an impact in the long term. People will continue to talk about "Moore's Law of Bandwidth" or "Moore's Law of Blue Lasers" etc.

I actually think this is ok, as long as you don't take it too far by changing it from a shorthand/descriptive term for bundling together a collection of trends in one area under the heading of a single heuristic curve into an actual "law" or driving force that pushes "progress" in an area. The most egregious example of such specious reification that I've seen is here:

http://www.corante.com/mooreslore/

This "Moore's Law of the Kitchen Sink" mentality is characteristic of a kind of futurist tech boosterism that looks for large-scale economic an technological trends that can be aggregated, simplified, reduced to a handy slogan or catch-phrase, and then marketed as indicative of the Next Big Thing. This way, the person who said "I told you so" can sell more books and get more speaking engagements.

At any rate, increases in network capacity that come not from advances in fiber but advances in signalling tech can sometimes be indirectly related to Moore's Law (in the "original" feature size sense), but I don't think that these kinds of connections are what DeLong had in mind.

Posted by: Jon Stokes on March 18, 2003 10:07 AM

How does one define overcapacity or undercapacity given the way the technology works?

In the year 2000 it was possible to calculate with widely publicized figures and technology trend rates that by 2010 only one network of similar cost would be needed to deliver simultaneously more than one streaming DVD quality video per person on earth. Of course this is ridiculous as it counts people without electricity and ignores that most data is transmitted locally (video on demand and videophone are mostly local transmissions).

So how do you account for this with methods like depreciation when it is not the fiber that loses value but the ever cheaper and better accessories which amplify the supply? The value of an installed fiber network is pretty much zero if there are many firms and nonzero only if there is monopoly or oligopoly.

Posted by: snsterling on March 18, 2003 10:51 AM

I've sold controllers for many years to machine builders in the semiconductor fab, disk-drive fab, and photonics (fiber) fab industries. As such, I've been in the middle of the common thread of these industries, the requirements for ever tighter positioning tolerances in the manufacture of these devices.

I've gotten used to the "gradual" tightening of tolerances the semiconductor and disk-drive fab people have demanded as chip and disk features have gotten finer over the years.

But I was blindsided by what the fiber multiplexing guys were doing in the late nineties. I remember a call I got one day out of the blue where they said, "We've gotten 3-picometer repeatability with your controller, but we've got to get it down to 1-picometer, so we've got some detailed questions about your algorithms." They explained that the tighter they could hold their wavelength tolerances, the more information could be pumped down a given fiber.

Nobody else in any industry was talking repeatability finer than a nanometer (10^-9 meters), and a picometer (10^-12 meters) is a thousand times finer. The thought still boggles me.

Posted by: Curt Wilson on March 18, 2003 12:51 PM

The difference between Kevin's and my positions on this was never stated clearly. Kevin believes that WorldCom's equipment has decreased in value rapidly due to Moore's Law. On the other hand, per Isen's beliefs, I think that the equipment was NEVER worth anything to WorldCom's shareholders, because WorldCom could not have made money offering services above Level 3 in OSI jargon. The mathematic difference between rapidly declining to zero, and dropping immediately to zero is not a lot, but it does have investment implications for new networks being built now (such as Wi-Fi, Wi-Max, etc.).

Posted by: Scott Rafer on April 20, 2003 09:50 PM

The difference between Kevin's and my positions on this was never stated clearly. Kevin believes that WorldCom's equipment has decreased in value rapidly due to Moore's Law. On the other hand, per Isen's beliefs, I think that the equipment was NEVER worth anything to WorldCom's shareholders, because WorldCom could not have made money offering services above Level 3 in OSI jargon. The mathematic difference between rapidly declining to zero, and dropping immediately to zero is not a lot, but it does have investment implications for new networks being built now (such as Wi-Fi, Wi-Max, etc.).

Posted by: Scott Rafer on April 20, 2003 09:51 PM

The difference between Kevin's and my positions on this was never stated clearly. Kevin believes that WorldCom's equipment has decreased in value rapidly due to Moore's Law. On the other hand, per Isen's beliefs, I think that the equipment was NEVER worth anything to WorldCom's shareholders, because WorldCom could not have made money offering services above Level 3 in OSI jargon. The mathematic difference between rapidly declining to zero, and dropping immediately to zero is not a lot, but it does have investment implications for new networks being built now (such as Wi-Fi, Wi-Max, etc.).

Posted by: Scott Rafer on April 20, 2003 09:56 PM
Post a comment