I posted this awhile ago, but maybe some need to read it again...............................
192.156.97.31
COVER STORY SYSTEM ON SILICON
------------------------------------------------------------------------
<Picture: ELECTRONIC BUSINESS TODAY>
The incredible shrinking OEM
As the system goes onto the chip, it may take the OEM's customer with it
------------------------------------------------------------------------
What does it take to start a PC company?
A credit card and a screwdriver.
--Silicon Valley riddle
------------------------------------------------------------------------
By Robert Ristelhueber
<Picture: story image>The technology to create an entire system on a single sliver of silicon, once a remote fantasy, is fast becoming a reality. As a result, the whole electronics industry is now grappling with the question: If the chip becomes the system, why do we need the systems company?
The advent of deep submicron technology is causing a seismic shift, overthrowing the existing balance of power in the industry. Intellectual property, and the profits it brings, is shifting away from the OEM and toward the semiconductor maker. This has been demonstrated most dramatically in the personal computer industry, where Intel Corp., Santa Clara, has grown wealthy selling the brains of the system, while Intel's OEM customers scramble for slim profit margins.
Given that the heart of the system has become a commodity available to all, how can OEMs differentiate their products and thus avoid becoming glorified distributors? And with more of the systems knowledge residing at the semiconductor maker, where is the boundary between the chip company and the OEM? How can the two sides adjust their relationship to recognize this new reality?
It's not just the PC industry that's feeling the heat. "Historically, you had a few consumer electronics [systems] companies, an oligopoly that would have the engineering wherewithal to develop a VCR or a camcorder," says Alexandre Balkanski, chief executive of C-Cube Microsystems Inc., Milpitas, CA, a manufacturer of video compression chips.
"As you go digital, the task of building a piece of electronics is much simpler. So we're enabling [OEMs] that were no-names until they became our customers. I think it will fundamentally affect the success and structure of the old consumer electronics industry," he adds.
As an example, Balkanski points to the explosion in video compact disc players sold in Asia as a result of the karaoke craze. Unlike past waves of consumer electronics products, the dominant Japanese firms like Sony and Panasonic were forced early on to share the market with systems bearing names like Idall, Malata and SAST, built in China and other developing countries. The difference this time was the wide availability of low-cost integrated circuits that perform the crucial video compression function, making it easier for competitors to quickly jump into the market, he contends.
Some argue that high-end systems, like the switches used to handle data networking, are immune to this trend. But the potential exists even there for a shake-up in the hierarchy. For example, one start-up IC company, MMC Networks Inc., claims that its chipsets can duplicate the functions of an Ethernet or asynchronous transfer mode (ATM) switch. <Picture: picture of Bami Bastani>Chip companies with systems architecture knowledge can dominate a business, contends Fujitsu Microelectronics Executive Vice President Bami Bastani.
"In the past, a company like Cisco did nearly everything themselves," says John Kennedy, director of marketing for Sunnyvale-based MMC. "Now they don't have to go off and invent the switch architecture, they can buy it from a company like ours. That makes the barrier to entry for non-traditional networking players much lower."
Cisco Systems Inc., San Jose, is using MMC's chipsets in its Lightstream 1010 ATM backbone switch. But MMC has found other customers for its silicon, including Hitachi Ltd., Tokyo, which used the chipsets to quickly set up a new networking division, Kennedy says.
The key for networking OEMs is adding value in the software, Kennedy believes. "Networking companies are going to rely more and more on software to differentiate themselves, because the hardware is becoming commoditized," he adds.
Software seems to be a key value that many OEMs can add.
U.S. Robotics Inc., for example, uses programmable digital signal processors (DSPs) from Texas Instruments Inc., Dallas, rather than buying hard-wired modem chips. This allows the company to add its own intellectual property.
"When all the intellectual property is trapped inside the silicon, there's very little value you can add," says Tony Francesca, vice president for OEM business at the Skokie, IL, company. "The reason we're able to stay ahead of the market is that our modems are softbased." The company has assembled a team of DSP experts to spearhead new product development.
When U.S. Robotics wanted to upgrade a large number of modems it had installed at Internet Service Providers (ISPs) from 28.8 to 33.6 kilobits/second, it was able to do so in a few weeks by downloading code. "It's not that easy if you have to replace hardware boards," Francesca says.
The company has grown phenomenally, from just $50 million in revenues in its 1990 fiscal year to more than $2 billion last year. It recently agreed to a $7 billion buyout from 3Com Corp., Santa Clara. And in an unusual twist, it has licensed its technology back to TI, which is using it in a new line of modem chipsets.
Nowhere to hide
But adding software may not be an option in the computer industry, where both operating system and applications software have become a shrink-wrapped commodity. That leaves systems makers scrambling for ways to stand out from the crowd.
Some contend that there's still room for OEMs to maneuver on the hardware front. "The way you put it together on a system level is very significant," asserts Bernard Guidon, vice president and general manager of Hewlett-Packard Co.'s computer operation in Grenoble, France. "The way you run your bus, the way you access your memories, graphics, I/O and multiple processors, can all be significantly different from one company to the other.
"This is not a static partition, what goes in a chip and what goes in a system, so the best thing you can do for your customer is to have very close cooperation between the system designer and chip designer," he adds.
Joseph Daltoso, chief executive of Micron Electronics Inc., Nampa, ID, a major producer of personal computers, agrees. His company spends a lot of time telling its motherboard and chip suppliers what Micron's customers want, and learning about new semiconductor products entering the pipeline.
Although Micron's research and development costs are minuscule, amounting to about 1% of sales, "we're putting more engineering resources into making sure everything works, integrating the chips and the chipsets," Daltoso says. "If anything, our job is becoming more complex because there are so many options with the new technology."
Nevertheless, few can dispute that control over intellectual property now resides at the chip level. Both sides recognize this new reality. In the case of the chip companies, it means walking a fine line between enabling OEMs with the latest technology and yet not competing with them directly.
Although he proclaims that Cypress Semiconductor Corp., San Jose, is a systems company because it understands how each component works and how to interface the components together, T.J. Rodgers says he's not competing with his OEM customers. "There's a clear dividing line," says the president of the San Jose-based company. "There are things we can make more efficiently, and there's a higher level of integration [the OEM] can make."
But that line seems to be constantly and quickly shifting toward chip companies.
At some point, as much as one-third of all chips could be one-chip systems, believes Larry Jordan, senior vice president and general manager of the System IC Division of Toshiba America Electronic Components Inc., San Jose. "In many [electronics products] there will be only one IC in the system," he says.
But OEMs needn't fret, he reassures. "As soon as we learn how to put everything onto one chip, then they say 'Let's put more into the system.' "
Cypress' Rodgers adds that "from a practical point of view, it would be really stupid for me to say to the market, 'Here's your switching system, better than Cisco's and smaller.' Not that I'm capable of it, but it wouldn't make sense for me to attack Cisco or AT&T, or annoy IBM."
Ed Sack, chief executive of Zilog Inc., a semiconductor company in Campbell, CA, is equally diplomatic. "Our philosophy is that the control will always reside with the OEM, even though the chip guy is doing more of the system," he says. "He's in touch with the end market and will always know more than we do."
Even Intel, which controls the life-blood of the PC industry, eschews the mantel of power.
"We can put more and more of today's functionality into one hunk of silicon, but the guys who are doing add-in cards and accelerator cards aren't going away," says Craig Barrett, executive vice president of Intel. "They're just taking the next step up and getting the next level of performance. It's this endless chase of innovate and integrate."
Intel perfected the art of tightrope walking between the two camps by becoming a major producer of motherboards and through its aggressive branding campaign. "People don't walk into a store and ask for an IBM compatible or a Packard Bell or Acer computer anymore," notes Tom Beaver, corporate vice president at Motorola's Semiconductor Products Sector, Phoenix. "They ask for an 'Intel inside.' "
Barrett disputes that by providing motherboards to OEMs, Intel and other suppliers are commoditizing the business. "They've just shifted their resources. Those guys are moving upscale in their capabilities, giving you workstation performance instead of desktop PCs." OEMs can add value in a number of ways, including software-hardware integration, he says. "There's also the service and reliability aspect of it, and the distribution channel, obviously."
Going with the tide
For their part, OEMs are being forced to accept the new power structure and learning how to swim with the tide, rather than against it. After all, the shift in the relationship between chip and OEM companies hasn't been caused entirely by the growing complexity of chips. The growing pressure to bring systems out faster has forced OEMs to rely more on canned solutions. <Picture: picture of Alexandre Balkanski>"We're blowing up the OEM structure," claims C-Cube Microsystems' Chief Executive Alexandre Balkanski.
"By leveraging other people's resources, they can meet the market's needs better. If the chip folks can give them a standard product, they don't need to use engineering dollars to duplicate that. That lets them differentiate using software, or in some unique part of the hardware," notes Don Ciffone, president of chip manufacturer Exar Corp., Fremont, CA.
The shriveling of the engineering departments of OEMs is forcing semiconductor vendors to alter their relationships with those companies, according to Motorola's Beaver. "Every one of these customers feels a severe pinch in his profit and loss statement, and their engineering line on the P&L has to shrink. So, you've got to provide more and more of the solution," he says.
Lean on me
Beaver uses the term "silicon community" to describe the interaction. "It's more than a buyer-seller relationship, or even a partnership." In addition to co-designing systems, chip suppliers are being asked by customers to coordinate the activities of their contract manufacturers and industrial distributors, to hold more of their inventories, and even to work with shippers like Federal Express to help their deliveries to clear customs in foreign countries.
Iomega Corp., Roy, UT, parlayed a tight relationship with Motorola into a successful launch of its Jazz removable storage drive. "It was a sweet marriage," says Wayne Stewart, senior vice president and chief operating officer at Iomega. "It allowed us to break through problems of time-to-market." The project was brought to fruition in just nine months.
The key was to combine proprietary logic developed by Iomega with Motorola's 68H16 microcontroller to create a single chip. Motorola stationed its engineers at Iomega at the earliest stages, and they worked closely with Iomega's employees to define the product. "They brought significant expertise in designing ASICs [application-specific integrated circuits], which we needed," Stewart says. "There was a clear division of responsibility over who would take possession of which piece of the design," adds Joe Janega, Iomega's senior strategic commodity manager.
Such relationships aren't possible with every customer, however. "Iomega knew we weren't getting into the disk drive business, so they weren't paranoid about what we were learning there. You have to have pretty progressive managements that go along with it," says Motorola's Beaver. And obviously, not every customer is large enough to warrant expending such large resources, he adds.
------------------------------------------------------------------------
<Picture>"When all the intellectual property is trapped inside the silicon, there's very little value you can add" - US Robotics' Vice President Tony Francesca
------------------------------------------------------------------------
In another instance, Motorola helped Uniden America Corp., a worldwide supplier of telephones, get a project off the ground. "We came up with the idea to take a cordless phone to connect with e-mail on the Internet," says Greg Jones, vice president and general manager of the Internet products group, Fort Worth, TX. "We needed to have a partner in technology to get to market quickly, because we had the expertise on the phone side but not on the Internet side."
Motorola put Uniden in touch with several companies that helped it with modem and software technology, and industrial design. "It's becoming the norm for semiconductor manufacturers to help a company like ours make this happen. No one company has all the technology," Jones says.
The shift in the balance of power to the chip companies doesn't come without a price, however. Chip companies are finding that OEMs expect them to have significant systems knowledge.
For example, Texas Instruments has invested millions of dollars to build a pair of laboratories in Dallas that are stacked to the ceiling with networking and wireless equipment purchased from a wide number of suppliers. "My value to my customers is my systems knowledge," says Richard Templeton, president of TI's Semiconductor Group. "That means I have to understand networks and wireless communications and disk drives deeply enough so that I can come in and say, 'Here's your three best options to partition this system.' "
TI acquired Silicon Systems Inc., Tustin, CA, last year largely to obtain that company's read-channel expertise, valuable in disk drives and digital versatile disc (DVD) products, Templeton says. "I don't think any one approach is sufficient to gain systems knowledge. You acquire companies, hire and train people, work closely with customers."
Where's the beef?
Reflecting the need to beef up their systems knowledge, more and more semiconductor companies are hiring systems engineers.
"There's a cross-pollination of talent. People are showing up at IC companies who used to be in systems companies. We spend a lot of time studying the architectures of imaging products and video products, looking for some commonality," says Exar's Ciffone.
In previous decades, Zilog would hire one systems engineer for every 10 chip engineers, notes Sack. Today, the ratio is nearly even.
"The number one differentiator for us is systems architectures," says Bami Bastani, executive vice president at Fujitsu Microelectronics Inc., San Jose. "It's a tough expertise to acquire, but semiconductor companies that develop systems architecture competence in a given area can absolutely dominate that area."
Whatever the future holds, it seems clear that to find the power in the electronics industry, one only needs to follow the intellectual property. And today, that belongs mostly to the chip companies. The best hope for OEMs seems to be to find a strong semiconductor partner and hold on tight.
"Systems companies used to compete on their engineering wherewithal, their system integration capability," notes C-Cube's Balkanski. "That's becoming less important. We're blowing up the OEM structure because we're enabling many more companies to participate." |
