Vendors are saying the all-optical switch is making a small comeback, although it's arriving too late to save Corning Inc.'s (NYSE: GLW) IntelliSense division, which began closing down this week.

In fact, optical-switch fans will tell you "comeback" is too strong a word for what's going on. "I don't know that it ever disappeared, really," says Marlene Bourne, analyst with In-Stat/MDR.

What has disappeared is the dream of massive, 1,000-port crossconnects. In its place is a more realistic plan, one that began emerging just after the peak of the bubble: a small all-optical switch used for express-lane traffic, accompanied by a traditional OEO (optical-electrical-optical) switch for channels that need to be terminated.

"There is a lot of discussion between customers and their suppliers," although most of that hasn't translated into huge revenues yet, Bourne says. And it hasn't been enough to save the optical-switch firms that are shutting down (see OMM: The End Is Near and Onix: Another MEMS Casualty). But it could be good news for those still standing, including Advanced Optical MEMS Inc. (AOMEMS), Calient Networks Inc., Continuum Photonics Inc., Lynx Photonic Networks, Movaz Networks Inc., Optical Switch Corp., and Polatis Ltd.

The hype for massive port counts hit after Nortel Networks Corp. (NYSE/Toronto: NT) made a $3 billion offer for startup Xros Inc. (see Nortel Spells Out Its Cross-Connect Strategy). "I remember talking to a particular customer who demanded [that we have] a road map to go to 2,000 ports," says Jas Sandhu, business development manager for Polatis. "I remember thinking, 'Is that where the game is?' "

Smaller numbers characterize the market now. Calient announced a subsystem version of its switch earlier this week, available in sizes from 32x32 to 256x256. Yesterday, Polatis announced a 4x4 switch -- a smaller version of its 16x16 product -- to be shown at next week's NFOEC (see Calient Shuffles to Subsystems and Polatis Shrinks Switches).

Calient CEO Charles Corbalis says the time is right for his company's subsystem, because equipment makers are looking again at technologies like optical switching that hit the back burner when the telecom boom went bust.

Corbalis seems aware that gear makers aren't going to bite on optical switches immediately. It could take 12 to 24 months for them to explore options and start deploying optical switching in networking equipment. In the meantime, optical switches can be used for several other applications, such as grid computing or lab testing.

For the latter, the technology offers a ready-made optical network for putting telecom devices through their paces. Without a switch, optical fiber must be manually reconfigured. Developers can save one-half to two-thirds on testing time with an optical switch, Corbalis says.

Polatis likewise sees the test market as a viable customer, one that's helped the company survive while laying low for the past year.

Polatis announced its 64x64 switch at the 2001 ECOC but had to pull back as the market collapsed. Company officials trimmed a Bragg grating product line and got busy raising money and hiring new executives, while refocusing development on a 16x16 switch(see Polatis Reveals Switch Secrets, Polatis Refocuses, Polatis Pulls Down $8.4M, and Polatis Appoints New CEO).

Polatis uses piezoelectric actuators to steer beams directly to the output fiber, making it one of the few non-MEMS (micro-electro-mechanical system) plays in optical switching. But MEMS still represents most of the efforts, and it's getting more vendors as Asian firms jump on board. LG Electronics Inc., Santec Corp., and SK Opto-Electronics Inc. (SKOE) are developing MEMS parts for optical networking; and Hitachi Ltd. (NYSE: HIT; Paris: PHA) has considered the idea as well, according to In-Stat's Bourne.

"These folks are being really stealthy and quiet about it," she says. "It's going to be interesting to see what they bring to the market."


IntelliSense Closes
Even if MEMS switching makes a comeback, it's not reason enough for Corning to hang onto IntelliSense, a provider of MEMS design software and foundry services, which was acquired in 2000 (see Big Vendors Acquire MEMS Makers).

IntelliSense's roughly 70 employees were notified early this week about the closing. Corning will continue to support IntelliSense's software "until it can be placed within a longer-term situation," a Corning spokeswoman says. Corning is also seeking buyers for some of IntelliSense's technology.

An IntelliSense closing has been rumored for a long time, as Corning has divested its photonics businesses (see Avanex Deal Reshapes Sector and Corning Chops Wavelength Blocker).

"Corning needs to focus on their core technologies, and if they're at a point where they need to cut costs, it makes sense -- kind of like JDS cutting off Cronos," Bourne says (see Memscap to Buy JDSU's Cronos).

Corning bought IntelliSense at a time when MEMS manufacturing capacity was scarce, but there's now an oversupply. And IntelliSense's design software is rivaled by products from Coventor Inc. and Memscap S.A. (Euronext: MEMS). So, the greater MEMS industry wouldn't be harmed if IntelliSense disappeared. "The pain would come more in the shock of a long-time respected company going under," Bourne says.

She does expect Corning to find a buyer for at least part of IntelliSense, though, considering the growing interest in MEMS in areas outside of telecom.

— Craig Matsumoto and Mary Jander, Senior Editors, Light Reading