Ever wonder what happened to Oplink Communications Inc. (Nasdaq: OPLK)?

The bubble-era IPO star vanished from most radars after seeing a merger with Avanex Corp. (Nasdaq: AVNX) crumble in 2002 (see Avanex and Oplink: Wedding's Off). Oplink's answer has been to escape into a new sector: manufacturing.

Companies such as Taiwan Semiconductor Manufacturing Company (TSMC) (NYSE: TSM) have built multibillion-dollar industries from building other companies' chips, selling the manufacturing service but not the chips themselves. Though it hasn't quite attained that level, Oplink has aspirations to be the optical equivalent, a foundry for optical subsystems.

Half of Oplink's revenue still comes from component sales, but psychologically, the shift is on. "We are not a product company any more," CEO Joseph Liu says.

Slowly, it's beginning to work. Oplink has moved all its manufacturing to a 500,000-square-foot facility in Zhuhai, China, which housed 598 of the company's 676 employees as of last June. Among its customers are systems houses such as Ciena Corp. (Nasdaq: CIEN) and Nortel Networks Corp. (NYSE/Toronto: NT), as well as fellow subsystems vendors Avanex, Bookham Technology plc (Nasdaq: BKHM; London: BHM), and JDS Uniphase Corp. (Nasdaq: JDSU; Toronto: JDU).

Wall Street hasn't taken much notice. True, Oplink stock has nearly tripled from its $1 level a year ago, but that wasn't unusual among the other gains of 2003 (see 2003 Top Ten: Explosions & Implosions). Oplink's market capitalization sits at around $400 million, compared with around $670 million for Avanex and $6 billion for JDS.

"I don't hear a whole lot from the analysts," Liu says. "Until the revenue shoots up, people are probably suspicious what Joe is talking about. All this strategy is still waiting to get proven by revenue growth."

Oplink's revenues are rebounding slowly but are nowhere near 2001 levels. "We have a business model and some preliminary success, but we have not seen it turn into dramatic, 50 percent quarter-over-quarter growth," Liu says.

Table 1: Oplink Annual Revenues

Doubts are understandable, as it's unclear how well Oplink -- or any other photonics play -- can sustain itself on manufacturing alone, and the answer won't manifest itself until the industry stabilizes. Oplink is off to a reasonable start, at least.

The problem is that photonics isn't quite like semiconductors. TSMC receives orders that can run in the hundreds of thousands of units, compared with the 200-piece orders found in photonics circles.

Moreover, optical components are an analog technology, inherently subject to tweaks and changes. Liu wants to be where analog-chip specialists Linear Technology Corp. and Maxim Integrated Products Inc. (Nasdaq: MXIM) are, pounding out relatively small quantities of high-end devices, collecting high margins along the way (the difference being that Linear and Maxim sell those parts themselves).

Oplink claims to have a cost advantage by having all its manufacturing in China, including the final steps of assembling subsystems. Other companies kept this "back-end" process in North America, because offshore countries lacked the expertise for it, Liu says.

The company got its first China foothold in 1999 but really took the plunge in 2002, after the Avanex fiasco. Then-chairman Liu took the CEO helm from Fred Fromm (Liu wouldn't discuss the handoff in detail) and began chopping down Oplink to fit the stark realities of the components market.

"I was drafted by the board to come up with a new business model for the company, because coming out of a failed merger hurts a company quite a bit. It hurts the board's and management's reputation to have recommended the merger."

Among Liu's decisions was to cut all U.S. manufacturing, hiring new teams in Zhuhai. The idea of becoming a foundry came up as Oplink was getting manufacturing jobs that the electronics manufacturing service (EMS) companies couldn't handle.

Those companies, whose ranks include Celestica Inc. (NYSE, Toronto: CLS), Flextronics Corp. (Nasdaq: FLEX), and Solectron Corp. (NYSE: SLR), jumped into photonics manufacturing during the bubble. Their efforts have toned down -- Solectron consolidated its optical groups into its Charlotte, N.C., facility, for example -- but all three are still taking photonics work. Flextronics' involvement might even increase, should it close a proposed deal to take over most of Nortel's in-house manufacturing (see Nortel Talks Outsourcing).

But Liu claims they aren't set up to handle the analog tweaks of the optics world. "The EMS guys are trained that when the design is complete, bang! you cannot change a thing."

JDSU, Avanex, and Bookham offer foundry services as well, but Oplink hopes to stand out by not concentrating on component sales, thus avoiding competition and questions over intellectual property.

Oplink does plan to keep its component product lines, but they'll be used to fill the gaps in customers' subsystems designs, ideally by inserting higher-performance parts than the customer could provide.

"Just like TSMC, I have to produce a lot of technology to differentiate myself," Liu says. TSMC does that by working closely with companies such as Altera Corp. (Nasdaq: ALTR) and Xilinx Inc. (Nasdaq: XLNX), using their highest-end designs as proving grounds for advanced manufacturing methods. Oplink plans to gather its technology through acquisitions such as Accumux Technologies Inc. and RedClover Networks Inc., and Liu says he's scouting for more candidates. (See Oplink Nabs Accumux and Oplink Plucks RedClover.)

Oplink continues product R&D, aimed at integration and smaller subsystem sizes rather than breakthrough speeds. Topping Liu's list is the integration of active and passive devices, an area where he hopes Oplink can make some advances.

— Craig Matsumoto, Senior Editor, Light Reading