Global Enterprise Management

Advice for high-tech companies:
Make better use of technologies you developed

By Bruce Goldman

Not that long ago in Silicon Valley, the name of the game was technology. If you had the new technology first, you won. But things are more complicated these days. That little garage startup is now a multibillion-dollar global enterprise building factories all over the world at $2 billion apiece.

As they have matured, info-tech companies' destinies have become increasingly intertwined. The new box that company X has just introduced may have hot new technology and may work just fine, but if it's not compatible with an off-brand component in a customer's system, the customer will be calling for answers. What can the customer service representatives at company X do?

At least part of the answer is that even high-tech companies need to make better use of the very technology that their products make possible. In order to coordinate their widely distributed networks, they need to develop "bullet-proof" information systems that use the Internet to integrate all their manufacturing sites, including the virtual ones.

That was one of the conclusions reached at a workshop, titled "Global Enterprise Management for the Internet Age," which took place on campus this spring. Its general theme was that information integration must extend beyond manufacturing to include other basic functions, such as design, investment and marketing. The gathering provided a forum for 60-plus representatives from high-tech companies to discuss the travails of being a global concern in a world where not only technologies but competition, economic cycles and business models seem to change at warp speed.

The meeting was sponsored by the Alliance for Innovative Manufacturing at Stanford (formerly known as the Stanford Integrated Manufacturing Association), a joint venture dedicated to exchanges of ideas and techniques between academia and industry. Chief organizer of the event was Ramakrishna "Ram" Akella, a visiting associate professor in Stanford's Engineering-Economic Systems and Operations Research (EES/OR) Department.

Akella and a number of high-tech manufacturers in industrial sectors ranging from semiconductor-manufacturing equipment to computer hardware, software and networking have been collaborating on attempts to coordinate all aspects of the business enterprise: design, investment and financing, marketing, supply-chain management, quality control and so forth. These efforts have yielded successful new models for increasing profitability.

Ironically, while most traditional companies have increasingly shifted their capital spending from bricks-and-mortar facility construction toward assembling state-of-the-art information systems, semiconductor outfits have been forced to spend a growing amount of their budgets on ever more complex fabrication facilities, or fabs, where the integrated circuits that serve as the basic building blocks of the information age are created. As process technology has grown trickier, the cost of a single fab has passed the $2 billion mark.

The semiconductor manufacturers have little choice. Technology upgrades keep coming on stream, and customers want the best. Oh, and by the way, they'd like it delivered yesterday.

As Chris Bajorek, senior vice president and chief technology officer of Komag Inc. of San Jose, which makes magnetic storage disks for computer hard drives, told the group, "We sell to practically every drive maker in the world. Our customers are increasingly sophisticated. They want four-hour delivery time. They want to be able to cancel or to hike their order by 30 percent overnight. And they want us to give them a Lexus for Saturn prices. Not so long ago, they paid $3 per megabyte of storage. Now it's two cents, and in a year it'll be one cent. Storage density doubles every year."

Meanwhile, product cycles are growing shorter. If there's any significant delay in getting your new product out, you may lose half your hoped-for sales.

To make things even stickier, companies, while producing their high-tech items in facilities all over the world, must nevertheless guarantee uniform quality. Their customers don't want to worry whether the chip they're using was made in Israel or Ireland or Idaho.

Komag and Intel, the giant Santa Clara microprocessor maker, each own multiple fabs. Both use a method for fab construction known as "copy exactly": Once a single facility has perfected the process for producing the newest generation chips, each new facility is built to be as exact a copy of the original as possible. To do otherwise with such a complex operation, both Bajorek and Intel fab manager Kirk Hasserjian agreed, increases the risk of catastrophic shutdowns when parts of the process don't mesh perfectly.

The "copy exactly" approach does seem to solve the quality-control issue and make speedy ramp-ups of new products more of a sure thing, the semiconductor representatives said. But it creates another problem: Engineers hate copying exactly. "Engineers by nature don't fall into this mode easily," said Intel Fellow Gene Meieran. "They're bright; they're innovative; they want to make an impact. When you tell them, 'No, you're not gonna do that,' it's like you're taking away their livelihood."

There's also a geographical issue. "To be the lowest-cost producer, we need low-cost factories," said Bajorek. Komag has four major fabs in Malaysia, two in Northern California, one in Japan and one in Thailand; about 75 percent of its volume production is in Malaysia, where the cost of labor -- including engineers and white-collar workers -- is one-seventh that in the United States. "But our R&D prototype and pilot production is in the U.S. How do you do 24-hour-a-day R&D with constant hand-offs across vast geographical regions? You need bullet-proof information technology," Bajorek said.

Then again, are you really sure you even need to build a factory? "Four years ago, demand looked insatiable," said Bajorek. "Everybody built fabs. By the time they were finished, business was down. A couple of local companies closed their doors." What's changed in those four years? "We've gone from a three-year to a nine-month product cycle."

This is where outsourcing starts to sound like a good idea. "It doesn't make sense to own too much capacity if your industry has wildly fluctuating demand," said Tom Zeuss, director of marketing at Oracle, a Redwood City database software supplier.

To avoid the consequences of huge, costly demand miscalculations, companies have taken to contracting out production while sticking to their core competencies of design and marketing. For every Cisco employee, there are six "virtual" Cisco employees, according to Todd Tuomala, director of manufacturing operations for the San Jose-based company, which makes Internet equipment. Cisco uses five main contracting manufacturers, and 14 global manufacturing sites, for the printed-circuit boards it uses in some of its products.

Outsourcing has created a whole new beast, the so-called "foundry": a contract manufacturer that makes the integrated circuits that other companies design. TSMC (an acronym for Taiwan Semiconductor Manufacturing Corp.) is one such foundry. Established in 1987, TSMC now controls 35 percent of global foundry work, and its market capitalization is the third largest of any semiconductor company in the world, behind Intel and Texas Instruments.

"Today we can offer state-of-the-art technology and, we hope, state-of-the-art service," said Magnus Ryde, president of TSMC America. A foundry's ability to produce huge volumes of an advanced, complex product tailored to exacting specs enables a small design house to act like a big manufacturer -- and that's just what's happening. The "fabless semiconductor" segment (i.e., companies that design and perhaps market, but do not produce, chips) now accounts for a good 10 percent of global revenues, Ryde said. This segment is growing nearly twice as fast as the general industry -- more than 20 percent annually versus about 13 percent.

Ryde predicted that, in addition to the semiconductor makers now embracing the foundry model, "the systems integrators, such as computer makers, also will start doing some design themselves and coming directly to foundries with them," bypassing the big-shoulders, brand-name semiconductor giants entirely.

That gives Stanford EES/OR Assistant Professor Blake Johnson pause. "Are we creating monsters?" he asked. "Sure, it makes sense to set these foundries up, but does it eventually create economies of scale that allow a TSMC to dominate the industry? What if the contract manufacturers start putting their own label on their white boxes? Will their clients eventually become mere marketing extensions? That's unclear." Once upon a time, Johnson reminded the audience, "it made sense for IBM to contract the development of its operating system to Microsoft."

The fabrication foundry provides a model for other sectors as well. For example, the food service industry, at $350 billion a year in revenue, is the third-largest retail industry, after grocery stores and automobiles. According to Ted Daly, vice president of Instill Corp. of Palo Alto, which occupies a business-to-business e-commerce niche for the restaurant business, "Chains like Pizza Hut and Marriott are realizing they can negotiate better deals with manufacturing giants like Kraft or Nestle than they can get through distributors. That's where Instill comes in."

Food is 35 percent of an operator's total costs, Daly said, and typical net margins for food service operators run about 5 percent. "So, if we can lower costs by $1 million, that's equivalent to a sales increase of $20 million."

One way to do that, he said, is to rationalize what is now a chaotic ordering process. "Instead of purchasing done by individual unit operators, we offer a centralized, web-based ordering system. Chains can deal with multiple vendors via a single interface, which makes for smoother back-office integration and significant cost benefits."

In a supermarket, every product has a UPC bar code for easy identification, said Daly. "With food service, there's no such animal. The same product from 20 different distributors will have 20 different numbers, so a buyer doesn't know what he's bought, or how much." Instill is addressing that problem by building a common product code for the industry, Daly said: "Now you can ask, 'Where did I buy this from? How much did they charge us? Who charged too much? Are there any discrepancies vis-à-vis my contract?' You can apply for the rebates you've earned, and see new opportunities to negotiate with suppliers."

All this contracting produces interdependence and, with it, the need for souped-up information technology. Information systems that were fine back when suppliers sent stuff to the factory, where it was assembled and sent to distributors, are no longer adequate. Now that suppliers ship parts directly to distributors for assembly and faster delivery to customers, it is much more difficult to keep track of everything.

"The ultimate core competency is not the product but the design of your supply chain," stated Oracle's Zeuss.

Intel's Meieran spoke of the uneasy but necessary balance between the tiny, creative, exuberant and often short-lived Internet-based companies, which he called "fruit flies," that are springing up to occupy the business-to-business information niche, and the slower, more bureaucratized behemoths he called "dinosaurs," which are staid, risk-averse and politicized but whose more measured pace supplies much-needed standards to a chaotic marketplace.

Like other creatures with short lifespans, fruit flies tend to proliferate rapidly when the climate is favorable. "The number of public 'dot-com' companies has increased 72 percent per year, compounded, and their market value has increased 365 percent per year. In 1999, they will lose about $2 billion," said Bob Neely, vice president for supply-chain management at A. T. Kerney Inc., an Alexandria, Va., consulting firm.

Associate Professor Samuel S. Chiu of EES/OR offered a plausible formula: "The price of an Internet startup's stock is proportional to the amount of money it's losing."

COMMENTS? Contact Richard Reis, Executive Director AIM (650) 725-0919

email: reis@cdr.stanford.edu