A Whole New Chip Game

Protected by guards and hidden from roadside view, the headquarters of IBM's mammoth chipmaking operation--the world's largest--is a forbidding place. Until recently, the 216,000-square-foot Somers (N.Y.) building didn't even have visitor parking. But now, Somers executives are throwing open their doors, and, more important, are offering to sell rivals IBM's most advanced technology.

It's just one sign of a major plunge by IBM, DEC, Hewlett-Packard, and a handful of other electronics giants. After years of making chips only for their own products, these companies are seeking new revenues from sales on the open market. The chip business is already so competitive that the newcomers won't turn it topsy-turvy, but they could shake out a few weaker players.

The biggest impact is on the neophyte chip-sellers themselves. Early this year, IBM broke tradition and publicly began selling memory chips--including state-of-the-art 16-megabit chips that haven't even appeared in many IBM products yet. On May 10, IBM offered to build custom chips called application-specific integrated circuits (ASICs) that are at the core of everything from mainframes to advanced voice-recognition systems. Hewlett-Packard Co. is also pushing ASICs, while Digital Equipment Corp. is to add to its microprocessor sales on May 25 with a new line of communications chips.

NO HERMITS. These owners of "captive" chip operations used to feel that outside customers would distract them from serving their internal needs. Today, outside customers have become a necessity. In an era of increasingly open standards, computer makers want to make their new chip designs industry standards by selling them to as many competitors as possible. The discipline of selling outside can also help producers reduce costs and boost quality. And with tomorrow's chip plants projected to cost $1 billion and up, computer companies can no longer afford isolation. Unless they offset the huge costs of running their chip factories, they face massive write-downs on existing plant investments, analysts say.

That's IBM's situation. With its computer hardware revenues growing at mere single-digit levels, IBM could not afford to fund chip R&D and operating costs that are rising 15% to 20% annually, says Jon E. Cornell, the executive organizing IBM's push into commercial chip sales. "This is not a hobby," says Cornell. "We have to do this." He says he's aiming for $1 billion in outside chip sales within two years-up from zero last year and an estimated $300 million in 1993-to customers in computers, communications, and consumer electronics.

In some ways, the timing couldn't be better. The semiconductor industry is undergoing a sea change--away from generic chips and toward entire systems on silicon that incorporate knowhow once scattered among several chips and software programs. After decades of building complex systems, computer companies are rich in such skills. They can take heart from the experience of American Telephone & Telegraph Co. It stumbled for years with generic chips, but then marshaled its systems skills and today has a newly profitable, $2.5 billion operation whose merchant sales grew 30% last year to No.20 in the world, according to market researcher Dataquest (box).

BIG ARSENAL. Because of such gains, semiconductor companies aren't taking the new entrants lightly. "We have to be concerned with them in the mid- to long-term," says Kevin McGarity, senior vice-president at Texas Instruments Inc. Hajime Sasaki, an NEC Corp. senior vice-president, says the entry of systems builders such as IBM and Hewlett-Packard into the chip market "is going to make life harder for smaller chip companies that don't have products with a lot of differentiation and don't have the same knowledge of systems."

Among the computer giants, IBM has the highest ambitions, with money-makers such as memory chips, microprocessors, and ASICs, of which it is estimated to be the world's largest producer. With state-of-the-art production lines in East Fishkill, N.Y., and Essex Junction, Vt., "IBM has some great technology," says Zenith Data Systems Executive Vice-President David F. O'Connor. "We're talking with them, and we're very excited."

Still, success in these markets will require huge changes in the way the captives make chips. High costs that used to be absorbed in the bookkeeping won't survive. Nor will idiosyncratic sizes and packaging techniques that throw the companies out of line with the chip industry. Already, changes are occurring: IBM bowed to the wishes of Motorola Inc., its partner on the PowerPC microprocessor, and agreed to co-design the chip with commercial software that Motorola knew well instead of using IBM's own design tools.

The open market can be a challenge for newcomers. IBM lost the chance to sign up its first national chip outlet when Intel outmuscled it to win over chip wholesaler Anthem Electronics. "They felt the return on Intel would be better short-term," says Intel director of distribution, sales, and marketing Dennis C. Reker. He calls the timing coincidental and not intended to thwart IBM's entry into commercial chip sales. IBM's Cornell says the company knows that it must expand its portfolio to interest wholesalers in carrying IBM products--an expensive proposition.

Some rivals say the changes may be so daunting as to cause a quick retreat. Wilfred J. Corrigan, CEO at ASIC maker LSI Logic Corp. in Milpitas, Calif., says assembling sales and technical help in ASICs will be slow and costly for IBM: "I think what IBM will find out is they ought to sell computers."

To prove Corrigan and other naysayers wrong, the newcomers are moving gradually, and with partners. IBM has made Motorola its key microprocessor partner. DEC is working with Japan's Mitsubishi Electric, and Hewlett-Packard with Hitachi and Samsung. HP is limiting its commercial ASIC work to 30%--enough, it figures, to defray the cost of building chips for itself. And DEC is coming to market with a narrow line of chips, mainly its bet-the-company Alpha microprocessor and a handful of logic chips that were developed for Alpha computers. Even IBM, the most aggressive newcomer, is starting small--beginning with just two centers where customers can design chips, vs. 50 run by LSI Logic.

The newcomers' outside sales aren't likely to rival those of Intel, NEC, Toshiba, or Motorola anytime soon. But IBM and the others are determined to give the commercial market a go. The truth is, they have little choice but to try.

Before it's here, it's on the Bloomberg Terminal.