Optical components

More on Infinera (née Zepton)

Ultra-stealthy startup Infinera (formerly Zepton Networks) has never said much about what it's doing, but it may have already said too much.

Word on the street is that the company is about to be the subject of a puffy cover piece in a well-known Silicon Valley magazine -- though said article gives very few new details about what the company is doing.

As previously reported in Light Reading, the company has top engineering talent but has kept its technology under wraps (see Zepton Has an $86M War Chest). The only major recent public development is the company changing its name from Zepton to Infinera (hmm, did it sound too much like Zaffire?).

Infinera officials are still tight-lipped about any new details. "We're probably still a way off coming out of stealth mode," says David Welch, Infinera's founder and CTO.

But the scuttlebutt has caused Light Reading to do a bit more investigating -- and it now looks as though Infinera is targeting an add/drop multiplexer system for metropolitan networks. But more on that in a bit.

First a bit of background. Infinera got started in May last year when it scored $50 million from top-tier VCs in a round led by Kleiner Perkins Caufield & Byers (see Zepton: Take Me to Your Leaders). A few months later it took in another $36 million.

Jagdeep Singh, founder and CEO of the company, is known in optical networking circles for helping found Lightera Networks, which was sold to Ciena Corp. (Nasdaq: CIEN) in 1999 -- and has turned into a smashing success there as the basis of that company's CoreDirector product. Drew Perkins, who was previously a founder and CTO of Lightera, helped Singh in the founding of Infinera. The third founder was Dave Welch, former general manager of the research and technology development group at SDL, which was sold to JDS Uniphase Inc. (Nasdaq: JDSU; Toronto: JDU). Singh and Perkins provide the systems savvy at the startup, while Welch is the components guru.

According to several sources, Infinera is developing smaller, optically-switched equipment for the metro and edge of the network. "The edge, that's where you need more switching," says one source, who did not wish to be named. "The edge is always changing, but once you put up a link from New York to San Jose, it's pretty fixed. There's an order of magnitude difference in terms of how often you switch, so the edge is where you're going to make the most cost savings."

Smaller port-count switches up to 32x32 -- which could be created using integrated optical circuits -- would be suitable for switching at the edge, he adds.

In his view, Infinera must have some clever technology for active components such as lasers and amplifiers, with the possible addition of MEMS (micro-electro-mechanical system) technology. MEMS employs tiny tilting mirros to enable optical switching. "If they're only doing [passive components] they have no edge, because passives are very easy to copy. Active components are still a black art."

Given Welch's background, and the fact that Infinera is working in indium phosphide rather than silicon, this version of events seems very likely. Indium phosphide is the only material in which it is possible to integrate active functions like lasers and amplifiers, along with passive functions such as multiplexing and switching.

Welch would have been involved in a variety of integrated optics work while at SDL, including waveguide-based modulators and multiple lasers integrated on the same chip. Some of that work spawned another startup, Santur Corp., which is developing tunable lasers. Approximately half of Welch's team at SDL is reported to have joined Infinera; the other half went to Santur, which was actually founded first.

As for the MEMS switching, there are ways of integrating MEMS functionality onto a chip. Nanovation (RIP), which was once a poster child for integration in indium phosphide, had developed such an approach using pop-up mirrors that sat in trenches between optical waveguides (see Nanovation Goes Bust). And an Israeli startup, GalayOr Networks, is thought to have a similar technology.

Infinera's executives presented a united front in refusing to comment on any speculation. Light Reading's sources seem confident of their guesswork, but, as they point out "If we knew for sure, we'd be under NDA."

One other interesting factor about Infinera is that it has opted to build a system around its components, rather than simply sell components. This presents a much harder and riskier task. But it also provides a faster route to market for the components. Rather than having to get the product qualified by the systems vendors, who then have to get it qualified by a carrier, the vertically-integrated supplier can cut out the middle man.

Having said that, other startups that wanted a fast track for their components have hit up against problems. Luxcore Networks Inc. for one, started out developing systems and later changed its mind, because it thought it could recognize revenue much faster by just selling components (see Luxcore Pulls a Switcheroo).

Applying this logic to Infinera, it seems that the startup will likely need more than the $86 million it already has, if it's to develop a system sucessfully.

— Pauline Rigby, Senior Editor, Light Reading
pooh-bear 12/4/2012 | 10:43:38 PM
re: More on Infinera (née Zepton) I got it in the mail last night.
mess 12/4/2012 | 10:43:33 PM
re: More on Infinera (née Zepton) As always LR gives partial information and interpertation.
Nanovation MEMS technology is quite different than Galay-Or's.
While the former moves mirrors back and forth (not poping them up
like OMM or C-Speed) to block the path of light, then Galay-Or moves waveguides (Like OpsiTech from France, see the Tel-Aviv university web site for a description of Galay-Or technology).
Recently Opsitech discarded their MEMS waveguides effort for both business (no customers for 1xN switches) and technology reasosns (low yields and low reliability) which may indicate the future of Galay-Or. Nanovation is not doing so well too.
As for Zepton/Infinera, working with InP is tough, and that's what analysts and readers don't know. Those who have been there know that this is the reason why we see less success in this front. Lasers have low yields, integrating them with passive and MEMS may be a killer. Unless Infinera guys found a clever way to do it, their success probability is low.
smoking_craters 12/4/2012 | 10:43:22 PM
re: More on Infinera (née Zepton) Sheesh!

I just finished reading the Red Herring "puff piece". Argh! Does Kleiner Perkins put horses heads into people's beds, or are writers just too stupid to do any homework?

The claim made is that Infinera has created the world's "first integrated photonic circuit".

Hardly! That honor belongs to the folks at AT&T / Lucent / Bell Labs who built the first commercial photonic IC in the 1980's. And there are plenty of other folks who have built them in between. In fact, I'd have to say that I saw at least two dozen products at OFC last week that were photonic IC's.

Hey, if it's all optical and it has optical circuits on it and they're fabricated on the same integrated substrate, guess what, it's a photonic IC!

I really wish that VC's like Kleiner Perkins would turn off the hype machine.
poster 12/4/2012 | 10:43:21 PM
re: More on Infinera (née Zepton) smoking-craters: have you seen the number of smoking craters on the moon KP calls their portfolio? these guys are the kings of 'keiretsu' hype - that is the only thing they actually do. KP is the Enron of venture capital. about 95% over-hyped and 5% substantive value. the fact that red herring still buys and prints all the bullshit tells you as much about red herring. red herring is an integral part of their bullshit-factory. it's like Cisco and Packet magazine.

grateful photon 12/4/2012 | 10:42:35 PM
re: More on Infinera (née Zepton)
InP photonic circuits have been rudimentally demonstrated for a couple decades. it tough and will require much time and money to make significant progress. the monolithic devices themselves are difficult enough, but it's the process integration that will be the limiter. while it's true that in the intervening years, semiconductor process technology (driven by Si, but beneficial to III-V too) has advanced considerably, the real tough problems remain in integration. the promise of some even low-scale integration PIC will likely be met with the reality of a few functions on a chip. at the end of that exercise, it's still uncertain that the economics make sense over micro-optics or assembled monolithics.

as for red herring, showing up on the cover is nearly a curse. if you are willing to hype your company to that level, i doubt you have substance.
flanker 12/4/2012 | 10:41:42 PM
re: More on Infinera (née Zepton) Hey, I read Packet magazine...

Anyway, Om Malik knows as much about telecom as Bobby Knight knows about self restraint. The whole Red Herring article is poorly researched if not slightly entertaining. Om should stick to personality profiles, because he'a a great ass-kisser.

His knowledge concerning the direction of the industry or how the various pieces (technology, vendors, SPs) fit together is rather light.

And you're right - Lepton/Incinera didn't invent the first ASOC, or comercially deploy the first ASOC.

It's all moot anyway, because the market for this product is exactly zero. Maybe three years ago Lucent or Nortel would have bought this company. Not today. They wont even but the components; they have their own engineers to feed.

Sign In