SAN MATEO, Calif. Intel Corp. has spent two years and several billion dollars buying assets and sculpting a strategy to tackle the telecommunications market. And the fight is just beginning.
The microprocessor giant is quickly moving from the access region of the network to the core, where the designs are trickier, the marketing ferocious but the margins greater. It's also where Broadcom Corp. is headed.
While there is plenty of other competition from communications-IC vendors in this space (Intel is, by one ranking, only No. 4 in the market and Broadcom isn't in the top 10), a clash of the titans may be brewing.
"We are trying to own all of the building blocks of the Internet economy," said Pat Gelsinger, vice president and chief technology officer for Santa Clara, Calif.-based Intel. "We absolutely see Broadcom as a strong competitor."
"We started out in the access segment, but now we are moving into the core infrastructure level," said Henry Nicholas, president and chief executive officer of Broadcom. Of the Irvine, Calif., company's recent spate of acquisitions, he said, "all . . . have been made with the objective of connecting every computer in every market over every medium."
Historically, the two chip giants hail from very different regions Intel in the computing segment and Broadcom in end-user broadband modems. But the two have transformed themselves through aggressive acquisition programs.
Much of this high-stakes M&A chess game to date has been played out in the access level of the network, where both the PC and broadband modems link to the Internet. But as Intel and Broadcom move beyond the access level into the transport segment, where margins and profits are better and the technical bar is set very high, it is unclear where the advantage lies.
Networking equipment is a new game, one in which vendors are not necessarily awed by the two giants, and homegrown ASIC solutions often win the day over chip makers' offerings. Moreover, the rapid expansion of the Internet has made it impossible for any single system vendor to supply the entire global network, a factor that allows multiple chip suppliers to compete for the same slots.
With so many slots in every box, and so many different systems needed to move data across the network, the vendor that can provide compatible chips for every slot has a significant advantage over a small startup with a single outstanding product.
"That's the motivation behind all these communications acquisitions," said Linley Gwennap, president of market research firm The Linley Group (Mountain View, Calif.). "Intel and Broadcom are both following the same strategy and are trying to build themselves into the dominant networking-silicon providers. There's no question that Intel and Broadcom are now direct competitors."
Intel's long march into Broadcom territory began two years ago with its $2.2 billion acquisition of Level One Communications (Sacramento, Calif.), whose product line then was Ethernet transceivers used in LAN hubs. A year later, Intel snapped up DSL component vendor Basis Communications Corp. (Fremont, Calif.) for $450 million and the cable modem unit of Stanford Telecommunications Inc. (Sunnyvale, Calif.) for an undisclosed price.
The moves did not go unnoticed at Broadcom, where chief executive Nicholas accuses Intel of copying his own acquisition strategy. Nicholas said Intel's $550 million cash buy of VxTel Inc. (Fremont) to get voice-over-packet technology was designed to counter Broadcom's acquisition of Silicon Spice Inc. (Mountain View). And Intel's January purchase of networking-card vendor Xircom Corp. was aimed at checkmating Broadcom's 1999 acquisition of AltoCom Inc. (Mountain View), he said.
To date, Intel has spent at least $7 billion on at least 16 communications-related acquisitions. Mark Christensen, vice president and general manager of Intel's network communications group, said that some of these deals have not been made public, so the total cost for Intel's shopping spree could well exceed $8 billion. Broadcom, for its part, has made 18 separate acquisitions in communications in the similar period and at a comparable cost.
Yet the two acquisition plans have some subtle but important differences, said Jeremy Bunting, senior analyst for communications chips at investment-banking firm Thomas Weisel Partners (San Francisco). The key is long-term vision. Bunting said that Broadcom's plan revolves around preparing for the wired home and the various products that OEMs will need to supply that market. In contrast, Intel is taking a more shotgun approach, Bunting said, one that does not suggest a clear vision of what the network will look like in five years.
"Broadcom's acquisition strategy says that in five years, every home is going to have some kind of box that serves as a broadband gateway, and it is going to develop or acquire every technology that will be required by that box," said Bunting.
He pointed to several deals that showcase this plan. In May 2000, Broadcom picked up Pivotal Technologies Corp. (Pasadena, Calif.), which produces chips that link TV sets, DVD players or monitors to the network, in a stock trade valued at about $242.5 million.
That came on the heels of the March 2000 acquisition of Stellar Semiconductor Inc. (San Jose), which develops low-cost 3-D graphics chips for set-top boxes or game consoles. The Stellar buy was another stock deal, worth about $140 million. Earlier, in April 1999, Broadcom bought Epigram Inc. (Sunnyvale), a key player in the development of home phone-line networking technology, for some $343 million in stock.
In Bunting's view, Intel's acquisition strategy is less cohesive, and lacks a sense of what the needs of OEM customers will be five years out. For instance, he said, Intel will offer the IXP1200 processor for routers, but it does not have a solid offering in the switch-fabric space, an arena Bunting termed just as critical.
Beyond connecting the home, Broadcom has moved further up the chain, by picking up some of the key pieces required to build the systems in the access and data aggregation levels. Arguably the most important of these was the November 2000 acquisition of SiByte Inc. (Santa Clara) for $2.06 billion in stock. SiByte develops network-processing devices, which are required to add intelligence into routers and switches.
These components will go head-to-head against Intel's IXP1200 network processor. (Ironically, Dan Dobberpuhl, who founded SiByte, was a key player on the original StrongARM processor team, the technology upon which IXP-1200 is based.)
As both companies advance toward the core, Intel appears to have an early lead. Just last month, the company announced eight new error-correction and physical-layer chips all aimed at the optical networking market. These devices support data rates as fast as 10 Gbits/second, and serve as the electronic link between the data aggregation level and pure photonic data transport. According to Jeremey Donovan, senior analyst for research firm Dataquest (San Jose), they are all the result of the March 2000 acquisition for $1.25 billion of the Danish chip company Giga A/S.
Early on, some observers questioned whether Intel could adapt to a more wide-open market than computing, one where open standards play a larger role. But internally at least, Intel seems to have gotten the message. And from its years in computing the company is adept at seeding and supporting the engineering community with everything from data sheets to developer kits and forums.
Intel could capitalize on architectural and market nuances as well. For instance, one key services aspect of networking is real-time virus detection, an activity that, in some subsystems, can eat up to 100 percent of the system's compute time.
"This requires very fast processing that we are very good at," said Ashok Kanagal, general manager in Intel's Integrated Access Operation (Chandler, Ariz.).
Broadcom's march into the core includes transceiver technology resulting from the August 2000 acquisition of Newport Communications Inc. (Irvine) for about $1.3 billion in stock, but those parts are not yet ready.
But Bunting of Thomas Weisel Partners said Intel's lead may soon evaporate. In the high-end networking segment, he expects to see Broadcom begin with the physical layer and gradually expand higher. When it does, he said, Broadcom will have an entire portfolio of products to sell. "The Giga products from Intel are good, but it's just a niche technology."
Despite its ambitious strategy, there is a chink in Broadcom's armor: the stock market. Chief executive Nicholas boasts that every acquisition has been a pure stock trade. That's fine if the market is strong, but now that the tech sector has slumped it will be hard for the company to continue on this path, said Tom Nolle, president of the telecom consulting firm CIMI Corp. (Voorhees, N.J.). Broadcom's share prices now hover in the mid-$30 range, after reaching nearly $300 last summer. "I'm sure that Broadcom will not be able to make any more acquisitions because its share price is too low," he said.
Another booby trap for Broadcom, Nolle said, is the fact that much of its business has been with startup ventures. "Broadcom's success has come from selling to customers that have no earnings, and are financed by venture capital," said Nolle. "Now that venture money has dried up, Broadcom's customer base will not be buying as much."
For Nolle, the key question is, "What is the device that will consume all these chips? Consumer broadband services are not yet mainstream, and it could be some time before they are."
In the meantime, Intel has its microprocessor revenues to help sustain it, as the company waits for its acquisitions to start producing revenue. "I think that Broadcom is the weak sister in this competition," Nolle said.