IMMI Switches Strategy
What's left of the all-optical switching market is littered with the broken shells of startups. Many of them never got past the design stage, but a few shipped product and even experienced a near miss with success.
Count Integrated Micromachines Inc. (IMMI) among them. Now calling itself "IMI," the company killed its all-optical plans a year ago, hoping for rebirth in the semiconductor test market. But company officials say they once landed a Cisco Systems Inc. (Nasdaq: CSCO) deal that could have revived their fortunes. From the sound of it, the deal might even have kicked some life into the all-optical business.
IMI sings a different tune today. "I've been out of it for a year, so now I can look back and say electronic switching kicks butt," says Raffi Garabedian, IMI's CEO.
The all-optical market still has its share of hopefuls, including Calient Networks Inc., Corvis Corp. (Nasdaq: CORV), Lynx Photonic Networks, and Polatis Ltd., but it's become better known for its collapses (see Calient Bags $20M More, All-Optical Still Kicking, OMM: The End Is Near, Onix: Another MEMS Casualty, and Network Photonics Shuts Down).
All-optical mania was triggered in 2000 by Xros Inc., a startup building a 1,000-port crossconnect that used microelectromechanical systems (MEMS) mirrors to route wavelengths. The goal was to avoid converting traffic from optical form to electrical, allowing carriers to bounce wavelengths from node to node -- the "only" way to handle the traffic tsunami that surely was coming. Xros had barely announced its presence when Nortel Networks Corp. (NYSE/Toronto: NT) acquired the startup for more than $3 billion in stock. IMMI sprang upon the scene just months later (see Xros Launches First 1000-Port All Optical Cross Connect, Nortel Spells Out Its Cross-Connect Strategy, and Switch Startup Raises MEMS Questions).
By the time Nortel killed off Xros in 2002, the market had flooded with competitors, most of them bearing MEMS-based ideas (see Nortel Shuts Optical Switch Effort and MEMS Make It Big?). With 1,000-port mania relegated to the joke file, a more pragmatic model emerged: Certain traffic would be packed into express lanes, wavelengths that would zip through nodes via all-optical switching. But the majority of traffic would get terminated at each node and converted into electrical form for scrutiny by routers and switches.
According to former IMI chief executive Denny Miu, Cisco was pursuing this model late in 2001 with a single box called the 5600. Garabedian says the design used IMI's 80-port crossconnect, which would be announced at the 2002 Optical Fiber Communication Conference (OFC), but also packed an electronic switch fabric with STS-1 (51-Mbit/s) granularity. Unlike previous all-optical projects targeting the core, the 5600 aimed for metro networks, a higher-volume play (see IMMI Intros Crossconnect).
Who knows, the 5600 might have changed the image of all-optical switching. IMI did its part by shipping product to Cisco, but the project was killed. "They liked what we were doing, but their customers said people weren't ready to buy wavelength services anytime soon," Miu says. (Cisco officials declined to comment on this story.)
IMI had already had a rough road. Miu says he oversaw four layoffs before stepping down as CEO in mid 2002. With the Cisco project dead, IMI gave up, as Garabedian reasoned that all-optical switching would find cheaper or simpler methods by the time the market recovered.
"There wasn't enough potential for the product that would justify the way we were spending money," Miu says.
IMI could have stuck to building 1x2 and 2x2 devices, frequently used for protection switching, but decided it wasn't worth the effort. "I wish I could be interested in it, but the market is so competitive, and there are so many technologies out there, that I don't see it as a profitable business for a startup," Garabedian says.
Late last year, IMI returned to the probe-card market for semiconductor testing -- ironically enough, the very market it had abandoned in 2000 to pursue all-optical switching. The new target has the advantage of, well, existence, with $604 million in 2004 probe-card sales forecasted by VLSI Research. Even here, IMI is battling uphill by offering a higher-end alternative to established technologies.
The company has at least one customer hoping to ship a probe card in 2004, and Garabedian is scoping the VC circuit for $10 million to get to volume production. As for Miu, he's still in the crossconnect game. He did some consulting with startup Glimmerglass Networks and funneled IMI's former customers their way. His latest gig is a stealth startup doing yet another crossconnect -- but this one is both electronic and enterprise-minded.
— Craig Matsumoto, Senior Editor, Light Reading