Gary and Thread - Smartmoney Article: Does Intel Want To Be The Next Cisco?
Article makes some poignant observations about what Intel might have to contend with in the networking and communications space. That said, Intel has made great strides in this sector since entering it with vigor in the LAN products sector. Article presents how some of the recent acquisitions and business initiatives (Server Farms) may tie together.
One analyst comments: "My gut feeling is Intel will have a real challenge here because it's going to be too small [a market] to capture their interest, and too fragmented a skill set for them to bring together the necessary expertise in time," Moosa says. "And I don't think culturally they have the focus in communications to deliver product after product after product. You've got to live and breathe this stuff."
From what I've heard, I would not be certain that my brethren over in the Networking and Communications Group do not already "live, eat, and breath" this stuff. My contacts seem to indicate that Intel is of a mind set to do whatever it takes to get a viable foothold as a fierce competitor in the Network and Communications Products supplier business.
Article also points out that this market is still pretty small at this point compared to what us Intel types are used to. However, it is predicted this market will grow at a rather large rate over the next few years. I suppose the markets for 4004's, 8008's, and 8086/88's were considered small at one point. Of course, I was in grade school then :-).
PB
smartmoney.com
======================================================================
Does Intel Want to Be Cisco?
By Tiernan Ray
THEY WERE waiting on unbearable lines for taxicabs in Las Vegas this week, and eating even more unbearable chow at the food court at the Bellagio and elsewhere, as networking enthusiasts got down to business at NetWorld+Interop, a semiannual trade show where those who follow Cisco Systems (CSCO), Lucent Technologies (LU), Nortel Networks (NT) and other makers of communications gear get to kick the tires on the new routers, switches and digital telephone systems. Quietly, Intel (INTC) was holding small private meetings to outline a product strategy that may shake up the networking world in the next couple of years, and which, without a doubt, represents the company's most dramatic departure ever from its highly successful franchise in computer microprocessors.
On March 4, Intel announced it would acquire Level One Communications (LEVL), a maker of semiconductors that power boxes that move packets of data around local-area networks (LANs) and across the Internet. The deal was approved on April 27. Level One, based in Sacramento, Calif., has soared 210% in the past two years as investors salivated over the chips it makes for high-speed T1 lines that connect companies to the Internet. Intel has an existing business in networking, manufacturing hubs and switches for office networks, as well as selling chips on the open market to various equipment makers, a business that puts the company in competition with Cisco and that has spawned persistent rumors of an acquisition of 3Com (COMS).
Given that, Intel might be expected to use the Level One purchase to produce more sophisticated chips for fast Ethernet and gigabit Ethernet, technologies that are popular in corporate America and that could sell in the kinds of high volumes that appeal to Intel, which last year sold 92 million microprocessors and 81 million chipsets (the circuits that tie together microprocessors and other chips), according to Mercury Research, a technical and market research firm based in Scottsdale, Ariz.
Intel's Network Market? Manufacturer Revenues ($Millions) 1998 1999 2000 2001 2002
Layer 3 Ethernet Switches * $637.40 $1679.50 $2725.50 $3393.70 $3918.00
High-end Routers 1501.80 1694.40 1925.00 2117.50 2286.90
Access Concentrators ** 1954.00 2750.50 3540.60 4393.30 5156.30
Total 4093.20 6124.40 8191.10 9904.50 11361.20
Growth Rate 20% 50% 34% 21% 15%
* Includes both Modular and Fixed Configurations
** Includes modem, cable modem and DSL concentrators/head-end gear
Source: Dell'Oro Group
Instead, however, Intel is going in the opposite direction, moving away from the relatively simple chips used for Ethernet LANs and offering what amounts to a microprocessor for wide area networks, or what's commonly known as the WAN. The WAN is where different companies' networks connect with one another to exchange traffic, and there is a level of software complexity in the exchange of traffic at these public hitching posts that seems tailor-made for a fairly complex processor like the kind Intel is great at building. In addition to just moving data, the chips might encrypt that data for privacy as it is sent into the world, establish firewalls to protect the interior of the network from sabotage and exchange information about traffic all across the public network.
It may also be one of the first places where vendors try out a much-touted technology called voice-over-IP, those Internet phone calls you've heard so much about. As such, these boxes increasingly must ensure so-called quality of service, making data seem as reliable as the phone network.
To do all that, Intel is in effect hoping to build a microcomputer for networks, a big, complex, general-purpose chip that can be slipped inside any number of networking boxes built by equipment makers. Such a chip would replace complex chips used today in routers, chips dubbed application specific integrated circuits (ASIC), which are built not for general computing tasks but rather for a specific set of functions a router needs to perform. Product details on the network processor are scant, so far; a Wall Street analyst traipsing around the show on Tuesday told me he's heard that Intel's network processor will consist of four StrongARM processors, the chips Intel acquired from Digital Equipment Corp. a year and a half ago, fabricated as a single chip.
Tom Halfhill, an embedded chip analyst with MicroDesign Resources in Sunnyvale, Calif., says the StrongARM may prove an asset to Intel precisely because it can be manufactured in a variety of flavors to suit the various niches in the WAN market. Halfhill says Intel has also talked about adding instructions to the StrongARM that would improve its ability to process heavy amounts of data, as would be needed in a router or switch.
The idea is not novel: Startups such as Softcom Microsystems and C-Port are already hard at work on this. At least one venture capitalist I know thinks programmable chips are the future. "Programmability is coming to networking; it's only a matter of time."
But something is amiss here. For one thing, the market Intel is initially targeting is small and fragmented. Broadcom (BRCM), perhaps the best example of a network chipmaker, made just $200 million in revenue last year -- peanuts to Intel. What's more, as the handy chart provided by my friends at the Dell'Oro Group shows, last year the three most likely product categories into which Intel could sell, boxes designed for the WAN, totaled $4 billion. That's expected to grow to $11.3 billion in 2002, but the market is divided into literally dozens of different kinds of networking boxes of various flavors that perform a wide variety of tasks. Each of these represents a small niche. Unlike the personal computer, different chips are needed for each kind of box. Intel says a general-purpose microprocessor could replace much of that chip menagerie.
Of course, the market will grow. But the bigger problem is that Intel lacks a software partner. The dream team, Intel and Microsoft (MSFT), created a standard combination of software and hardware that fueled but also guided the computer business for years. In the networking world, Cisco and Lucent design their own software and their own ASICs, which are then built to specification for them by a variety of chipmakers, some of whom owe their entire business to Cisco and a few other networkers. The close relationships Cisco and Lucent enjoy with these chip fabricators, and with companies such as Broadcom that provide them with other, less-exotic chips, are very tight collaborations that Intel will be hard-pressed to equal or replicate.
Intel says that using a reprogrammable processor, box makers will be able to add their own software, their own intellectual property, in effect, and compete with Lucent and Cisco as well as relieving Intel of some of the software burden. Nortel is considered a prime candidate for such an arrangement
But another way for Intel to fudge its lack of expertise at the software and systems level is for the company to force the industry to remake the router or switch in its own image, much the same way it has gradually shaped the design of the PC to suit the Pentium chip. Elias Moosa, who follows PMC Sierra (PMC) and others for Thomas Weizel Partners in San Francisco, thinks Intel is considering a change in the overall design of routers. Today's simple switch architecture would be replaced by Intel's network processor driving a mesh, or matrix, of smaller switching chips from vendors such as Abrizio of Mountain View, Calif., and Sweden's SwitchCore. Kind of like putting a miniature Internet inside the box.
Nonetheless, Moosa is skeptical about Intel's ability to reduce the complexity of so many product lines.
"My gut feeling is Intel will have a real challenge here because it's going to be too small [a market] to capture their interest, and too fragmented a skill set for them to bring together the necessary expertise in time," Moosa says. "And I don't think culturally they have the focus in communications to deliver product after product after product. You've got to live and breathe this stuff." Moosa thinks Intel might have to spend several billion more to acquire a networking leader such as Vitesse Semiconductor (VTSS) or PMC.
Another trick is for Intel to find a completely new software partner, or a set of partners. The company recently announced a strategy of setting up so-called cyberbricks, basically server farms run by Internet service providers. It's a bizarre strategy considering how far afield it is from Intel's core business of making chips. One observer tells me, though, that the cyberbricks may be Intel's way of getting ISPs signed up as software partners.
David Smith, a consultant with Technology Futures in Austin, Texas, says that because of the increasing complexity of the Internet, ISPs may take advantage of the reprogrammability of Intel's chips to configure routers and switches with their own instructions, rather than those of Cisco, in effect diminishing the importance of the box maker and turning parts of the Internet into a private lane. "ISPs are all looking for a way to differentiate their networks; one way to do that is to implement some commands [in a reprogrammable processor] that say, 'Treat this packet one way if it is from my network, and another way if it is from someone else's network,'" says Smith.
At any rate, Intel will have to contend with the increasing sophistication of the merchant vendors such as Broadcom. They are coming at this problem from the opposite direction of Intel, pursuing the "dumber gets smarter" theory of networking chips. Broadcom is basically working its way up from so-called physical layer chips, which just move bits, to switching chips that require more complex circuits, using the combined knowledge of both as a series of stepping stones to protect its market. Broadcom has reportedly won the rights to design the brains of an upcoming product from Nortel's Bay Networks unit, the BayStack 450, by combining the physical transceiver chip with switching logic on a single chip.
Joe Osha, who tracks both Intel and Broadcom for Merrill Lynch, tells me that he thinks this combination could prove to Broadcom's advantage. "Level One wants to do that kind of integration as well, but it's going to be harder for them; Broadcom is better positioned because they've focused on digital signal processing for their transceivers from day one."
Obviously, Intel has some catching up to do in a segment of the chip world the company has skirted for the most part; that may be the chief reason Intel partnered recently with Analog Devices (ADI) to develop the kind of expertise in digital signal processors that Broadcom possesses. One thing's for sure: Without having the networking business handed to it on a silver platter, as IBM did with the PC microprocessor business in the early 80s, it's going to be a tough slog for Intel in this new territory.
|
