Chapter 10 - Networking: Market Competition 1981-1983
10.26 Ethernet Chips, Boundless Hope and Market Confusion
By October 1982, the progress of Ethernet semiconductor development had become generally known. At first there would be four Ethernet chips: Source of Ethernet technology
Exhibit 10.26.1 Ethernet Chip Partnerships
| Vendor | Semiconductor Foundry |
|---|---|
| Ungermann-Bass | Fujitsu |
| 3Com | Seeq (Silicon Compilers) |
| Xerox (DIX consortium) | Intel |
| DEC (CMC) | AMD and Mostek |
The availability of Ethernet chips had already driven the price of a controller from two printed circuit boards to one and from $3,000 to less than $1,000. The dataPBX will still be cheaper (roughly $650 per connection), but it did not enjoy the steeply declining cost curve of Ethernet, or of other network connections that could be reduce to a chip - that is software and logic becomes a chip. This advantage, when combined with the communication speed that could be supported at the connection, or device, of 10 million bits/second, not a paltry 19,200 its/second, gave Ethernet an ever-gaining edge. The dataPBX and CBX speed per connection seemed like enough in the paradigm of terminals connecting to computers, but not when the paradigm became computers to computers.20 A paradigm bursting into existence with the introduction of the IBM PC, a paradigm that will revolutionize what the Information Economy will look like.
The Ethernet chip will also create substantial early mover advantages since token ring chips from TI would not be available for an estimated two years; two years that will become four years. Token bus, a broadband design, required the use of RF modems, not the simple transceiver of Ethernet, and modems will preclude token bus from ever becoming a chip in time. So Ethernet vendors were about to seize a huge market advantage.
The market for LANs was projected to become a $1 billion somewhere between 1986 and 1990; depending upon whose forecast you believed; this up from a market, ranging between $50 and $100 million in 1982.21 The more optimistic forecast proved correct, implying a growth rate between 70-110% per year. An extraordinary growth rate that would propel those firms with established presence into real companies.
For the firms selling local area networks, especially Ethernet products, the lofty market estimates might have made raising money easier, but in the short-term it invited unmitigated competition. In 1982, estimates of the number of firms with announced products ranged as high as 200. There would be 60 Ethernet vendors by 1984.22 In addition to the sheer number of competitors, there were the complications of competing against the lumbering CBX and PBX vendors, the promises of the integrated CBX-LAN vendors, the dataPBX firms and dozens of proprietary local area networks extolling their special advantages. The endless debates of what were better baseband (i.e., Ethernet) or broadband (i.e., token bus) added yet another layer of confusion.23 In sum: chaos.
- [20]:
It wouldn’t have been because bit-mapped graphics and laser printers would have driven a personal computer into the market if IBM had missed the boat.
- [21]:
Datapro, Nov. 1982 c11-010-103, p. 103
- [22]:
Red Herring, 1984
- [23]:
The confusion of this debate already had UB introduce an Ethernet over broadband (UB) and token ring was seen as working with either.