SAN MATEO, Calif. This is the image: Intel Corp. hasn't made heads or tails of the communications industry. As the company bought up startups sometimes paying two or three times the already inflated prices of 2000, some analysts say it was simply collecting one of everything, with no central strategy for interlocking the pieces. That Intel is a parent on Christmas Eve discovering the true meaning of "some assembly required."
It's Sean Maloney's job to shatter those perceptions. As vice president and general manager of Intel's communications group, he bears most of the responsibility to make this the year, at last, that Intel becomes a serious player in communications.
In fact, Maloney says there really is an underlying strategy to Intel's comms work, and it's one that exploits the Santa Clara, Calif. chip giant's greatest strength: manufacturing.
"The name of the game in the next three to five years is high-end capacity and what you do with it," Maloney said. "We're the guy with the factories. We believe that capacity is a big deal in the next few years. And integration of design and manufacturing is a big deal."
A unifying mission would be good, because its missteps have made Intel look more like a weekend warrior than a serious communications player. Critics say some acquisitions, such as Level One and Basis Communications, were all but flops in terms of delivering products. A Gigabit Ethernet physical-layer device (PHY) didn't pan out, forcing Intel to borrow one from a competitor. The Blackfin DSP developed with Analog Devices Inc. hasn't set the wireless world aflame. Only the IXP1200 network processor has been a standout and Intel lucked into that one, having nearly thrown it away after the Digital Semiconductor acquisition.
In a January report, analyst Joe Osha of Merrill Lynch & Co. Inc. (New York) said that with the exception of the IXP1200, Intel's communications efforts have "largely failed" and that "Intel has not done a good job of diversifying away from its core microprocessor business."
At the same time, no one wants to bet against Intel. The consensus is that the company that once morphed from DRAM player into microprocessor powerhouse has the money and therefore the time to change faces again. "To date, they haven't proved to be a tier-one player in the LAN or enterprise, but I wouldn't be too dismissive of Intel," said Alex Gauna, an analyst with UBS Warburg LLC (San Francisco).
In addition to muscle and might, Intel's got motivation. To remain a superpower, the company needs a cushion against the falling margins of PCs and microprocessors, and chief executive Craig Barrett has flagged communications as the escape hatch of choice.
Looking for leverage
With manufacturing, Intel might have found its "in." The comms industry largely lacks the kind of manufacturing prowess and product planning that have become key talents at Intel, Maloney said. Done right, Intel's comms strategy could introduce new levels of automation and cost reductions. The company's acquisition of LightLogic Inc., a maker of transceiver modules, was a key step in this direction. Even bigger was Intel's March 5 announcement to set up a fab for passive optical components a wholly new business for Intel but one that could use some manufacturing know-how.
And Intel "can out-manufacture anybody on the planet," said Will Strauss, president of consulting firm Forward Concepts. "Some year this decade, probably around 2007, the industry's going to ship 1 billion handsets. There's only a few companies that can supply one-third of that."
Moreover, Intel has proven it can develop products and bring them to market. "The execution record on the comms suppliers' side is abysmal," Maloney said. By contrast, "We're hitting all of our schedules" in desktop and server Gigabit Ethernet devices and net processors. "We haven't slipped a week in 18 months."
Fits and starts
Intel has nonetheless been beaten to market, as in the case of Atheros Communications Inc.'s 802.11a chip, which Intel is using in lieu of its own. It's also been outdone, as when its Gigabit Ethernet PHY proved inadequate, prodding Intel to use a PHY from Marvell Semiconductor Inc. (Intel is "very happy" with the Marvell relationship, Maloney said, but he refused to comment on whether Intel still hopes to produce its own Gigabit Ethernet PHY.)
Those might seem like small setbacks, but they get magnified in the communications arena, foreign turf for Intel and a world where it can't call the shots. "They're used to being the de facto standard in everything, but standards have to be in place first in communications, and that takes away their natural advantage," one analyst said.
Intel has been good in areas where it controls the infrastructure, but "when it's an open-market jump ball, they do less well," said Gauna of UBS Warburg. For example, he noted that Intel's chips for 10-Gbit/second Ethernet have found one major buyer Intel's own network interface card while the rest of the market has embraced Broadcom Corp.
It's not just individual products that have fizzled. Acquisitions have stalled as well, or failed to produce the results Intel wanted. DSP Communications, for example, was supposed to be Intel's ticket into the CDMA market. But those plans faded when Qualcomm Inc. stole DSP Communications' lone customer Kyocera Corp. and held its lead in CDMA. Separately, the Basis acquisition yielded some digital subscriber line chips that have already been discontinued, analysts said.
Maloney was quick to stress that Intel has picked up core talent along the way. "We have lost almost nobody in the R&D space from the acquisitions," he said, noting that Intel's R&D pool has grown to include RF experts in Israel and Southern California, and former space engineers in Russia. Expertise inside the company has broadened to include hard-core comms disciplines such as SS7 software stacks and voice-over-packet protocol capabilities, he said.
Intel is done with its acquisitions, at least for now, Maloney said. More important, he said that Intel has finished melding these startups into the company, a burdensome task that meant integrating disparate project-management methods and metrics into a single process. Maloney contends that project management is one of the key facets missing from most comms component suppliers.
"Anyone in the communications market has to accept that they're going to have a lot of projects," he said. "To have a high batting average on execution, you have to have best-known methods or project measurement and project management."
With the acquisitions complete, the manufacturing muscle comes next. By bringing in the heavy machinery, Intel believes it can help the comms industry bring its costs down and productivity up. For example, Maloney said LightLogic already has begun shipping an OC-192 and 10-Gbit/s Ethernet module that's one-fifth the price of competing products, thanks to LightLogic's patented method for computer-assisted fiber alignment.
Still, optical in general has a way to go. "The cost of other companies' optical modules has to come down by another order-and-a-half of magnitude," Maloney said.
Intel has optical-module customers going to production with their next-generation products, and LightLogic manufacturing is ramping up heavily as a result, Maloney said. LightLogic's founder, Jean-Marc Verdiell, is now director of Intel's optical efforts and is "in a driver's seat position" to make Intel a leader in innovative optoelectronics manufacturing, Maloney said.
Intel might be able to make a big contribution with its optical-components fab in south San Jose, Calif. The fab has been operating for roughly one year but its existence wasn't made public until earlier this month. There, Intel is using equipment and techniques from its other fabs, applying large-scale, automated manufacturing that is largely absent from the optical-components industry to make Bragg grating filters and planar waveguides.
"The supply chain we use is really the same supply chain that we use for our semiconductor tools," said Rama Shukla, the general manager in charge of Intel's optical-components business. "The chemistry and recipes are different, but other than that, we are talking about an operation which really runs like a semiconductor fab."
Mass production isn't in the cards just yet, however, as Intel is starting its optical-components division by making customized parts, a model similar to LSI Logic Corp.'s ASIC business.
Anthony Cataldo contributed to this story.