Silicon Photonics Signals Red Alert for 100G
LOS ANGELES -- OFC/NFOEC 2012 -- Silicon photonics appear poised to make a run at the 100Gbit/s module market, but it's not certain this should have established module vendors freaking out.
The technology was a hot topic at OFC/NFOEC primarily because Cisco Systems Inc. (Nasdaq: CSCO) just acquired Lightwire Inc. for US$271 million -- a mind-boggling figure for a company with no known revenues. (See Lightwire Points Cisco Toward 100G.)
That's got some analysts wondering if silicon photonics, a technology that's been lingering in the market for years, might suddenly be on the verge of something big.
For example, components vendor Kotura Inc. might be ready to take on the optical modules makers such as Finisar Corp. (Nasdaq: FNSR) and JDSU (Nasdaq: JDSU; Toronto: JDU), according to analyst James Kisner of Jefferies & Company Inc. "Our checks suggest Kotura aims to launch a proprietary 100Gbit/s transceiver on a similar timeframe as Cisco (by July 2013)," he writes in a report published Wednesday morning.
His report focuses on the damage this could do to Finisar. He thinks the Cisco acquisition alone "puts 5 to 8 percent of Finisar's current revenue at risk" and that silicon photonics "could potentially reduce the cost of a [40Gbit/s or 100Gbit/s] transceiver by 90 percent."
So we're looking at cheap competition and the prospect that Cisco, a big module buyer, might possibly build its own stuff. Sounds like a recipe for an optical freak-out, at least among data-center module vendors (most of the silicon photonics talk relates to the data center, not the telecom network).
But not everybody is buying into the theory.
Playing with blocks
For one thing, Kotura denies that it's making a transceiver module.
"We haven't decided that that's the right place for us to be, but definitely we want to do the chips for those," said Arlon Martin, Kotura's vice president of marketing.
A better way to describe the situation is that Kotura has built an "engine" that can be the core of a 100Gbit/s module, says analyst Brad Smith of LightCounting . Kotura, which has made much of its money selling variable optical attenuators (VOAs) into multiple markets, was at OFC/NFOEC discussing its 25Gbit/s modulators and detectors, elements that could be part of a 100Gbit/s module. And providing the building blocks that others can use to build products would be more Kotura's style, Smith says.
(Kotura has technically broken into the 100Gbit/s market, because it's sold VOAs into some vendors' 100Gbit/s modules, Martin notes. But the VOA isn't a central part that makes the module "go." No one calls you a module maker just because you have VOAs.)
Kotura's readiness for 100Gbit/s shouldn't be that surprising. Luxtera Inc. , the other longtime silicon photonics startup, announced in November a 100Gbit/s chip that can be the basis for an optical transceiver module. (See Luxtera Tries Its Hand at 100G.)
The big kahuna
And then there's Cisco. Some analysts have speculated that Cisco's real interest is in using silicon photonics for optical backplanes and on-board optical connections -- improving routers with faster speeds and lower power consumption.
That's true to some extent, but officials tell Light Reading they've really got optical modules on the brain.
"Initially, expect us to deliver product in the form of optical transceivers in 40Gbit/s and 100Gbit/s, and in terms of distance, let's go beyond distances that VCSEL [vertical cavity surface emitting laser] technologies can address," says Massimo Prati, vice president of Cisco's Central Engineering Operations.
So, three companies have silicon photonics parts that could go into a 100Gbit/s module. Even so, optical module makers don't need to worry yet, says Rafik Ward, vice president of marketing for Finisar, a (cough) optical module maker.
"In the business we're in and the volumes we're in, it's not clear to me that this is an economic game-changer," Ward says.
One stumbling block is that silicon chips aren't necessarily cheap to build when you're talking about a high-end, low-volume technology such as 100Gbit/s. Ward points out that high-end chips are expensive to build, with lots of money going into the mask sets that are used in manufacturing. "The economics may or may not make sense," he says.
Analyst Smith agrees: "I don't see that this is going to impact the price so much. This is a high-end technology. This is 100Gbit/s stuff." Moreover, 10Gbit/s is going to dominate data centers for a while, he notes. Silicon photonics vendors aren't targeting that area yet.
He does think silicon photonics has a major chance to crack the data-center market, though. It's because VCSELs are struggling with high speeds. The devices can go 25 Gbit/s -- examples were dotted around the fringes of the OFC/NFOEC expo, Smith says -- but if they can't go further, or if 25Gbit/s VCSELs prove an unviable market, silicon photonics might step in.
While he doesn't buy that the Cisco/Lightwire combination will hurt Finisar that much, he does think it's going to have repercussions.
"I predict there's going to be a number of mergers in the transceiver space as a result of the Cisco thing," Smith says. They won't necessarily be silicon photonics -- more likely, the deals will be for any intellectual propoerty that helps with 100Gbit/s modules.
To see all our OFC/NFOEC coverage, go to http://www.lightreading.com/ofc-nfoec.
— Craig Matsumoto, Managing Editor, Light Reading