PHOENIX " The word from the 2004 Semico Summit here last week was that the electronics industry has rebounded. Yet the industry's captains have yet to confront what the downturn hath wrought: a customer model that has changed fundamentally and forever.
Designers: Your customer isn't the OEM any longer; it's the OEM's customer. The era of design-for-your-mother, with all the complexities that entails, has dawned. CEOs and CFOs: Toss the data on historical cycles. You can no longer extrapolate from it, nor can CTOs rely on the steadfastness of Moore's Law.
Managers today talk of the consumer market the way they spoke of convergence a decade ago and PCs before that. But they view it as a market at which to throw technology and the tried-and-true techniques that served other markets during the industry's first 50 years: military/aerospace, industrial, PCs, cellular.
Consumer is different. It's a million applications wanted by a billion people, all increasingly tech-savvy and impatient. MIPS Technologies Inc. president and CEO John Bourgoin may have offered the best sound bite here last week: It's not about the next killer app, he said, it's about the next "killer experience."
The problem is that an experience is an innately emotional reaction " and emotional reactions often beget irrational desires, such as, "I want this faster, cheaper and cuter, today."
In the old days, a tiered system of gatekeepers protected designers from the sometimes-fickle demands of the consumer; now those barriers are gone. Engineers and engineers-turned-CEOs knew just what technology the world needed; today the world calls the shots. That has ramifications from the engineering bench all the way to the corner office " where high-tech CEOs seem to remain in place permanently, whereas their counterparts in other industries last an average of five years.
In this new world, stubborn forces are butting up against one another other. Leakage current, for example, rises tenfold every 1.6 years, yet the consumer expects annual improvements in battery life and performance. Materials and clever techniques are being brought to bear on the problem, but no one agrees on an approach " and if there's no agreement, then the myriad point solutions are by nature expensive.
National Semiconductor Corp. chairman, president and chief executive officer Brian Halla has talked enthusiastically of a sensor-networked world using "harvested" power from piezoelectric or solar sources. President and CEO T.J. Rodgers has spent some of his own money investing in solar technology for Cypress Semiconductor Corp. They wouldn't be doing so if leakage could be solved.
System-on-chip designs, considered a driver for the next wave of electronics, are a world away from the methodologies and incremental efficiencies of the past. Aside from the well-worn issues of intellectual-property integration, verification and test, SoC design poses unique problems of economics.
Jordan Plofsky, senior vice president at PLD vendor Altera Corp., has argued that there are very few profitable markets available for SoC designs that cost $30 million to $50 million to develop. The target market has to be at least $1.5 billion for a project to be considered a success. Plofsky used a calculation that assumed the project R&D was 20 percent of revenue and that eventual market share of that $1.5 billion market would be roughly 10 percent. So the revenue from the project would total $150 million.
Even Tsugio Makimoto, corporate adviser to Sony Corp. and formulator of the eponymous Makimoto Wave describing semiconductor industry trends, hedges, saying, "Will SoC be profitable? Yes in some cases, but not always." Not the most encouraging assessment of a technology.
Cost is king in the consumer world, and this is perhaps the biggest challenge facing semiconductor companies. Every chip supplier has the "right" cost solution, but often it's the right solution vis--vis competing solutions, not consumer perceptions. Consumers who revel at paying $30 for a DVD player at Wal-Mart won't be able to rationalize paying $150 for a solid-state tape recorder. A Sharp organizer is $12.88, but a Palm Vx is $79? What's with that?
The ASIC camp filed into town last week saying the market is coming back to a point where it may see an increase in the high-volume applications that have helped suppliers hit their numbers in the past. They're rallying around the concept of structured ASICs, in which a few metal layers are left for customization, saving time and cost.
Six-month cycles, 100k units
But the consumer world of the future won't be populated by Nintendo Game Cubes and Sony Playstations cranking out hundreds of millions of units. They'll have six-month product life cycles in the hundreds of thousands of units.
Ah. That's the purview of the programmable-logic camp, where never is heard the NRE word and the costs are reasonable in modest volumes all day. There's just one catch: The form factor of some consumer devices is far smaller than that of many programmable-logic dice.
The applications that open in the consumer world are virtually infinite. They range from smart phones with color displays that flash when you go to a disco to smart textiles, ubiquitous sensors and bioelectronics. These do not lend themselves easily to the 50-year-old notions of memory plus logic plus analog.
For 10 years now, marketers have been promising the consumer "experience on demand" or "video on demand" or "information on demand" anywhere, anytime. Stymied by tiffs over standards and gladiator-style battles over content protection, they have yet to deliver.
Now that the consumer is in control, the industry may simply have to come hat in hand and adjust the expectations it's built up over the years.
"Listen to the technology," Carver Mead said " but today there are too many to choose from, and too many unproven for a venture-capital and investment community that may now be permanently risk-averse. VCs not long ago could, by sheer force of will and capital, pull the industry out of a funk by making a few clever investments. Today, VCs are demanding cash-outs in two years (a ludicrous time frame) and booting out CEOs at the slightest sign of trouble.
Designers have a notion of where to go, but they're under the thumb of managers who have one eye on the bottom line and the other on the CNBC crawl. The problem isn't solved by 100 guys in a Bangalore office park with a T1 line, either.
National's Halla, who predicted the upturn and is fond of reminding us of that, demurred last week on the future of this cycle. "The next downturn is," he said, pausing for effect, "extremely difficult to predict. We're in a recovery " and it's completely different from anything any of you have ever witnessed."