Dawn of the Superchip
The past two years have been brutal on tech companies. Lots of promising ideas have died, underfunded, on the vine. Despite that climate, on Oct. 15, IXI Mobile Inc. of Palo Alto, Calif., landed $15 million in venture capital to push ahead with some snazzy gadgets of the future. These included a wristwatch that can receive e-mails and surf the Web and a ballpoint pen that can take photos and wirelessly transmit them over the Net.
All very James Bond, except for one thing. These aren't pricey, one-off prototypes for spies or millionaires. IXI's devices will hit the market early next year for as little as $30 apiece, plus another $150 for a slim router that can power them all from your coat pocket. The key: The gadgets run on a single Texas Instruments Inc. chip that's half the cost and one-quarter the size of previous designs. With a price like that, says CEO Amit Haller, "the response [from cell phone makers] has been overwhelming."
In the chip biz, less is turning out to be more. And raw horsepower is no longer the be-all, end-all of chip design. These days, semiconductor powerhouses such as Intel (INTC ), Texas Instruments (TXN ), and STMicroelectronics (STM ) are racing to usher in a generation of chips that can replace whole colonies of present-day processors by combining all their functions into one sliver of silicon, a "system on a chip."
To be sure, the push to combine functions is as old as the integrated circuit itself. But these systems represent a new stage. Industry insiders say that in the next few years, the rollout of powerful all-in-one chips will strengthen the hand of the behemoths at the expense of the industry's lesser players. "Just a small number of companies have the capability," says STM CEO Pasquale Pistorio.
For consumers, there is no downside: Single chips mean smaller, cheaper, and less-power-hungry devices. Some manufacturers, however, will struggle to survive in an era when the number of chips in every gadget is shrinking. The likely outcome is a market share battle of unprecedented intensity.
In fact, it has already begun. In the market for wireless chips, for example, deep-pocketed chipmakers are racing to combine on one semiconductor all the functions of a cell phone. Next up will be other high-volume gadgets such as digital cameras, MP3 players, and even personal computers. All of this is devilishly difficult to do, requiring the smarts to integrate a hodgepodge of functions onto one chip and the capital to crank them out in big numbers using the most advanced manufacturing processes. But those who can execute will gain a significant competitive advantage, says G. Dan Hutcheson, president of San Jose (Calif.) Consultancy VLSI Research Inc.
With the chip sector in such dismal shape, the whole move to systems on a chip has the air of a crusade. As demand for new tech gear has waned, total chip sales plummeted to $139 billion in 2001, from $204 billion in 2000. That has left the industry riddled with excess manufacturing capacity and desperate for growth. At the same time, the only way to shore up prices--which have been falling even as customers have begun restocking--is by pushing chips to do more. "I look at this as survival," says Michael A. Yonker, chief wireless technologist at TI, which is planning a single-chip cell-phone design for late 2004.
Nowhere is the competition tougher than in cell phones. Half a dozen wireless chipmakers are spending billions to build a superchip that will drive the cheapest possible phone. TI thinks its design--which reduces the total number of components in a cell phone to 25 from 185 today--will streamline assembly and cut handset prices in half. That would boost sales in developing markets such as India and China. "By lowering the cost, we can find the next billion users for wireless," says Richard K. Templeton, TI's chief operating officer.
TI's competitors are drawn to the same alluring economics. Intel Corp., the leading supplier of memory chips for handsets, is developing a single-chip cell phone with which it hopes to chisel away at TI's lead in processors. TI, meanwhile, expects to add enough memory to its single chip that handset makers will no longer need to buy it from Intel. And ST, a leader in analog and memory design, is pumping 70% of its $1 billion research budget into developing systems on chips for cell phones and other products.
The all-in-one chips should do more than just cut manufacturing cost and time. They also promise to rev up the introduction of technologies that bring with them bigger opportunities for profit. At Intel, for example, engineers have accelerated plans to build wireless modems into many of the company's chips by mid-decade. The result will be inexpensive broadband modems, digital audio receivers, portable game players, and video recorders that can connect wirelessly to the Net. "Every chipset, every processor, every memory device literally can become a radio," says Intel Chief Technology Officer Patrick P. Gelsinger.
Meanwhile, the chip wizards at TI are planning to embed global positioning systems into their cell-phone chips within a few years, helping emergency services locate distress calls and enabling advertisers to target customers by their location. Toshiba Corp. (TOSBF ) and IBM (IBM ) are spending hundreds of millions of dollars to design the all-in-one chips necessary for Sony Corp.'s (SNE ) next-generation PlayStation. "This is the technology that will set us apart from the others," boasts Yoshihide Fujii, vice-president of semiconductor operations for Toshiba.
The race will be hard on small chipmakers. In the scenarios of some forecasters, only the largest vendors will be able to shrink the filigreed circuit lines on chips beyond today's superdense limits. And companies must mix analog and digital parts so that real-life signals from voice and pictures can be processed by the same chip. Engineers will also need to isolate elements such as radio transceivers that can interfere with delicate digital circuitry and add complex software that can deftly manage these multitaskers.
Most single-chip designs will require the use of cutting-edge lithography techniques that can create transistors only 90 nanometers wide, a big reduction from today's thinnest lines of about 130 nm. Intel expects to be operating at 90 nm in its factories by mid-2003, and TI, IBM, and Advanced Micro Devices (AMD ) are also close. But only a handful of other chipmakers will be able to afford to join them, according to Hutcheson, who cites ST, Motorola (MOT ) , Taiwan Semiconductor Manufacturing Corp. (TSM ), and Samsung as the most likely.
Systems on a chip, in other words, may spell the demise of smaller, specialized chipmakers over the next few years. Intel, for instance, contends its efforts to integrate wireless networking into all its chips will render existing modem chipmakers obsolete. "If that's your domain, I suggest you move on," warns Intel's Gelsinger.
Some small players are sure to find ways to ride the industry's wave toward fewer and fewer chips. Silicon Laboratories Inc. (SLAB ) in Austin, Tex., which outsources production to TSMC, has managed to shrink the radio transceiver used in cell phones from 100 parts a year ago to 21 today. Next year, it will roll out a chipset with just one piece of silicon that will take up one-eighth the real estate of typical chipsets. With fast-growing customers such as Samsung Electronics and China's TCL Mobile, the upstart's quarterly revenues have soared 156%, to $41.2 million.
Growth is one thing. Profits are another. Analysts point out that the virtue of the do-it-all chip is that it allows semiconductor companies to siphon profits not just from niche chipmakers but from the packaging specialists that now assemble chips on boards, as well as other bit players in the electronics manufacturing food chain. TI says that when its single- chip design reaches volume production, it could boost its share of total cell phone revenues by 50%. "The system on a chip is a new market accelerator," says Richard Doherty, president of Envisioneering Group. Right now, any acceleration would be welcome.
By Andrew Park in Dallas, with Cliff Edwards in San Mateo, Calif., Andy Reinhardt in Paris, and Irene M. Kunii in Tokyo