the boxes. They want to be able to offer many models—fancy, expensive ones for gadget buffs and simple ones for other folks. If the network company supplies the boxes, there is no profit for the retailers. The cellular telephone industry solved this competition with a partial subsidy: You acquire your cellular phone from any retailer, but the price is partially underwritten by the cellular company that you commit to buy service from.
Cable and phone industries will be the primary, but not the only, competitors to provide the network. Railroad companies in Japan, for example, recognize that the rights-of-way they have for their tracks would be ideal for long fiber-optic cable runs. Electric, gas, and water utilities in many countries point out that they, too, run lines into homes and businesses. Some of them have argued that the energy savings from computerized management of home heating alone might defray much of the cost of stringing fiber-optic cables because energy demand would be lowered, which reduces the need for expensive new generating plants. In France most of the cable TV connections are owned by two big water companies. But outside France, at least, conventional utility companies seem less obvious candidates for building highway connections.
You may wonder why I haven’t mentioned direct-broadcast satellites and other technologies as mainstream competitors of the telephone and cable companies. As I said earlier, current satellite technology is a good interim step. It delivers a great broadcast video signal, but there would have to be a major technological breakthrough before it could provide a unique video bandwidth feed to every television set and PC. For the United States market it would have to go from today’s 300-channel-per-satellite system to a 300,000-channel system, even assuming fewer than 1 percent of the displays needed a unique feed simultaneously.
Because these satellites also have a problem delivering data from the home back into the network (the back channel) to provide true interactivity, applications such as videoconferencing aren’t possible. A partial solution is to use the phone for the back channel. Direct-broadcast satellites such as the Hughes Electronics’ DIRECTV system use your regular home telephone line to submit back to their billing center a record of any pay-per-view programs you have chosen. With a special add-in circuit, direct-broadcast satellites can send data to PCs as well as to television sets. Data broadcasting is a valuable interim conduit for some applications.
Teledesic, a company that my friend, cellular telephone pioneer Craig McCaw, and I have invested in, is working on overcoming the limits of satellite technology by using a large number of low-orbit satellites. The scope of the proposed system is quite ambitious. It involves nearly 1,000 satellites orbiting fifty times closer to Earth than traditional geo-stationary satellites. Being closer to Earth means these satellites require 2,500 times less power and have increased two-way channel resources. This solves the back-channel problem. Also, the significant transmission delay associated with satellites is also overcome. Over long distances these low-orbit satellites can provide transmission speeds comparable to those available on fiber. Teledesic has regulatory, technical, and financial challenges and it will be several years before we’ll know whether the company can overcome them. If it can, Teledesic or other systems like it may be the first, the cheapest, or indeed the only way to bring the highway to many parts of Earth. Most of the population of Asia and Africa, for example, is unlikely to have local access to fiber connections within the next twenty years.
Another rapidly advancing technology is ground-based wireless communication. Television signals, which have been broadcast over the air using wireless VHF and UHF, will be carried primarily on fiber. The purpose of this change is to enable everyone to have a personal video feed and interact. Meanwhile, voice and other low-data-rate connections are moving from the wired infrastructure to wireless transmission in order to support increased mobility. The ideal system would allow for the sort of personalized high-quality video and mobility I talked about having with a wallet PC. So far, that combination cannot be supported by any of today’s technologies, because wireless systems can’t provide the bandwidth for individual video feeds that a fiber network can.
Early on, competitors will race to supply the first interactive services to communities, but once all of the attractive territories are served by one company or another, rivals will begin head-to-head competition by entering markets already served by others. Interestingly, in the cable TV business, the few places a second system was constructed, the “overbuilder” never made money. Having two or more general-purpose connections running into every home would aid competition, but the extra cost is immense.
Servers for the information highway will have to be large computers with gigantic storage capacity that run twenty-four hours a day, seven days a week. Competition to supply them will be intense. Various companies have different ideas about the right design for the servers, and strategies for developing them. Not surprisingly, the positions held by the various potential competitors are influenced by their areas of expertise. If your only tool is a hammer, pretty soon every new problem starts to look like a nail that needs pounding. Mini-computer companies such as Hewlett Packard envision using clusters of mini-computers as servers. A variety of companies that primarily make personal computers believe that inexpensive PCs, connected together in large numbers, will prove to be the most cost-effective and reliable approach. Mainframe specialists such as IBM are adapting their big machines to be servers. They cherish a fond hope that the information highway will be the last bastion of Big Iron.
Software companies naturally see their product as the answer. Software is so inexpensive to duplicate that substituting it for costly hardware reduces system costs. Another competition is shaping up to supply the software platforms that will run these servers. Oracle, a database-management company that makes software for mainframes and mini-computers, envisions the server as a supercomputer or mini-computer running Oracle software. AT&T, with its experience in the network business, will probably try to embed most of the system’s intelligence in the servers and switches of the network and put relatively little processing power in information appliances such as PCs and set-top boxes.
At Microsoft, our only “hammer” is software. We expect that the highway’s intelligence will be evenly divided between servers and information appliances. This arrangement is sometimes called “client/server” computing, which means that the information appliances (the clients) and the servers will run cooperating software applications. We don’t believe that giant supercomputers, mainframes, or even clusters of mini-computers will be necessary. Instead, Microsoft, like many of the PC manufacturers, sees the server as a network of dozens to hundreds of what are essentially personal computers. They won’t have the familiar cases, monitors, and keyboards, and may be housed together in large racks at the headquarters of a cable system or the central office of a telephone system. It will take special software technology to harness the computing horsepower of thousands of such machines. Our approach is to make the coordination of the highway a software problem and then use the highest-volume (and therefore cheapest) computers to do the work—the same ones used in the PC industry.
Our approach focuses on taking full advantage of all the advances going on in the PC industry, including the software. The PC will be one of the primary devices used on the highway. We think the set-top box should share as many technical features as possible with the PC to make it easy for developers to create applications and services that will work with both. This will allow the Internet to evolve upward into the highway in a compatible fashion. We believe tools and applications available on the PC today can be used to build new applications. For instance, we think set-top boxes should be able to run most of the CD-ROM titles for PCs that will appear over the next decade. One could argue that we are thinking too narrowly by trying to imagine the new world in terms of the PC. But there are more than 50 million PCs sold every year worldwide. The population of installed PCs will provide a substantial starter market for a prospective developer of any application or service.