IBM and Intel Step Into the Ring
By Alec Appelbaum
CHIP TRIP
FROM NEW YORK factories to California campuses, people are talking. And emailing and surfing and e-buying. Phone companies crave all that traffic, and the folks who make telecom equipment want to deliver it quickly and dependably. So telecom-equipment companies are looking for fast chips that communicate like dervishes and don't do much else. Which is why two veterans of the general-purpose microprocessor, IBM (IBM) in New York and Intel (INTC) in California, have announced new chip families that put them in direct competition in the networking world.
The back-to-back announcements -- Intel's on Sept. 1 and IBM's the next day -- had market-watchers settling back for a heavyweight bout. IBM had just announced a partnership with networking giant Cisco Systems (CSCO), and if it also had a communications chip to sell through its huge service division, then the strength of its rep with networking companies should portend a cleaning of Intel's clock?right? Actually, the networking world changes as fast, and twists and turns as crazily, as information coursing over the Internet. Intel was up slightly at midday, while IBM sank. Clearly, if this is a battle, these announcements are just the start.
Before we submit the companies to a weigh-in, it's worth understanding what a networking chip does. Unlike the chips in your computer, which run everything from writing to graphics, networking chips are more specialized and rigged for speed. This means everything to the Internet gear known as routers, which send information from a network to a computer or phone, says Aberdeen Group research analyst Andy Cray. Networking chips have fewer instructions than general-purpose chips, and so they use less power and dish out more speed.
But unlike the standard chips currently used for special functions such as PalmPilots, IBM's and Intel's babies are programmable. After they're shipped, customers can make changes to them in a matter of days to reflect advances in computer language or software. Cray says this will radically improve life for startups, which now have little recourse when technological advances render their products obsolete in the months between business plan and testing. "It will be possible to implement new networking protocols on the fly," he says.
OK, so networking chips are a good idea, and both IBM and Intel have been working on them for upward of two years. Does the fact that both of them are stepping out now mean that one will have to prevail at the other's expense? Not necessarily. According to spokesmen and analysts, the two are following somewhat different strategies. Intel is aiming to establish itself as a fertile programming source beyond the PC, while IBM is trying to play off its fame in research and services.
By combining its new programmable chips with its service reputation, IBM gives its networking offerings a new and potentially happier home. When IBM tried to sell telecom gear directly to phone companies on its own, says Giga Information Group analyst Stan Schatt, the Big Blue marvel "didn't really have a channel to reach large enterprises." By treating telecom-equipment giants such as Nortel Networks (NT) and Alcatel (ALA) as its customers, IBM gets to sell more gear -- and provide more of the services that produce roughly a third of its revenue -- with bigger margins than it can get through equipment sales alone. "This is an expansion move and an acceleration move for customers who we help in time-to-market," says Steve Longoria, a director in IBM's wired-communications group.
Meanwhile, Intel has been warming up for life outside the PC box ever since it announced plans to buy networking-chip specialist Level One in March. Level One produced Intel's new chip instructions, called IX architecture. According to spokesman Tom Beermann, Intel's mousetrap has 13 "support chips" that provide "blueprints" for software and hardware developers to run with in any direction. While IBM's Longoria says Big Blue wants to become "the technology-solutions provider," a font of consulting as well as of silicon, Intel appears to be hunting for a software partner that can dominate networking the way Microsoft (MSFT) dominates PCs. It's trying to make it "simple and easy" for programmers to design applications on its networking chips, says Beermann, and to sell lots of chips as some of those designs take off.
This whole market is very young and complicated. IBM announced three customers -- Nortel, Alcatel and Newbridge Networks (NN) -- while Intel disclosed a fistful of "supporters," including Newbridge and Cisco Systems. Huh? Didn't Cisco just commit to $2 billion worth of IBM goods? Yes. With the market and technology so fluid, these equipment makers can't tie themselves to any one supplier.
Ultimately, IBM's and Intel's success in networking chips will depend on how easily new services can run on systems that incorporate them. IBM would seem to have an advantage here, because of its proven expertise in the switching software that makes such systems work smoothly. But Bob Merritt of Semico Research says there's little to prevent Intel from going out and buying comparable expertise. Indeed, if Intel really wants to shine in consumer appliances like cable modems, Schatt suggests it might do well to go buy 3Com (COMS).
So how do we watch this fight? Patiently. "Nobody knows how to measure these processors very well, because in many cases they're intended to address different parts of the network," says Merritt, who was briefed on IBM's product. And this isn't just a two-player race. Startups like Agere, Pivotal Communications and C-Port (which is working on standards with IBM) could get folded into bigger companies; chipmakers Texas Instruments (TXN) and MMC Networks (MMCN) can go after the shapeless mass of futuristic applications with their own silicon. Craig Johnson, whose Pita Group tracks the networking world, advises investors to look for "architecture wins" -- broad contracts with equipment makers that entail a whole series of core and support products -- as a means of keeping score. All the same, there doesn't have to be a loser in a market this hungry. "Every equipment provider has got to have a fast chip," says Aberdeen's Cray.
What is clear is that you, by reading this site and letting your kid bash your neighbor's kid in some Sega gamescape, are making telecom-equipment makers ravenous for faster, smarter and even-faster-than-they-realized-they-wanted chips. When equipment companies start signing deals in the next six to eight months, both these chipmakers had better be ready.
IBM and Intel Step Into the Ring
By Alec Appelbaum
CHIP TRIP
FROM NEW YORK factories to California campuses, people are talking. And emailing and surfing and e-buying. Phone companies crave all that traffic, and the folks who make telecom equipment want to deliver it quickly and dependably. So telecom-equipment companies are looking for fast chips that communicate like dervishes and don't do much else. Which is why two veterans of the general-purpose microprocessor, IBM (IBM) in New York and Intel (INTC) in California, have announced new chip families that put them in direct competition in the networking world.
The back-to-back announcements -- Intel's on Sept. 1 and IBM's the next day -- had market-watchers settling back for a heavyweight bout. IBM had just announced a partnership with networking giant Cisco Systems (CSCO), and if it also had a communications chip to sell through its huge service division, then the strength of its rep with networking companies should portend a cleaning of Intel's clock?right? Actually, the networking world changes as fast, and twists and turns as crazily, as information coursing over the Internet. Intel was up slightly at midday, while IBM sank. Clearly, if this is a battle, these announcements are just the start.
Before we submit the companies to a weigh-in, it's worth understanding what a networking chip does. Unlike the chips in your computer, which run everything from writing to graphics, networking chips are more specialized and rigged for speed. This means everything to the Internet gear known as routers, which send information from a network to a computer or phone, says Aberdeen Group research analyst Andy Cray. Networking chips have fewer instructions than general-purpose chips, and so they use less power and dish out more speed.
But unlike the standard chips currently used for special functions such as PalmPilots, IBM's and Intel's babies are programmable. After they're shipped, customers can make changes to them in a matter of days to reflect advances in computer language or software. Cray says this will radically improve life for startups, which now have little recourse when technological advances render their products obsolete in the months between business plan and testing. "It will be possible to implement new networking protocols on the fly," he says.
OK, so networking chips are a good idea, and both IBM and Intel have been working on them for upward of two years. Does the fact that both of them are stepping out now mean that one will have to prevail at the other's expense? Not necessarily. According to spokesmen and analysts, the two are following somewhat different strategies. Intel is aiming to establish itself as a fertile programming source beyond the PC, while IBM is trying to play off its fame in research and services.
By combining its new programmable chips with its service reputation, IBM gives its networking offerings a new and potentially happier home. When IBM tried to sell telecom gear directly to phone companies on its own, says Giga Information Group analyst Stan Schatt, the Big Blue marvel "didn't really have a channel to reach large enterprises." By treating telecom-equipment giants such as Nortel Networks (NT) and Alcatel (ALA) as its customers, IBM gets to sell more gear -- and provide more of the services that produce roughly a third of its revenue -- with bigger margins than it can get through equipment sales alone. "This is an expansion move and an acceleration move for customers who we help in time-to-market," says Steve Longoria, a director in IBM's wired-communications group.
Meanwhile, Intel has been warming up for life outside the PC box ever since it announced plans to buy networking-chip specialist Level One in March. Level One produced Intel's new chip instructions, called IX architecture. According to spokesman Tom Beermann, Intel's mousetrap has 13 "support chips" that provide "blueprints" for software and hardware developers to run with in any direction. While IBM's Longoria says Big Blue wants to become "the technology-solutions provider," a font of consulting as well as of silicon, Intel appears to be hunting for a software partner that can dominate networking the way Microsoft (MSFT) dominates PCs. It's trying to make it "simple and easy" for programmers to design applications on its networking chips, says Beermann, and to sell lots of chips as some of those designs take off.
This whole market is very young and complicated. IBM announced three customers -- Nortel, Alcatel and Newbridge Networks (NN) -- while Intel disclosed a fistful of "supporters," including Newbridge and Cisco Systems. Huh? Didn't Cisco just commit to $2 billion worth of IBM goods? Yes. With the market and technology so fluid, these equipment makers can't tie themselves to any one supplier.
Ultimately, IBM's and Intel's success in networking chips will depend on how easily new services can run on systems that incorporate them. IBM would seem to have an advantage here, because of its proven expertise in the switching software that makes such systems work smoothly. But Bob Merritt of Semico Research says there's little to prevent Intel from going out and buying comparable expertise. Indeed, if Intel really wants to shine in consumer appliances like cable modems, Schatt suggests it might do well to go buy 3Com (COMS).
So how do we watch this fight? Patiently. "Nobody knows how to measure these processors very well, because in many cases they're intended to address different parts of the network," says Merritt, who was briefed on IBM's product. And this isn't just a two-player race. Startups like Agere, Pivotal Communications and C-Port (which is working on standards with IBM) could get folded into bigger companies; chipmakers Texas Instruments (TXN) and MMC Networks (MMCN) can go after the shapeless mass of futuristic applications with their own silicon. Craig Johnson, whose Pita Group tracks the networking world, advises investors to look for "architecture wins" -- broad contracts with equipment makers that entail a whole series of core and support products -- as a means of keeping score. All the same, there doesn't have to be a loser in a market this hungry. "Every equipment provider has got to have a fast chip," says Aberdeen's Cray.
What is clear is that you, by reading this site and letting your kid bash your neighbor's kid in some Sega gamescape, are making telecom-equipment makers ravenous for faster, smarter and even-faster-than-they-realized-they-wanted chips. When equipment companies start signing deals in the next six to eight months, both these chipmakers had better be ready.
v |
