What will it take to return to prosperity?

What will it take to return to prosperity?
Technology has a life cycle. When introduced as the applied practice of discovered science, new technology usually takes a while to catch on. But when-and if-it does, it quickly proliferates. The proliferation phase of technology's life cycle is, in many ways, its heyday. New companies are spawned and great wealth generated for its practitioners and the cognoscenti.

Widespread acceptance is the next phase. The wealth effect of the technology spreads to the marketplace at large and to society. The technology is widely used, standardized, available and subject to the laws of supply and demand. In other words, it becomes a commodity. The strong suppliers of the technology get stronger at the expense of the weaker. Unit volume continues to grow, but it gets harder to make money. Suppliers concentrate on manufacturing efficiencies. Startup activity slows.

Next is decline through displacement by a newer technology. Rarely does a technology disappear entirely. Displacement technologies may offer significant advantages but usually are one-to-one replacements. Thus, the older technology lives on as a niche.

The three great revolutionary technologies of the early 20th century were railroads, automobiles and aircraft. All were designed to connect and make the world smaller.

Railroads are in decline but live on as cheap, albeit slow, mass transport. Autos and aircraft are commodities. In fact, autos are a particularly egregious example of commodities as we define them. As a general trend they are sold in increasing volumes each year, but manufacturers are making less money on each car. Profitless prosperity is a term that well applies to the car-making business.

Our purpose here is to examine the life cycle of semiconductor technology, the newest and most powerful of the connectivity technologies. As the industry emerges from the severest revenue and profit downturn it has faced, our objective is to examine what it will take to make it prosper again. Or, will we find that the challenges it faces are so severe, on so many fronts, that profits-and therefore the health of the industry-will be in jeopardy for some time to come, until there is a permanent change in the landscape and its inhabitants?

One thing is clear: As it approaches $150 billion a year in factory sales, the semiconductor industry is not only among the largest manufacturing industries in the world, but also, arguably, the most important. No one need point out to readers of this issue that semiconductor technology has enabled the computing and communications revolution (and the Internet, for that matter) of the latter third of the last century.

Further, the longer-term sales prospects of the industry appear to be undimmed. PCs and cell phones are commodities today, but they continue to sell in large numbers. And, other, more powerful and more featured variants are in the works. Those will continue to consume more powerful semiconductors.

Devices that combine entertainment, mobile communications and computing are emerging as likely engines for the industry's near-term growth. The semiconductor business has always been pulled by certain equipment sectors. In its infancy, the industry was nurtured by demand from military systems. Later, it was the growth of the mainframe computer that sustained the demand. Later still, the cell phone and the Internet caused the explosive demand that peaked at the end of the 20th century.

In particular, the new engine for the industry stresses connectedness-specifically, pervasive interconnectedness, not just for consumer electronics but for all kinds of applications. That bodes well for semiconductors for wireless applications and for power management.

Even without brand-new applications that enable new kinds of interconnection capabilities, the silicon content of electronic equipment has been steadily on the rise. "There has been a stepwise increase in the silicon content of electronic systems with every generation," says Theo Classen, the CTO of Philips. "Digital equipment has more silicon content than analog equipment. Silicon is replacing copper and iron. In the TV, for example, the CRT is being replaced by LCDs, and the same sort of thing is happening in transportation," where drive-by-wire is replacing electromechanical linkages.

Transportation electronics-in cars and planes-may, in fact, be an important engine for the semiconductor industry. The electronic content of autos is rising rapidly, with semiconductors taking over engine and drive train control, cruise control, creature comfort and safety, and, more recently, navigation and multimedia entertainment. But it is likely that not all semiconductor vendors will find this market attractive. It takes a broad portfolio of capabilities, such as sensors, power handling, flash memory and controllers, which not all vendors can provide; the design cycle is long; and existing relationships are hard to dislodge.

A little further down the road, a vast new applications arena is about to unfold for semiconductors at the intersection of health care and electronics. "In 10 years, medical-monitoring systems will have a similar impact and be a similar driver to what the mobile phone has done for our industry in the last 10 years," says Sir Robin Saxby of ARM Holdings. "Biotech applications are on the way, and I think they will stretch the definition of a system-on-chip."

While the future is bright, the industry faces significant short-term challenges that will affect who makes money and, thus, who will survive. The landscape will be shaped by a number of converging scenarios.

The first is what STMicroelectronics chief economist Jean-Phillippe Dauvin calls "endemic overcapacity," caused by the bursting of the demand bubble over the past couple of years. Such overcapacity hangs like a financial millstone around the necks of most manufacturers and puts customers in the driver's seat in terms of pricing. That, in turn, affects how much money can be made and jeopardizes the future of the industry's weaker players.

"Semiconductors are maturing," says Ulrich Schumacher, CEO of Infineon. "There is a clear consolidation measured in terms of the number of players per segment of the industry." Schumacher points out that the top three players in DRAMs, for example, now hold a collective 75 percent share of the market, whereas 10 years ago the top 10 players held about half the market. Altera CEO John Daane points out that only two companies now control over 80 percent of the programmable-logic segment and that startup activity in PLDs is at a standstill.

DRAMs and PLDs have been around for a while, so a case can be made that consolidation is inevitable in older segments. But even in the newest and hottest growth segment, there seems to be concern: "Wireless semis are next," predicts Schumacher.

The decline in startup activity-while primarily affected by the economic turmoil-may be another indication of maturity. "If a startup does get as far as raising capital to cover design and manufacturing, then they have the challenge of breaking into a first-tier customer. Coupling product complexity with shorter product life cycles does indeed pose some challenges for established companies as well as startups, says Roland Pudelko, CEO of Dialog Semiconductor.

The second factor is mindset. The semiconductor industry, for the most part, has a decades-long and therefore unshakable belief in Moore's Law, which postulates that integrated circuits will double in complexity at a predictable rate with a concomitant decrease in manufacturing costs. The added benefit is that circuit complexity increases integrated functionality and reliability while reducing size, weight and power consumption. The industry has always believed that its ability to make things faster, cheaper, smaller (and therefore more packed with features) will spur an emerging market.

With the flattening of the industry revenue curve since 1995 (though it peaked massively in the year 2000), the industry has had to face the very real spectre of saturated demand for the computers and cell phones that have been its most recent drivers. The industry has had to face the possibility that, for the first time in its history, its driving markets may not be infinitely elastic.

The industry's technology leaders have reacted predictably to the situation. If demand drops, then reduce the manufacturing costs by increasing complexity: smaller, denser, more capable circuits on larger wafers.

If the leaders succeed in adhering to Moore's exponential, they will distance themselves from the pack. In moving to the next generation of technology, they will have imposed a tremendous financial burden on those with underused installed capacity in previous technology generations.

But it's not yet clear whether they will succeed, and at what cost. "For our industry," Moore himself has said, "many of the exponential trends are approaching limits that require new means for circumvention if we are to continue the historic rate of progress. Business as usual will certainly bump up against barriers in the next decade or so."

In a now-famous keynote speech delivered at the 2003 International Solid-State Circuits Conference, Moore pointed out that "a new and more fundamental barrier must be confronted in the next decade: the fact that materials are made of atoms and technology is approaching atomic limits."

The problems to be solved are many. They fall into two categories: increasing device speeds and controlling leakage currents. The solutions are not very clear. Undaunted, technology leaders are looking for "boosters" that wring incremental performance out of current technologies. Among them are such techniques as "straining" the lattice silicon to improve the mobility of charge carriers, high-k gate dielectrics, metal gate electrodes and fully depleted silicon-on-insulator. Researchers are also exploring a departure from planar structures in the fundamental building blocks of the IC, in the form of multigate devices such as FinFETs.

But will those efforts pay off even as the technological obstacles are scaled? As Moore put it in his ISSCC speech, "The technological challenges continue to escalate, but so do the financial challenges. Clearly, in an industry where revenue growth seems to be slowing down, this represents a formidable obstacle."

Already, the costs of scaling the obstacles are so huge that only the largest of the semiconductor manufacturers can hope to work on them. What about the rest of the industry?

Some companies are resorting to joint R&D and manufacturing efforts. Others are betting that the specialized manufacturing foundries will bail them out. "It used to be that the foundries were a half-step behind the state of the art, but not anymore, " says Wilf Corrigan, founder and chairman of LSI Logic. In fact, the foundries may have an inherent advantage over captive fabs. Foundries can fill their fabs with products from a number of designers. Captive fabs need increasingly scarce high-volume products to make their operations more effective.

On the other hand, the more complex-and hence the more subject to instability-the manufacturing process, the closer its link with the design process must be. A tighter, more iterative integration of design and manufacturing may present new challenges to the foundries, their fabless customers and the EDA industry. "The EDA industry is somewhat behind our requirements. The problem is that they can never agree on standards," says Corrigan.

Already, one of the experiments in the disaggregation of design and manufacturing has not lived up to the hype: The plug-and-play virtual component business model has not caught on, with the notable exception of ARM Holdings, whose microprocessor intellectual property is ubiquitous.

The third problem the semiconductor industry faces has to do with the supply-chain dynamic. The technology is unique in that, by increasing "packing density," semiconductors can be made both smarter and cheaper at the same time. The industry has used this to grow end markets. But it also means that with each successive technology generation, the chips inside the box have accrued more and more value.

Nothing demonstrates that better than the PC. Virtually all of the intellectual property of a PC resides in the chips inside. PC makers-led by Dell, which made a virtue of a necessity-have been forced to abdicate the creation of value around product functionality in favor of creating value around assembly, distribution and marketing.

What happened in PCs is about to happen in cell phones. A host of semiconductor manufacturers is readying standard chip sets, even single-chip phones, that could be assembled as phones with value creation beyond that relegated to assembly, distribution and marketing. If control over the architecture and domain knowledge in cell phones shifts to the semiconductor manufacturers, then handset makers like Nokia may be doomed. As articulated by Cypress founder and CEO T.J. Rodgers, this may a fundamental law: "If I can make it, you're hurting, because I have control over the silicon."

That, in itself, may not appear to be bad for the semiconductor industry. The problem has to do with the peculiarities of the consumer market. Today and in the near future, the industry is looking to the personal consumer electronics market as the high-volume engine. That means that, as it absorbs more and more of the value in consumer electronics products, the industry is getting closer to the individual consumer. And consumer markets are notoriously low-margin markets.

"The consumer market is not the place to expect high margins, " says Ulrich Shumacher, CEO of Infineon. "When growth rates were 30 or 40 percent per year, everybody was a winner. Now, the wireless communication market is in the process of being commoditized."

But that's nothing new to LSI's Corrigan. "The consumer was always the ultimate driver," he says. "Today the boundary between the consumer and the chip is getting transparent."

The problem is that the semiconductor industry does not have a stellar track record in serving the consumer directly. In previous generations, the industry migrated up the value chain by manufacturing watches, videogames and calculators, only to realize that it should not have been in those businesses.

While LSI Logic does not make a fully packaged DVD player, it does create almost all of the functional value. But given the highly efficient distribution mechanisms in place for consumer electronics, the time-to-commoditization for a new product has shortened dramatically. For an $80 DVD player, which is generally reckoned as a price point for an impulse buy, the margins on the chips inside are severe.

The consumer's heart
Though it's clear that the semiconductor industry is getting closer to the individual consumer, it's not all clear that it really understands the dynamics of taste and style that create perceived value in the consumer business. "We [the semiconductor industry] don't understand the mass market yet," ST's Dauvin says flatly. "We have to shift from selling applications to selling usage. For example, personalized casual electronics may need a different approach than optimizing the price/performance ratio." The exception may be Japan, whose semiconductor makers, beleaguered of late, are hoping that their expertise in consumer electronics will enable them to regain their prominence.

The only solution, from the semiconductor industry's point of view, is to make manufacturing more efficient. But that requires high-volume-standard, not custom-products. One way the industry intends to do that is to move from "pure components to platforms," says STMicroelectronics chairman Pasquale Pistorio. The platforms-say, for handsets-would be high-volume standard semiconductors that could be customized through software. Platforms would be available through a number of sources. Semiconductor makers would accrue manufacturing economies, and handset manufacturers would differentiate through software-supplied not by Microsoft, but by the semiconductor maker.

"Software will be more important than the silicon, " says Pistorio. "The world needs interoperable platforms. The price of a PC would have come down faster had there been one."

Platforms are application-specific standard products that are programmable for a small set of high-volume applications. But, in general, the idea of programmability may be the way to balance the requirements for high-volume products with the custom needs of disparate applications.

"There is a definite trend toward programmability," says Altera's Daane. "For R&D payback, we need high-volume chips in several markets for several customers." Interestingly, the platform approach, advocated by standard-product manufacturers, and the less application-specific programmable approach advocated by programmable-logic manufacturers appear to be converging. Altera and others are adding more structure, at the expense of pure flexibility, to their FPGAs. "We're providing IP blocks targeted at a certain customer base. It's an ecosystem. You want to integrate what's most often used, but not do too much of it," says Daane.

Both platforms and programmable logic are approaches designed to reduce the manufacturing costs of advanced technologies. They are driven by the industry's infinite belief in the veracity of Moore's Law.

But what of the capabilities available in current-generation technologies? "The problem is finding applications for existing technology that is stable," says Corrigan. "How do we apply this technology?'

The answer, to some extent, may be found in the business model of the analog segment. Analog semiconductors do not tax the state of manufacturing art. Further, they find more widespread application, and, they are standard, generic building blocks. The bottom line is that analog-semiconductor companies have generally fared better in the downturn, since they are not overly dependent on the vagaries of one application segment, and they continue to be hugely profitable. "We make more margin today than we did four or five years ago," says Linear Technology founder and chief executive officer Bob Swanson. For the most part, they've done it through clever designs. Can digital CMOS find inspiration in the analog model?

For starters, here's a fundamentally different approach from the digital world's focus on reducing manufacturing costs. "People will always pay for performance," says Swanson. "I'd rather spend my time on trying to figure out how to increase the value of a chip by 50 cents than trying to reduce cost."

It helps that every digital system needs some complement of analog circuitry. "It doesn't matter what drives the market," asserts Swanson. It also helps that power-and power management-and high voltage are exclusively the domains of analog. "Power is the analog piece," says Swanson. And, in addition to clever circuit design, the analog segment of the semiconductor industry has led the way in packaging.

In the end, though, it may be a matter of mindset. The analog industry has aggressively gone after new applications because it has had to do so. It has adhered single-mindedly to increasing value, not reducing manufacturing cost. But while analog manufacturers don't push the state of the art, they maintain absolute control of manufacturing technology. "I have to have a lot of control," says Swanson. "I have to tweak the processes."

Cypress' Rodgers is taking that one step further. Cypress' push into leveraging its silicon expertise is exemplified in its investments in solar cell technology and MEMS. Neither of these use state-of-the-art manufacturing processes; rather, they use the fundamental economics of silicon processing to create totally different applications that diverge from traditional CMOS pursuits. "What separates people are the architecture and the application. Manufacturing silicon is the game," says Rodgers.

Everyman and everything
If so, then the semiconductor industry needs to find those applications-beyond applications in computing and communications. It needs to find new high-volume applications that can be tackled without relying exclusively on the ability to make ever-smaller transistors. The historical resourcefulness of the industry in solving seemingly intractable problems in the past is rooted in engineering innovations. Now it may need to look beyond engineering and better understand the human condition first-what makes life better, convenient and safer-and then supply the technology solutions.

LSI's Corrigan is optimistic. "Thanks to the Internet, we have millions of futurists thinking up applications. The Internet has made Everyman the man who knows where to find everything."

Girish Mhatre has a long history with EE Times as semiconductor editor, editor in chief, and publisher, and continues to be one of the industry's great influencers.