Given war and the looming possibility of a second-dip recession, it was startling to find the Optical Fiber Communication conference in Atlanta to be somewhat upbeat. Sure, attendance didn't top 20,000, some 10,000 less than in OFC's heyday three or four years ago. But many attendees said they were glad to see the feeding frenzy of financial speculation gone (let's hope for good).

Two related OFC trends, however, suggest there is a continued problem of overcapacity at the bottom of the food chain, where both photonic-component and semiconductor manufacturers reside. First, many companies had vanished from the OFC landscape, victims not of a traditional Chapter 7 or 11 filing, but of a slow parceling out of assets that left only tiny holding companies in their wake. Second, the survivors were attempting to snap up assets at bargain rates and cobble together manufacturing lines and patent portfolios in a way that made sense for some sort of integration, be it planar lightwave circuit or silicon optical bench.

Many companies had recast themselves as custom contract manufacturers, from kSaria on the former manufacturing-equipment front to Oplink Communications, previously known as a supplier of DWDM and erbium-doped fiber amp components (see March 31, page 10). Hey, the fabless semiconductor model might indeed work in the integrated optoelectronics industry, and that might make the companies concerned more agile in seeking applications outside telecom, such as medical imaging, physics and biometric security.

This doesn't relieve the pressure of small-lot orders, though. Maybe only 10 customers wanted your standard electroabsorption modulator in the past, with unit sales in the tens or hundreds. But how many of the custom sales will hit the jackpot for large-volume shipments? Many of the cobbled-together fabs might resort to high nonrecurring-engineering costs, just as in the ASIC market, to help recover the expenses of keeping fabs open. At least the specialty 4- and 5-inch compound-semiconductor fabs don't have to face the gargantuan expenses of 12-inch, 90-nanometer CMOS fabs.

Playing cut-and-paste with distressed assets and using the resulting manufacturing line like a foundry may be a decent strategy for survival in optical communications. But it only works for a time. It might take survivors into short-reach optics for enterprise and last-mile access applications. Those working on 40-Gbit ultralong-haul components had better find a different model, as the long-haul market shows little sign of recovering before the second half of the decade.

http://www.eet.com