Main | April 1998 »

October 21, 1996

Backhoes Don't Obey Moore's Law: A Story of Convergence

“I need you; like the flowers need the rain, you know, I need you, I need you.”
     -America, I Need You

We would like to propose a theory that you may find either obvious or unimportant or both. This theory, which is actually borrowed from a friend at Teledesic, is that backhoes don’t obey Moore’s Law. Before we explain why this is important, let us first make sure everyone is on the same page.

Moore’s Law was first noticed in 1965, when Intel’s chairman (Gordon Moore) noted that the price/performance of memory chips was increasing by a factor of two every eighteen months. Moore’s Law has been the lifeblood of the computer industry. This consistent increase in processing power at a given price is a clear contributor to the insatiable demand for personal computers. As long as Moore’s Law continues, we will likely witness continual price-elasticity.

The backhoe is a piece of construction equipment that is used by several industries for general productivity. One industry that has been particularly active in the backhoe purchasing market is telecommunications. You see, backhoes are used to dig trenches, in which telephony companies lay cable – historically cooper, but more recently fiber optic.

Just in case you are uncertain as to whether or not backhoes obey Moore’s Law, we have done some research. This January, John Deere (DE, 43 7/8) will release an upgrade to its 310 series of backhoes, the first such upgrade in about six-years. We aggregated the improvement along six different metrics and came up with an improvement of 13.5% over approximately six years. Admittedly on one feature, the allowable weight in the dipper stick at maximum height, has increased 29.7%. However, this represents an annual improvement of 4.4%, which falls well short of Moore’s Law equivalent (59.7%).

By now you may be wondering where this is going. You hear a great deal about convergence – how the computer industry has collided with telecommunications to deliver this great medium we know of as the Internet. What you do not hear about is the downside of convergence. Convergence creates dependency. Hopefully you have followed the hints we are giving. The computer industry is now dependent on the telecom industry which is dependent on backhoes which don’t obey Moore’s Law.

Backhoe dependency is really just the simple side of our message. It is our impression that the majority of the players in the computer industry bring a “computer centric viewpoint” (CCV) when analyzing the issues that exist with the Internet. This computer-centric view could prove hazardous. Not only will it lead to disappointed expectations, but it may also lead to a less than accurate vision of the future.

We would now like to walk through five separate Internet issues, where we believe that the computer centric viewpoint is misleading. It is our goal to inform, provide a new way of thinking about some of these issues, and hopefully improve the accuracy of the general outlook. As financial analysts, we are extremely sensitive to the perils of unmet expectations.

HOW SMART IS THE INTERNET?
We suspect that many of you have heard that the Internet is this wonderfully flexible network that was built to withstand a nuclear war. You have also likely heard that the information that traverses the Internet is divided up into little packets that each travel along independent optimal paths. While we are not in a position to comment on nuclear wars, we will submit that packets on the Internet typically travel along the exact same path. As a result, the standard way of improving network performance is not through network optimization, but rather by providing raw bandwidth.

This has interesting implications in terms of how you measure Internet usage. Some people will tell you that there is no bandwidth crunch coming, based on the belief that only 10-20% of the Internet infrastructure is being used. The problem with this argument is that there is no way to effectively allocate this excess capacity to the areas of the Internet where the bottlenecks do exist.

THE VALUE OF INTERNET TELEPHONY
It seems to us that the majority of the players in the computer industry believe that Internet telephony is cheaper because it is digital rather than analog and because it bypasses the bureaucratic telecommunications companies. In case anyone has not told you this yet, Internet telephony is an arbitrage play. During long distance calls, the long distance provider is required to pay each RBOC involved 3.25 cents per minute. That accounts for 65% of the $0.10 per minute that Candice Bergen likes to promote.
EVERYTHING OVER IP
One thing that you continually read about in financial journals is that one-day all media types will traverse on a single network which will be some descendent of what we now know of as the Internet. This would offer the best of both worlds, the efficiency of a packet based network with the reliability and functionality of a circuit based network.

If something sounds too good to be true, it likely is, and we have no reason to believe that this case is any different. Different data types require different back-end networks. The satellite is great for broadcasting video, voice, and even data, but it is horrible at two-way communications. The telephone network is phenomenal at two-way communication, but it is horrible for video and Internet traffic is currently stretching the limits of the central office switch. Then there is the Internet. It is great for asynchronous communication. But then again, that is what it was designed to do. Force feeding synchronous connections over an asynchronous network will likely jeopardize the intended benefits of the IP network.

We envision a world where a single device connects our LAN or home to several best-of-breed back-end networks and allocates the data according to media type. We like this idea for two reasons. First, it fits with the generally accepted theory that intelligence should reside at the edge of the network. Additionally, it fits with reality. Ascend (ASND, 62) and Shiva (SHVA, 42 3/8) are separately working to deliver products that strip data traffic from telephony lines before it clogs the central office switch.

UNLIMITED-USE PRICING
We have run into several people that believe that unlimited use pricing is good for the Internet and good for the computer industry. We disagree whole-heartedly. When you allow unlimited use of a limited resource you suffer the “Tragedy of the Commons.” Freeloaders will use more than their fair share and actually decrease the overall value of the resource. If we really want this thing to go forward, we should put the use of the network in the hands of those who would find it most profitable.

We are certain that some people will find the theory of a move to usage-base pricing as inflammatory. Yet, all we are really talking about here is capitalism -- using price to properly allocate resources. Unlimited pricing, on the other hand, is socialistic. And history seems to suggest that socialistic policies have not had a positive impact on infrastructure development.

MINING FOR GOLD
Over the past eighteen months, many financial analysts (including ourselves!) argued that the way to make money on the Internet was to invest in the equipment companies, not the ISPs. This theory was based on the lesson that the real profits during the gold rush came from selling pick axes, pans, and blue jeans to the miners. The problem with this theory is that it requires that there be at least some gold in the mountain.

The companies that we are counting on to build out the infrastructure all face issues that could potentially limit their ability to deploy capital. The ISPs are experiencing significantly negative cash flows and have limited access to capital. The RBOCs may lose up to 40% of their operating profits as a result of access charge reform, came out on the short-end of the stick with respect to the Telecommunications Act of 1996, and their investors, unrealistically perhaps, expect a dividend. The cable companies are overly leveraged, have a poor service reputation, and are under assault in their core business from the satellite vendors. Lastly, the odds-on favorites, the long-distance companies, may be surprised when the fax, an inherently asynchronous data type, goes to the Internet. Fax may account for as much as 25% of their profitability.

We recently attended a conference where web masters were advised to acquire as many T1 lines as possible, based on the theory that there may be a local-loop T1 shortage in front of us. Now T1-hoarding seems a little drastic. After all, we can always lay more lines. But then again, laying lines requires backhoes, and backhoes don’t obey Moore’s Law.

Posted by Bill Gurley on October 21, 1996 at 08:00 AM | Permalink | Comments (0)

DISCLOSURE: The information contained Above the Crowd has been obtained from sources believed to be reliable but is not necessarily complete, and its accuracy cannot be guaranteed. Any opinions expressed herein are subject to change without notice. The author is a general partner of Benchmark Capital, a venture capital firm in Menlo Park, Calif. Benchmark Capital and its affiliated companies and/or individuals may have economic interests in the companies discussed herein. © J. William Gurley 2005-2006. All rights reserved.