Robert --
I'd like you and Zeev and jg and other technology-gifted minds to read the following article and contemplate the possibility the Merced push-back has less to do with the chip itself and more to do with Intel's over-all focus.
If processor speed isn't the cause of the bottleneck and bandwidth is, then wouldn't it behoove them to focus on these technologies? Say, on ATM over ADSL for the customer premise? Perhaps look for a way to install modems in boxes before they leave the factories?
I am a complete idiot when it comes to semiconductors, so be gentle when you respond.
This article was pulled out of my old Amati files --- so complete, I'm expecting the Smithsonian to come knocking any day. [USRX promise of V.Everything by mid-'97 especially noteworthy.] :))
Pat
<<<
Intel is running out of easy ways to expand microprocessor sales. It may have to resort to territorial aggression to keep its 50,000 employees busy.
Excuse me if I invade your business
By Nikhil Hutheesing
INTEL CORP.'S AGGRESSIVE MOVE in the networking business did a lot of instant damage to one company: The market value of 3Com Corp. fell $2 billion in just two days in February. But this could be only a small taste of what is in store over the next decade for companies situated dangerously close to Intel's microprocessor segment.
Intel gets most of its $21 billion in revenue, and almost all of its $5.2 billion in profit, from the processor chips that do the thinking for desktop computers. The core business has been so good Intel has had scant reason for branching out. But it has no scruples about invading other folks' turf. In 1993 it greatly expanded its business of manufacturing motherboards, the circuit boards that represent the guts of a PC. That move infuriated Compaq and other PC makers that buy chips from Intel, but they were powerless to stop it. Intel has also been stepping up its efforts to sell the helper chips that surround a microprocessor. There it competes with such firms as VLSI Technology and Opti Inc.
What did Intel do to hammer 3Com? It announced a price-cut in network adapter cards. These are the circuit boards that connect a desktop computer to a network cable.
Intel's professed intentions are pretty innocent-defensive, for that matter. Without fast and affordable networks people will be reluctant to buy the ever-faster microprocessor chips that Intel is trying to sell.
"Our motivation is to provide more bandwidth so that customers can take advantage of the processing power offered by the PC," explains Craig Kinnie, Intel's vice president in charge of the networking division.
Indeed, as George Gilder has pointed out (Forbes ASAP, Dec. 5, 1994), the bottlenecks in computing are not inside the box but in communication lines: connecting the PC to a colleague's machine on the next floor, or to an Internet site far away.
ÿ
Oracle Applications
ÿRemember the apologies Intel made when it moved more heavily into motherboards: We're just doing this to solve a bottleneck in getting Pentiums out into mass-produced computers.
It's like the tears Kodak has been shedding for small photoprocessing shops as it moves in on their business. Sort of a ritual bow to the victim before eviscerating him. Kodak says it has nothing against the small shops. It just wants to insure that people who buy its film and photo paper aren't disappointed with the results.
In its motherboard encroachment Intel was competing with its customers, the PC manufacturers. Compaq retaliated by switching purchase orders to Intel rival Advanced Micro Devices and refused to participate in Intel's "Intel inside" advertising. But Compaq needed Intel more than Intel needed Compaq. Compaq now buys more than 90% of its processors from Intel.
Intel's long-term strategy for keeping its processors busy is to feed more video and multimedia applications to PCs. "A few years from now every computer will be multimedia-ready and network-management ready," declares Andrew Grove, Intel's chief executive. "Computers without those things will be as meaningless as computers without memory."
So how is that strategy driving it into new businesses? Here's the problem facing the chipmaking giant. Today customers who buy Intel's new Pentium chip with MMX technology can get slick, full-motion video-but only if they can get data at 8 megabits per second. That is pushing the limits of the typical 10-megabit-per-second speed of corporate PC networks; two customers couldn't download different video files at the same time on a 10-megabit cable.
That's why Intel's Hillsboro, Ore. research lab has been working on networking equipment and networking software. 3Com invented the adapter card for the PC in 1981 and had the market largely to itself until Intel jumped in, in 1991. Today both companies make flexible adapter cards that run on either 10-megabit or 100-megabit Ethernet networks. According to market researcher International Data Corp., Intel has 36% of the sales of these dual-speed cards and 3Com has 40%.
On Feb. 5 Intel slashed the price of its 10/100-megabit card 40%, to $99. 3Com responded by cutting the price of its card 30%, to $94. "We have been doing this kind of stuff in microprocessors for 25 years," says Grove. "We start at the high end, and when manufacturing costs come down we cut prices and move the chips into the mainstream. We are applying the same technique to networking."
Understandably, 3Com takes a less casual attitude toward the price-cuts. "Intel's price-cut was very broad and extremely large, and that surprised us," says Eric Benhamou, 3Com's chief executive.
ÿ
"A few years from now every computer will be multimedia-ready and network- management ready. Computers without those things will be meaningless."
ÿ
ÿ ÿIntel's territorial imperative takes two forms. One is the marketplace push to keep its sales growing at the 32% annual rate of the past decade. The other is electronic. The more functions that can be consolidated on the master chip-or near it, on the motherboard-the faster and cheaper the computer is. That's why it is that so many companies (like Weitek) that used to sell ancillary chips surrounding Intel processors fell on hard times.
Right now Intel installs neither modems nor network adapter cards on its motherboards. But someday it could. Hey, why not put the adapter card right on the microprocessor? Says G. Dan Hutcheson, president of VLSI Research, "There is no reason why Intel couldn't integrate the networking functions into its processor. Then they could make a fast server with routing functions integrated into the server, instead of having a separate router box."
Intel says that wouldn't be practical-just now. But Moore's Law predicts that processors, which will soon have 7.5 million transistors, will have a billion by 2011. Those poor transistors will run out of things to do.
Are the networking companies worried? Bay Networks does $2 billion a year in revenue from selling routers, switches and hubs (the equipment that lets you move data packets around), plus network management software. Its chief executive, David House, was until December the senior vice president of enterprise computing at Intel. He puts a brave face to Intel's move. "Intel might be able to gain ground in adapter cards. But as you move up in the corporate network [to equipment like switches and routers], Intel doesn't even start to offer the equipment."
Sniffs 3Com's vice president of marketing for desktop products, Rakefet Kasdin, "They [Intel] think they are a networking company up there in Oregon, but they aren't. They are a microprocessor company."
Intel may prove them wrong. On Jan. 13 Intel closed a deal to acquire Case Technologies for $72 million. The newly acquired company sells switches that can route network traffic at speeds from 10 to 100 megabits per second. Switches are a faster alternative to routers, and even Cisco has had to make a defensive move in this segment, paying $4 billion to acquire switchmaker StrataCom.
Intel is now developing so-called backbone switches to be used for fiber connections between floors of a building as opposed to the larger switches bought by telecom companies. Also, in February Intel paid $52 million for a 12.5% equity stake in Xircom, a company that manufactures 100-megabit adapters for notebook computers.
Could Intel one day compete with Cisco and Bay to develop switches for sale to phone companies? After all, one of the reasons people don't buy faster machines for their homes is that they are usually limited to phone lines that connect them to the Internet at a creaky 28,800 bits, a thirtieth of a megabit, per second. Local phone companies are now, with some reluctance, deploying Integrated Services Digital Network connections at 128,000 bits per second. In the offing is the Asymmetric Digital Subscriber Line, at 1.5 to 10 megabits. Bandwidth is coming, but maybe not fast enough to suit Intel's taste.
Intel is already experimenting with satellites to connect home PC users to data networks. Since 1993 it has also been in the odd business of helping people get ISDN connections from their local phone companies-this as a way to foster demand for Intel videoconferencing technology. (And why is Intel interested in videoconferencing? It keeps processors busy.)
"I just wish there was the magic of 100 megabits per second available for the home user," says Grove. "I wish the telecommunications carriers approached this [problem] in the same fashion we do with our adapters, to move ISDN or ADSL into the mainstream so that it is available at the same price as plain old telephone service. That's what I have been dreaming about for five years."
Is Andy Grove just being wistful, or is he plotting something? Watch out, Ma Bell. ÿ
>>>>
ÿ |
