Finisar Meets Luxtera at Gilder Telecosm

The Gilder Telecosm 2006 conference was punctuated by sponsor company presentations. One session had the CEOs of Luxtera and Finisar (FNSR) presenting, followed by moderator and audience Q&A. I was really looking forward to this discussion given the work I’ve done on SFPs and Cisco.

Conference Title: On the road to 100G Ethernet

Most readers of this blog are familiar with Finisar, the largest provider of low-end optical modules for enterprise applications. I’ve written about them extensively. You may not be familiar with Luxtera. They are building integrated electrical/optical chips in silicon and flip bonding laser sources to the top of the die. The result is a single die with all of the electronics and optical components resulting in a highly compact solution. I would be interested in what readers have to say about the pros and cons of their solution.

It was also interesting to see how many people were familiar with the Cisco Monopsony, though most people I spoke with seemed to attribute the work to Lightreading.

Jerry Rawls (CEO of Finisar) didn’t offer anything spectacular in terms of technology. His presentation and discussions on various panels throughout the day left the impression that Finisar realizes what it should be- a low cost producer of commodity product. There isn’t a lot of flash in what he presents, and that’s fine by me.

Here are my notes from the conference. Italics are my comments. Plain Text is my transcription

Jerry Rawls- Finisar Chairman and CEO

Optics are back. Standard fare here, bandwidth and storage explosion, nothing revolutionary. Cites Parkinson’s Law. Never heard the law used in this context before, interesting stuff.

How do we get to 100G? Direct Modulation, WDM. Phase Shift Keying. IEEE starting to meet with volume production capability in 2011. He thought WDM would be the first path. Note I think Luxtera’s technology is ideal for such a solution (long wavelength, high integration) - what do you think?

Alex Dickinson - Luxtera CEO

CMOS always wins. Ah, that’s always a no lose throwaway argument. Some good charts on optical vs. electrical power requirements for transmission over equivalent distance.

Not really presenting exactly what Luxtera does. “Transition of optics to CMOS”. Not doing the technology justice here guys… too dumbed down. Let’s get hardcore.

Problem: Electronics are printed, optics are assembled. OK, getting better, I like that a lot! Playing pretty fast and loose, claims they have a chip that integrates 2 full 10G XFP’s into a single chip… that is not really true - they can’t integrate the light source - that still needs to be assembled. They also don’t highlight that this thing only works in the 1310nm lambda range.

Lots of tangential applications, very stable oscillators for wireless applications, medical applications, consumer optical interconnect. HDMI replacement? They should be targeting consumer applications if this is so low cost.

Q&A From Moderator

Question: Finisar spending a lot on 10G. How will you compete with Luxtera technology? Rawls: Thinks Luxtera is a neat technology, but there are some problems. Not a near term issue. Translation: Build 1M of these things with decent yield then ask me that question. He handled a tough question with tact.

Question: When does Luxtera go into production? Dickinson: First product will be a quad (I didn’t hear a date). Challenging when Cisco is the largest customer.. a .. uh … searching for the word (MONOPSONY - You heard it from me).

Audience Questions:

Question: What about SFP’s? How does Luxtera succeed without addressing SFP’s? Dickinson: Address markets with faster turnover, more sensitivity to cost. Possible that cost benefits of this solution would eliminate the SFP and 10GE boards ship with optics integrated? This makes sense if 50% of applications are short reach only.

Question: How do you make money given that Cisco makes so much more reselling modules than you do selling modules to Cisco (Uh, clearly this guy reads Lightreading or my blog) Rawls: Indicates that margins are going up and yes, while Cisco does make a lot of money off reselling modules it is not as high as what the author of that article claims.

Question: I identify myself as the author of “The Optical Illusion” and ask the question “I will not ask Jerry because as a major supplier to Cisco, I don’t want to put him in a tough position. But how could Luxtera break the Cisco Monopsony? Answer: Need to stop using identical technology and deliver a unique component with an advantage over the state of the art. Disrupt. De-commoditize. Jerry adds in - Finisar gross margins have gone from 20% to 40%. Vertical integration works too and we are a big player (I Agree, good answers from both).

Full Disclosure: I hold positions in Finisar

Comment on this article

Comments

[Bill Woodruff:]

Andrew,

Help me out here. I thought that Luxtera was basically a silicon based optical modulator w/ the ability to add some other circuits in CMOS. And, the optical waveguide integrated on the silicon was opitmized for 1550nm.

Is the silicon sufficiently transmissive at 1310nm? Or is this a technology limited to 1550nm?

Bill

2006 Oct 13 12:41 PM

[Andrew Schmitt:]

My article should read 1550, not 1310. Theoretically it could work at 1550. Is this the help you needed?

How are you Bill?

2006 Oct 16 02:57 PM

[numinary:]

Regarding Luxtera technology as a panacea for 100G, as noted they currently have some deficiencies such as (a) operate at only 1550nm wavelength, (b) Si modulators have no capability to offer phase control necessary for phase shift keying or other modulation techniques, (c) modulator technology does not offer good extinction ratio for longer distance signal propagation, (d) laser chips must be custom designed to correctly mount to their IC platform. The Luxtera technology is very interesting, but as the entire communications industry revolves around published standards, Luxtera is in a bit of a bind -- their tech does not line up well with any of the existing standards. Luxtera will sell custom interfaces as they try to manipulate their technology and the emerging standards to line up better with one another.

FNSR is in the midst of a squeeze as low cost high volume optical module shops from Asia provide constant pressure, while at the high end they have had little to no success at 10G modules. This is why it is critical for them to continue to invest in 10G and other markets. Of course good companies tend to use challenges such as this as a mechanism to develop new expertise and remain market leaders -- it remains to be seen if FNSR can do this, but they have a better chance than other public optical module companies.

2006 Oct 16 03:27 PM

[Andrew Schmitt:]

I think you are wrong on FNSR. I used to think the same but have corrected myself.

2006 Oct 17 12:13 PM