In Search Of the Perfect Chip

A new wave of semiconductors is going to disrupt every business.

May 1 2005 by Chief Executive

Everyone knows that modern communications have transformed the nature of business. Chief executives are certainly aware, having watched telecommunications and related computing trends€¦quot;from cell phones to wireless LAN to Internet backbone routers€¦quot;change how their companies work.

Get ready, because another period of convulsive change is about to rip through not just the tech sector but through every company that uses technology. That future can be glimpsed in the semiconductor industry’s race for “convergence,” which is happening alongside fundamental changes in how chips are made, where they are made and how much can fit on them.

Convergence, a term used to express how all manner of telecommunications and entertainment are coming together in single devices, has made connectivity a part of the very DNA of electronics. The market is demanding that voice, video and data be transmitted on the same network and that these services be available anywhere on most any device. Chips are at the heart of the systems that will allow that. “I think it is all driven by the desire of people to be mobile and to have as much connection to the world as possible,” says Agere Systems John Dickson, whose Allentown, Pa.-based company is at the heart of the fray.

The hunt for the winning semiconductor, a kind of a search for the Holy Grail, will transform the chip industry, as companies that can’t combine all the functions that are in demand fail to survive. The sectors that exist today within the chip industry, such as communications, graphics and data processing, will be smashed together. Freescale Semiconductor, Austin, Tex., is the former Motorola semiconductor unit that was spun off. The company has concentrated on communications chips and will increasingly compete directly against Intel, the dominant provider of chips today. All of that will likely lead to market consolidation, with some innovative startups disappearing and the concentration of control of the technology in the hands of a relatively few companies.

That could profoundly affect pricing. CEOs who would normally see the chip market as having no practical impact on them will find that they need to pay attention for the sake of their own profitability€¦quot;because the accustomed combination of new capabilities and falling prices may cease to be business as usual.

Technology Truly Born of Demand
Because communications technology is so important to every industry, it is easy to forget that, compared to the 40-year-old integrated circuit, these specialized semiconductors are adolescents. The reason is that the standards-based data communications they support are relatively new. Cisco didn’t sell its first network router until the mid-1980s, a few years after IBM introduced the PC. As data communications became increasingly important, the semiconductor industry began to turn its attention to creating specialized chips to support these types of applications. Motorola marketed the first “communications processor” in 1989.

The communications semiconductor industry blossomed because the demands of companies and consumers have blurred the lines between types of data, content and communications€¦quot;and convergence was born. Laptops had modems, then Ethernet support, then wireless. Mobile computers supported low-speed data connections, then broadband mobile connections. Business people needed remote anywhere, anytime access to the corporate network and databases. Consumers wanted to download thousands of songs to their pockets, read email that arrived on PDAs and get stock alerts delivered by phone. Even wireless transmission of TV programming, sometimes to cell phones, is becoming common in countries such as Italy and South Korea, notes Svetlana Issaeva, research analyst with Pyramid Research’s U.K. branch. “The big trend in the industry is that all these consumer electronics products are just becoming features of a cell phone,” says Paul Jacobs, CEO-elect of Qualcomm.

Looking beyond the handset-centric view of the San Diego, Calif.-based company, it’s clear that communications have become ubiquitous, whether as the remote assistance capability of an automobile outfitted with OnStar or factory-floor equipment connected to a manufacturing control center thousands of miles away. “You’re merging networks just built for data, video networks just built for data, and voice networks into a single architecture,” says Steve Rago, an analyst for electronics market research firm iSuppli, El Segundo, Calif. Increasingly, the premium will be put on selling chips that do more than offer such specialized functions as Ethernet, wireless support or Voice over Internet Protocol (VoIP). It’s a premium that will put many of the smaller innovative communications semiconductor companies out of business.

There are other forces destined to shake up the semiconductor industry. Since the mid-1960s, integrated circuits have largely obeyed the observation of Intel co-founder Gordon Moore that the number of transistors that fit on a chip doubles roughly every 18 to 24 months. What has made the march of power a reality is new manufacturing processes that allow companies to etch circuitry onto silicon in ever thinner strands. These days companies need only 90 nanometers, or just over 1/500th the thickness of a human hair, to create a transistor.

But the advances have come at a price. “At 90 nanometers, the design time [testing and production setup] will cost a couple of million bucks or maybe even more,” says Len Rand, managing director of San Francisco-based Granite Ventures. That price is up fourfold from just a few years ago, and even more powerful and expensive manufacturing processes are on the horizon. There are two results: Small semiconductor companies find it increasingly economically risky to bring new chips to market, and all the companies have to bear higher costs per chip. Because electronics equipment manufacturers are highly price-sensitive, the chip vendors must add more functions per chip to justify increases.

Another megatrend is the downsizing of engineering departments among the chip companies’ customers. Greg Lang, CEO of communications semiconductor company Integrated Device Technology, Santa Clara, Calif., says he has seen his clients cut their engineering staffs literally in half. “Our customers are going to have to rely more on off-the-shelf and standards-based solutions,” he says. “It translates into semiconductor guys delivering more value than they have in the past.”

Value means more elaborate reference designs, more software to run the chips and more help from semiconductor engineers. It is a second reason that companies must integrate a greater number of functions, because their customers have fewer people who can create a design out of more chips. And when looking overseas, many electronics companies would rather save the expense of a larger staff.

The companies that assemble electronics, in China and elsewhere, are demanding that more and more functions be integrated on the chips they buy. In effect, they want to buy brains off the shelf. “They’re perfectly happy to do nothing but design the casing and pick the color,” says Rand, because leveraging existing designs from semiconductor vendors lets them get to market faster, and in the electronics industry, a delay of a few weeks can literally cost a company one-third to one-half of its total margin on a product.

According to Indro Mukerjee, chief marketing officer and executive vice president of Philips Semiconductors, the willingness of the Chinese companies to use almost-ready-made designs is paying off. “We had some Chinese customers go from concepts to products in the shops [using our mix of hardware and software] in six or seven months,” he says. Such a head start translates into lower cost of development and premium pricing from early buyers, and thus, greater profitability from the product line. The practice can be found in more established companies as well: Apple used a hardware reference design for the guts of its iPod and relied on case design and software interface to create its popularity.

If all that wasn’t enough, consumers constantly demand portable electronics that are smaller, lighter and capable of running longer on a battery charge. Given the way that engineers design modern hardware using integrated circuits, it’s almost always better to replace a group of chips with a single one that takes up less space and that consumes less electricity.

The only practical way for a semiconductor vendor to supply one of those chips and to command a price high enough to offset the increased cost of chip design and manufacturing is simply to provide more functions on its chip.

Communications Takes a Back Seat
Unfortunately for the communications chip industry, intellectual property for most communications functions is easier to license or buy for integration into an integrated circuit than other know-how. That suddenly makes every company making integrated circuits for specific applications a potential competitor. One example is a leading graphics semiconductor company, NVIDIA, Santa Clara, Calif. It has a chip used in Microsoft’s Xbox that provides not only 3D audio and sound effects, but also the unrelated ability to provide broadband connectivity, a highly unusual combination. Competitor ATI Technologies, Ontario, Canada, offers set-top box manufacturers high-end audio, video streaming, interactive electronic program guides, a graphics engine and peripheral connectivity in one unit. In each case, the signal processing of audio, video or graphics is usually far more difficult to acquire or reproduce than some aspect of communications.

In other words, communications will quickly become a standard feature in many chips. The only hope a company has, says Agere’s Dickson, is to provide 60 to 70 percent of everything the customer needs for any given application. “Each of us will be finding different ways of bringing in intellectual property€¦quot;either license it, develop it ourselves or [get it through acquisitions],” he says. “The issue is how well do you acquire or develop the intellectual property and how well do you integrate it.” The new roles are likely to cause business frictions, as well. “Traditional partners will suddenly find themselves competitors in parts of their business,” Rand says.

That’s not to say that more specialized semiconductor companies haven’t a chance€¦quot;far from it. Qualcomm, for example, has continued to maintain existing key patents and develop new ones in the cell phone industry, making it difficult for a competitor to go behind its back.

Then there are companies like Freescale, which builds high-powered chips for the niche market of networking infrastructures. These will have a ready supply of customers because networks will need increased speed to withstand the onslaught of traffic that the increasing demand for communications services will inflict.

But smaller communications integrated circuit companies will face significant problems. They won’t have the existing broad base of areas of expertise to which they can add communications. For solutions, customers will likely look first to companies that offer what is perceived as the main function needed. For example, given its reputation in graphics, it would be easier for an ATI to add connectivity into one of its chips than for a company with experience in wireless LAN chips to start providing intensive graphics processing.

And while the smaller company may be able to continue developing its communications offerings, it is unlikely to add at the same time the application-specific exper-tise it will need. Because the small firms will find it increasingly costly to design and produce independent offerings, they might well turn to licensing their intellectual property to application chip companies. “There are a whole ton of companies out there now whose raison d’etre is just to create intellectual property and license it,” says Dickson. Realistically, though, most won’t survive, and with them will go important sources of innovation.

The winners aren’t home free, either. Because much of communications is becoming a commodity added in to sweeten another product offering, the vendors face the danger of thinking they must give the functions away to stay competitive. “You can destroy the value of integrating if you’re not careful,” says Mike Fister, the former Intel executive who is now CEO of electronic design tools vendor Cadence Design Systems, San Jose, Calif. Companies that can’t segment product lines, giving away some communications features and charging premiums for others, will need to make up for their costs in increased sales volume or market share, and they won’t have the financial justification to hotly pursue R&D.

Looming is at least one big unknown. As communications becomes more a part of integrated solutions, the semiconductor companies need better understanding of their customers, which means recruiting management talent from other industries. Companies are having to make trade-offs, finding people with one strong background, say in broadcasting or content, and then complementing them with people from other backgrounds, such as networking or consumer electronics. “You’re going to see an emergence of new executives and new opportunities within companies,” says David Westberry, senior client partner and leader of the semiconductor practice of Korn/Ferry International, headquartered in Los Angeles, Calif.

It’s impossible to know just how fast the industry will develop and perfect the new generation of chips, so it’s also impossible to know just how quick an impact they will have on the broader IT marketplace. But the trend lines seem clear. A “next new thing” is about to disrupt the IT industry and even nontechnology products. In a decade, communications technology will find its way into virtually everything, from furniture and building products to fabric and paper. The CEOs who can anticipate this change and invest their IT dollars wisely will be the winners.