Process Boy, Article..Intel-on-a-chip-PC MPU giant weaves plan to dominate embedded market..
June 9, 1999
ELECTRONIC BUYERS NEWS : Silicon Valley- If Intel Corp. had invested in gingham and calico instead, this could have been the world's biggest quilt.
The company's effort to stitch together a patchwork of opportunities outside the PC sector has triggered equity stakes in technologies as seemingly mismatched as Internet cable services and Linux software.
In the short history of its equity program, the microprocessor giant has set an instant halo over the heads of scores of promising start-ups. But by seeding everything from broadband backbones to online video content, Intel's strategy assumes a formless, almost random appearance. And then, from this furious hodgepodge, patterns emerge.
Equipped with its x86, StrongARM, and legacy architectures, Intel is evolving beyond its traditional sphere of influence in the PC market by steering its microprocessor developments toward application-specific embedded designs for servers, smart phones, and other network appliances.
And to complement its hardware efforts, the company is sealing an apparently endless series of acquisitions, investments, and joint-development deals in an effort to piece together the sundry elements needed to take on embedded applications and sustain its own meteoric growth.
"Intel is tacitly, if not explicitly, focusing on the embedded customer, " said analyst Drew Peck of SG Cowen Securities Corp., Boston. " Putting all their eggs in one basket in the mainstream microprocessor market is fraught with risk. Intel's in the middle of a massive transition, conveying the impression that everything's normal when it's anything but."
While PC sales remain robust and will serve as the foundation of Intel's business for the foreseeable future, the value of the semiconductor content in the average system is falling faster each year. The $280 an OEM spent on a 233-MHz Pentium in 1997 paid for a 350-MHz chip last year, with enough change left over for two or three other key PC components, according to International Data Corp., Framingham, Mass.
And while IDC reported that PC-microprocessor revenue has more than doubled since 1995, to $18.1 billion, the popularity of sub-$1,000 PCs and a host of small, nettling competitors are causing Intel to concede margins in the low-cost segment.
Advanced Micro Devices, Integrated Device Technology, National Semiconductor, and Rise Technology took a $2 billion slice out of the x86 processor pie in 1998, according to In-Stat Group, Scottsdale, Ariz. At the same time, Apple Computer's iMac system has emerged as an unexpected threat to Intel's Windows-based ideal.
No shrinking violet, Intel's move into the embedded space is viewed not so much as a reaction to competition in the PC arena as a recognition that it needs to diversify.
"Most people think of PCs when they think of Intel, but in the past few years we've expanded into servers on the high end and embedded in the low end, " said a spokesman for the Santa Clara, Calif., company. "We might be interested in seeing someone's tool set run best on StrongARM, for example, so we'll say, 'Let's work together, let's be partners, and oh, by the way, we'll invest in you, too.'"
And invest it has.
By its own estimate, Intel is one of the largest corporate venture capitalists in the technology industry, with a portfolio that includes more than 250 companies and is valued at roughly $3 billion. Last year, the company made about 130 investments worth $830 million, including a $500 million stake in Boise, Idaho-based DRAM maker Micron Technology Inc.
A quick look at Intel's investments so far this year reveals specific interests in the embedded market. Among those investments are minority stakes in Sente Inc., an Acton, Mass., company that develops software tools for power-consumption analysis in ASICs; Abstract Inc., a Fremont, Calif., outfit specializing in system-on-a-chip design-verification tools; and San Jose-based LogicVision Inc., which makes embedded test circuitry.
Intel's investment pattern is fueled by a recognition that a fundamental difference exists between the homogeneous makeup of a PC platform and the application-specific nature of the embedded space.
To date, the company's embedded plans have followed two paths: high-end devices based on the Celeron processor, and volume-driven consumer and communications applications based on the StrongARM. Both routes are expected to take Intel in the direction of system-on-a-chip development.
"Networking and consumer markets are very lucrative in volume and revenue, if not in margins, and both are intensely pointing at system-on-a-chip," said In-Stat analyst Max Baron. "In order to play there, Intel will have to support that technology, either as an ASSP supplier or by empowering ASIC companies [with its processor IP]."
How is Intel faring with its SOC plans?
So far, it's hard to tell. Though Intel's equity investments have been broad, they haven't been especially deep-typically less than $10 million each. In fact, the company has yet to lay out a clear SOC policy, from either an investment or an architectural standpoint, according to analysts.
But sources speculate Intel is quietly amassing an IP arsenal and the tools to build powerful Internet devices around its processor cores.
The company's pending buyout of Level One Communications Inc., last week's $780 million bid for computer-telephony integration (CTI) specialist Dialogic Corp., and an earlier agreement with Analog Devices Inc. to jointly develop a next-generation DSP all point to Intel's intention to make a name for itself as a broad-line communications-chip supplier.
While little is known about Intel's DSP development efforts, the company is clearly pushing its StrongARM processor into a number of embedded sectors. In April, Intel said it would field a software-programmable network processor for enterprise- and service-provider-class communications equipment-a chip since identified as using the SA-1200 StrongARM architecture.
And although Intel's interest in Dialogic has more to do with revving up system-level building blocks for CTI applications than with pushing its own silicon, Dialogic might replace its Intel i960 and 80C188 embedded chips with StrongARM processors.
"Actually, we fell in love with the StrongARM even before the Intel acquisition," said John Landau, vice president of marketing for Parsippany, N.J.-based Dialogic. "It's a neat chip."
But gaining support in the embedded space won't be easy. Despite its domination of the PC arena, Intel has little influence over companies such as ARM, IBM, MIPS Technologies, and Motorola, which have an installed base of software-development tools-a fact that could help explain Intel's apparent willingness to seek partners in this segment.
Intel's heavy reliance on proprietary software and its relatively conservative approach to integrating functions into its processor cores may also help illuminate its embedded strategy.
When developing application-specific designs using the StrongARM, for example, Intel does not modify the processor, but designs and modifies a companion chip. Only if enough applications share the same peripheral functions will the company consent to integrate them into the StrongARM.
"Our strategy is to do high-performance, low-power processors," said Mark Casey, director of marketing for Intel's StrongARM Division in Chandler, Ariz. "The second part is to develop highly integrated solutions for target market segments."
Whether this two-tiered approach gives Intel the flexibility it needs to keep pace with rapidly evolving embedded applications is a point of debate.
In the fledgling network-processor market, OEMs are already moving away from general-purpose RISC chips such as the PowerPC, MIPS, and StrongARM to more specialized programmable devices that handle a complex array of LAN/WAN protocols, said Lawrence Ebringer, director of marketing for Sunnyvale, Calif.-based MMC Networks Inc., a pioneer in the area. "I don't know if you can say that Intel is validating our business model, but we're keeping a close eye on them," he said.
And while most embedded-RISC licensees incorporate hardware building blocks into their designs, Intel relies on the native processing strengths of the StrongARM, which it bolsters with software. In fact, about half of the 1999 budget for Intel's Embedded Microcomputing Division is allocated to improving the company's software tools, such as VTune, which helps OEMs optimize Windows NT software code on Intel chips.
Particularly in the set-top box market, where the MIPS core is an entrenched architecture, Intel's use of a central processor is a departure from the dedicated hardware blocks that populate the majority of devices.
Intel claims to offer customers both "hard" and "soft" technology, but it insists that OEMs are looking for standard, off-the-shelf processors that can be differentiated through software.
"Nortel, for example, is now outsourcing all of their computing platforms, " said Joe Jensen, director of marketing for Intel's Chandler-based Embedded Microcomputing Division, which sells embedded x86 processors. " They've come to realize that the value-add is in software."
However, some observers see this as a PC-centric view, something that Intel has traditionally encouraged but may have to change as it moves into new markets.
"A PC by its very nature is a general-purpose machine," said Tom R. Halfhill, an analyst with MicroDesign Resources in Sunnyvale, Calif. "The embedded market demands more customizable parts, and there's a strong trend towards customization and configurability. You can't take off-the-shelf parts and build a specialized device like you'd take general-purpose parts and build a PC."
Faced with increasing chip complexity and shrinking product life cycles, Intel has yet to follow other IC makers by turning to independent technology developers and their promise of design reuse and shorter time-to-market. The notable exception is Rambus Inc., which is providing a high-speed memory interface with Intel's Camino chipset for next-generation PCs.
Whether Intel will continue to acquire the technology it lacks, or adopt the IP licensing model typified by the ASIC industry, is unclear. Certainly, the company is getting into position with strategic investments in EDA tool makers such as Magma Design Automation.
But if it really wants to harness the capabilities these IP companies bring to bear, Intel must do more than invest in the industry-it must break with tradition and open its proprietary methodology to third-party design tools, according to Rob Chaplinsky, a partner with Mohr, Davidow Ventures in Menlo Park, Calif.
"Building a system-on-a-chip is more than just buying IP," Chaplinsky said. "It's buying IP you can integrate into your flow."
By Crista Souza, Mark Hachman, and Mark LaPedus
|
