Pulling the plug

-September 01, 1998

By Robert Ristelhueber
photograph by Larry Goldstein

In the summer of 1996, James Diller, CEO of Sierra Semiconductor Corp., brought a stunning proposal to his board of directors: scrap the company's traditional business lines, uproot its Silicon Valley headquarters and move everything to British Columbia in Canada.

It would have been a bold plan for any CEO to suggest, but in Diller's case it was particularly wrenching. He had founded Sierra Semiconductor in San Jose, had nurtured the product strategies and hired the people to manage them. Now, he was about to sweep away much of the work of the past dozen years and embark on a new course.

President and CEO: Robert L. Bailey
No. of employees: 430
Headquarters: Burnaby, British Columbia
  Fiscal year ended
  Six months ended
June 28, 1998*
Dec. 31, 1997 Dec. 31, 1996**
Net Revenue: $74 million $127 million $188 million
Net Income: -$31 million $34 million -$48 million
ROA: -3.4% 2.3% -3.7%
ROE: -1.9% 4.9% -7.4%

*Results for the first six months of 1998 include a charge of $50.8 million related to the acquisition of Integrated Telecom Technology Inc.
**Results for 1996 include a restructuring charge of $69.4 million related to the company's shutdown of modem and other non-networking product lines, plus a $7.8-million charge related to the acquisition of the assets of Bipolar Integrated Technology.


"I was faced with a dilemma," recalls Diller, 63, from his Silicon Valley home. "I'm a great believer in shareholder value, but the value of one side of our business wasn't being appreciated." The company's stock price was languishing that summer because investors identified Sierra with the cutthroat PC market, while overlooking its flourishing networking chip business.

"I was struggling with what to do, but I became convinced that we had to shut down the business that had to do with the personal computer," which accounted for well over half of the company's revenue, he says. That conclusion would lead to Sierra's laying off over 300 workers and moving its headquarters from the epicenter of technology, Silicon Valley, to a suburb of Vancouver, where the networking chip operation resided.

Although it might have seemed like a giant gamble at the time, Diller's decision proved to be a home run. The company, now called PMC-Sierra Inc., has grown rapidly and become a major force in the burgeoning networking chip market. (See "PMC-Sierra at-a-glance," p.78.) In May, the firm won the prestigious 1998 Kachina award from Scottsdale, AZ-based market researcher Cahners In-Stat Group as the best-managed semiconductor company operating without a wafer fab. (See "All dolled up," p.78.)

The company's unusual odyssey began routinely enough in 1984 when Diller, a veteran of Silicon Valley firms such as Fairchild and National Semiconductor, launched Sierra Semiconductor primarily as a maker of custom analog chips. Proving his willingness to change course, Diller a few years later scrapped that strategy when "we came to the conclusion that we couldn't have sustainable profits unless we moved more to standard products," he recalls. Custom chips required too heavy an engineering commitment and presented too great a risk that a particular customer would not be successful in the market.

The new direction was toward the PC, which was fast becoming a huge new market for chip makers. As desktop computers evolved beyond text applications, Sierra began developing circuits to process color graphics and sound. It also created modem chips used to help PCs communicate, at one point becoming a leading supplier of 2400 baud modems. Apple Computer Corp., Cupertino, CA, became a major customer, eventually representing as much as 20% of Sierra's revenue.

Diller also switched gears by becoming an early convert to the fabless model. In the late '80s, he dumped Sierra's wafer fab and transferred its process technology to Chartered Semiconductor Manufacturing Inc., a start-up wafer foundry in Singapore. Diller became a member of Chartered's original board. "Having a fab was too expensive for a company our size. It did not have the economies of scale," he says.

Although Sierra had successfully managed its transition to PC products, Diller began looking for ways to diversify its product line to be less dependent on a single end market. By coincidence, he was soon approached by a small chip design group from British Columbia that was looking for an investor.

PMC meets Sierra

Pacific Microelectronic Center (PMC) had been created in 1984 as a chip design unit by MPR Teltech, the research arm for BCTel, British Columbia's phone company. "I was talked into leaving Mitel in 1986 to start this new group," recalls Greg Aasen, currently chief operating officer for PMC-Sierra. His mission at PMC was to create circuits for switching and multiplexing gear built by Microtel, the manufacturing division of BCTel.

The unit "struggled for many years" to gain critical mass, says Aasen. The situation wasn't helped in the late 1980s when BCTel sold parts of Microtec to Northern Telecom, forcing PMC to search for outside revenue. "We were just a small group barely paying the salaries from contracts."

A turning point came in 1991, when Aasen and associates attended a European trade show, Telecom '91. "I came back from Geneva blown away by this technology I hadn't heard about," he says. That technology was asynchronous transfer mode, or ATM, a high-speed switching scheme for local area networks and wide area networks. After returning to Vancouver, Aasen discovered that one of his designers had already learned about the technology and incorporated an ATM mode on a chip used in high-speed fiber-optic networks. It gave PMC an early start in what would become a lucrative market.

Aasen received approval from BCTel management to seek outside financing to spin out PMC as a separate company. Within months, friends of Aasen's working at Sierra had contacted Diller, and by mid-1992 Aasen had talked him into investing $5 million for a 60% stake in the start-up.

"I felt it was potentially a great way to diversify outside of the PC platform and into telecom," says Diller. "The engineering team there was just outstanding. They had a fairly broad product line even in the beginning."

PMC's engineers primarily had experience at systems companies like Northern Telecom, AT&T and Mitel, giving them good insight into the needs of end customers. Using the technology inherited from its days at MPR, the group created chips for communications technologies such as SONET (synchronous optical network) and T1. Revenue in 1992 was about $3.5 million, and the subsidiary, now called PMC-Sierra, had just 28 employees.

Aasen credits Sierra's management team with transforming PMC from a design group into a real business. Diller visited the outfit's Burnaby, BC, headquarters monthly and Sierra board member Charles Bass, the co-founder of Ungermann-Bass (no longer in existence), also joined PMC's board to impart his marketing knowledge. The start-up also was able to use Sierra's small European sales force, and Diller introduced them to the same wafer foundries used by Sierra: Chartered and Taiwan Semiconductor Manufacturing Corp. (TSMC).

All dolled up

Cahners In-Stat Group bestows its Kachina Award each May on the two best managed semiconductor companies, operating with or without a wafer fab. PMC-Sierra snagged the award this year in the fabless category, based on a composite of 10 indices including: net income as a percent of sales, cost of sales as a percent of sales, R&D as a percent of sales, return on investment, inventory turnover, sales per employee, long-term debt to equity ratio, net operating profits as a percent of sales, quick ratio (current assets excluding stock, divided by current liabilities) and total equity to total assets. Here are this year's Top 10 rankings and their respective scores (the lower the score the higher the ranking).
1. PMC-Sierra Inc. 58
2. Adaptec Inc. 64
3. C-Cube Microsystems Inc. 65
4. Lattice Semiconductor 68
5. NeoMagic Corp. 76
6. Xilinx Inc. 78
7. Altera Corp. 83
8. SEEQ Technology Inc. 84
9. OPTI Inc. 91
10. Micro Linear Corp. 92

Diller recruited Robert Bailey in 1993 to be president of the subsidiary. "I was running the ASIC division at AT&T Microelectronics at the time," Bailey recalls. "Conventional wisdom said it was a high-risk venture. There were really no start-up companies focused on [the ATM chip] market that I knew of, and there had never been any semiconductor company in British Columbia."

It took quite a bit of lobbying to persuade Bailey, recalls Aasen. "He had to be sold. I spent a lot of time on the phone with him, and took him on tours of Vancouver to show him it isn't such a bad place."

After several months of cajoling, Bailey began to see the potential of the venture. The Internet was just beginning to take off, and equipment vendors would soon need to upgrade their systems to provide more bandwidth for the Internet as well as for emerging high-speed data networks.

Established equipment vendors such as Alcatel, Northern Telecom and AT&T had internal design teams creating the ASICs needed for their hardware, but a new breed of networking equipment vendors without those resources--such as Cisco Systems Inc., San Jose, and 3Com Corp., Santa Clara--could provide a ready market for PMC's standard products. "That's where they made a huge contribution, in solving the problem of circuit design talent," says Brent Little, director of strategic marketing for Applied Micro Circuits Corp., San Diego, which makes chips that both compete with and complement those made by PMC.

The breakthrough product came in 1993 with the S/UNI, which addressed the physical layer of ATM. The following year a new version of the chip called S/UNI-Lite caught the attention of Cisco, which used it in an interface card for its high-end 7000 Series router. Today, Cisco is PMC-Sierra's biggest customer.

"It was the first chip of its kind available, and highly integrated in terms of having all the SONET functionality, which was something datacom people in general didn't understand then," says Graeme Fraser, vice president and general manager of Cisco's optical internetworking business unit in San Jose. "Having an off-the-shelf product available was important. PMC made a gutsy move because the specs on ATM were still evolving. They built it on faith."

The success of S/UNI enabled chip sales to more than double in 1994, to nearly $20 million. Sierra quickly exercised its option to buy the remaining 40% of PMC-Sierra for $145 million. Sierra's original $5-million investment was now worth $218 million.

Sierra takes a dive

But back in San Jose, things weren't looking so rosy in Sierra's other product lines. "The PC platform was becoming fiercely competitive," says Diller. A stampede of companies was pursuing the graphics chip market, making a shakeout likely. On top of that, "modems had evolved into a commodity, where we were selling silicon by the square inch." It didn't help that Apple Computer, still a major customer, was in serious trouble and pulling back orders.

"The modem business was still slightly profitable in '96, but we could see storm clouds on the horizon," recalls Bailey. "The number of engineers Rockwell and Lucent were putting on this were more engineers than we had in the entire company. It would have taken a lot of additional investment on our part just to tread water."

PMC's success in networking chips was being overshadowed by the trench warfare in the PC products, dragging Sierra's stock price down. "It became obvious to me that the only way to get value for the company and the shareholders was to completely separate the two lines," says Diller. "And the only way to do that was to shut down the PC business."

"It was a pretty radical move," says Bailey, who was present when Diller brought the proposal to the board. "It would require a massive restructuring of the company." But the board accepted Diller's reasoning and quickly embraced the plan, says Bailey. "It was total support. The board was very excited about PMC and very concerned about the modem and graphics business. It was almost a relief Jim decided to do this."

The decision was not without an emotional cost. "It was more difficult for Jim because he had founded Sierra," Bailey adds. "I have to give him tremendous credit for doing the right thing even though there was an emotional element for him."

Diller says he wrestled with the decision for several months before going to the board, and insisted on giving employees months of notice and generous severance deals to ease the pain. "It was difficult, but there's no question the company's stronger today because of it. If we had tried to fight it out, it would have dragged down the networking business. You've got to be sensitive to employees and the community, but I believe strongly in shareholder value."

The company announced its decision in August 1996, taking a $69-million restructuring charge to cover severance costs and an inventory write-down. A red-hot job market in Silicon Valley meant that the 300 terminated employees generally had their pick of new jobs there, says Bailey.

In the end, only a single worker, a product manager, relocated to Canada. "It's hard to get people to leave Silicon Valley," Bailey notes.

The following June, Sierra Semiconductor formally adopted the PMC-Sierra name, with its headquarters in Burnaby. Bailey became president and CEO, replacing Diller, who stayed on as chairman. "I could've moved up to Vancouver, but once we completed the transition we had someone eminently qualified to run things, and that's Bob," says Diller.

PMC's core market, ATM chips, is expected to hit three million units next year, a tripling since 1996, according to Cahners In-Stat Group. To tighten its grip on that market, PMC-Sierra purchased chip maker Integrated Telecom Technology Inc., San Jose, for $55 million in cash and stock earlier this year. In late 1996, it acquired Bipolar Integrated Technology Inc., Beaverton, OR, for about $10 million to enter the Ethernet chip business.

These acquisitions, in addition to a hiring spree, have boosted employment to around 430, about the level prior to the restructuring. The company enjoyed its most profitable year ever in 1997, and now has a list of blue chip customers that includes Cisco, 3Com, Bay Networks, Alcatel, Newbridge and Nortel. Its stock, which hit $7 in the summer of 1996, was selling for $36 in mid-August of this year.

PMC competes with a broad range of suppliers, from giants such as Lucent Technologies, Hitachi and NEC, to start-ups such as TranSwitch, Galileo and Xaqti. But the company has managed to win raves from customers such as Dennis Rainville, vice president of hardware engineering at Ascend Communications in Westford, MA.

"When we talk with other vendors about establishing partnership relationships, we always use PMC as an example of what we'd like that partnership to be," he says. "They're very good in presenting their roadmap. They're very responsive, right up to Bob Bailey doing periodic treks here to listen and ask us what they can do better."

That kind of relationship allows PMC to get an early look at where hardware is heading, says Elias Moosa, a research analyst with BancAmerica Robertson Stephens in San Francisco. "They have tremendous mindshare and credibility with customers, and their real strength is the depth of knowledge of the broadband communications market. That really creates a huge barrier to entry compared to competitors."

Moosa doubts the main threat to PMC would come from outside the company: "The biggest thing they have to worry about is complacency and overconfidence." He's not betting on that, though, as he's currently recommending the stock. "They've done really well, and they're set to do better."

Loading comments...

Write a Comment

To comment please Log In