Optical components

Can Vendors Build Their Optical Components?

What if optical equipment vendors went back to the days of building their own optical components?

The big systems vendors are starting to consider it, some industry executives say. And as far-fetched as the idea might sound, a few of those executives are convinced that vertical integration is in the optical industry's near future.

"This is their way of trying to differentiate against somebody like Huawei Technologies Co. Ltd. ," says Paul Meissner, CEO of laser vendor Santur Corp. "Alcatel-Lucent (NYSE: ALU) and Nortel Networks Ltd. are very worried that they've leveled the playing field by spinning out this technology."

Systems vendors pinged by Light Reading say they're not interested -- but then again, the way they're handling the 100-Gbit/s product generation actually supports the idea of vertical integration. (More on that in a minute.) And no source was claiming that any big acquisitions are in the works yet.

On the surface, vertical integration will be outshouted by 100-Gbit/s products at OFC/NFOEC next week. But for years, the same question has lingered over that conference: What's it going to take to make money here? (See Optical Components: Still Too Crowded.) According to some, vertical integration is being considered as a possible answer.

The big mistake
Proponents cite two arguments supporting vertical integration. The first is that it gets cutting-edge equipment to market more quickly, with Nortel and its early 40-Gbit/s and 100-Gbit/s products being held up as an example.

But some sources believe there's a deeper unrest behind this: Systems vendors, they claim, are realizing that a world of interchangeable components, all adhering to multisource agreements (MSAs), has boiled away too many of the differences in equipment, and software doesn't make up for the shortfall.

"The systems players are seeing they're all competing on price, and they're all using the same contract manufacturers," says one components executive requesting anonymity. "All these guys have got to be talking about it."

Underlying either theory is a belief that it was a mistake for systems vendors to divest their components divisions. They point to the financial misery that followed two of the biggest moves in that trend: Nortel selling its components division to Bookham in 2002, and Alcatel Optronics to Avanex in 2003. (See Bookham Buys Nortel's Components Biz and Avanex to Buy Alcatel, Corning Units.)

"I think that the de-verticalization was an unnatural act. It was a financially driven act," says Karl May, CEO of systems vendor Vello Systems (formerly OpVista). "Our view is that vertical integration does have to return. If all the systems are based on the same underlying IP [intellectual property], how exactly do you distinguish yourself?"

But even if vertical integration is sensible -- and not everyone agrees on even that point -- it may be too late to try.

"I don't think I buy that," says Andrew Schmitt, an analyst with Infonetics Research Inc. "The toothpaste is out of the tube. People are going to have to find a way to use the building blocks that are out there."

When it works
Infinera Corp. (Nasdaq: INFN) certainly shows how vertical integration can pay off, but the company started out by developing the chip, then built the system to go around it. Most systems vendors would be doing things in the other direction.

So it's Nortel that gets cited more often as a do-it-yourself example. The company got a headstart in 40 Gbit/s and 100 Gbit/s by designing its own Coherent receivers and its own optical modules. "They've got all this electronic dispersion compensation in a series of chips that are doing an unbelieveable job, and to date, nobody's been able to mirror that," says Eve Griliches, an analyst with ACG Research .

That doesn't guarantee that Nortel wants to make more of its own modules and components, though.

"We're not interested in the head-to-toes approach," says Mike Adams, director of portfolio strategy for Nortel's Metro Ethernet Networks group, which is soon to be acquired by Ciena Corp. (NYSE: CIEN).

But he concedes that, "for those types of things where you can differentiate in the marketplace, I can see why other people are beginning to talk" about vertical integration.

That sets up a likely limit for vertical integration. Companies will do it only to the extent that they have to, Infonetics' Schmitt says.

"Somebody like Nortel right now is winning at carriers they'll be shipping to for a decade. It's very important to be first, and for this market [100 Gbit/s], to be first they had to do it themselves," Schmitt says. "It only works early in the cycle. Once what they have is no different from what they can buy outside, they'll buy from outside."

Huawei, like Nortel, built its own chips for a 100-Gbit/s coherent receiver. (See Huawei Touts End-to-End 100G.) But again, the company says it's not interested in doing everything itself.

"We only do the pieces such as the DSP [digital signal processor]. Other components we get from other vendors," Glen Fang, director of Huawei's WDM product line management, told Light Reading last month.

In the weeks since that discussion, though, multiple sources have told Light Reading that Huawei does indeed want to start an optical components division. In-house optics would certainly seem to fit Huawei's culture; the company gets some of its chips through a spun-off ASIC division called HiSilicon Technologies Co. Ltd. (Huawei could not immediately be reached for comment.)

Small bites
To vertically integrate on any substantial scale would certainly fall under scrutiny, so it wouldn't be an easy decision for the big companies. "I don't know that anybody's ready to pull the trigger. We've seen them fly by, but none of them have put an offer on the table," Santur's Meissner says.

Many companies might choose not to take the plunge unless the competition does it. One or two component acquisitions "could put a real dent in the market," Griliches says. "Then we could get to a situation where everybody starts snapping up the component vendors as a defensive move."

It's unlikely the trend would start with a bang, though. It would be dramatic to see a systems vendor snap up a big components and/or subsystems player -- but Finisar, JDSU, Oclaro, and Opnext have all been beefing up via acquisitions. (See Finisar & Optium Challenge JDSU, Bookham, Avanex Tie the Knot, Oclaro Makes Its ROADM Bid, and Opnext Steps Up With StrataLight.)

"The question is: What size is the bite?" says Larry Schwerin, CEO of components vendor Capella Photonics Inc. "You could bring in a lot of stuff you don't need."

— Craig Matsumoto, West Coast Editor, Light Reading

paolo.franzoi 12/5/2012 | 4:40:54 PM
re: Can Vendors Build Their Optical Components?


The only way it makes sense for people to make their own components is if they can either:

a) Get enough competitive advantage because of the capability of the component


b) Lower the cost of the end product by more than the R&D cost (adjusted for risk and inflation, etc.)


This generally means that the end product is in volume or going to be in volume.  Remember the components need to be invested in and have a roadmap comparable to components available on the merchant market.  The way things are right now, nobody can afford the risk adjusted investment that I can see - unless they are Cisco.



ROADM 12/5/2012 | 4:40:54 PM
re: Can Vendors Build Their Optical Components?

Optical components like transceiver and transponder is do-able to assemble if vendors have in-house optical centre.

However, thing will be more complicated for thing such lasers where you need wafer fabrication to make them, espicially now, we are talking about parallel optic, where we have such more than 10 arrays of laser in one small chip.

Yield is low and a lot of headache indeed...


nelson3748 12/5/2012 | 4:40:54 PM
re: Can Vendors Build Their Optical Components?

basic finance 301, in the long run specialization beats integration every time; however, tech has a history of ignoring basic finance and destroying capital, over and over again, caused by hiring CFOs from inside who have tendencies to tinker destructively. NT is a great example of what NOT to do; winning 100G contracts in the future is worthless when your model bankrupts you.

specifically: if Integrator A buys OC A, then Int B will not buy components from OC A (most likely). this will not work if OC A and OC B are better at different products: if A has been maximizing profits buying from A & B, then they will be hurt by the inability to buy from B (ie, their output will be more expensive and/or lower quality).

what it all boils down to, are the OCs adding value between the maufacturers and the integrators? if so, they will stay independent, because they are actually increasing the value of the integrators' output. if not, they will be bought (or worse). based on what A Schmidt has said, i am assuming that the OCs do add value and they will stay independent.

Stevery 12/5/2012 | 4:40:54 PM
re: Can Vendors Build Their Optical Components?

So it's Nortel that gets cited more often as a do-it-yourself example.

So a company that drove itself into bankrupcy is cited as a good example of DIY?

In truth, I think vertical integration might be the right thing.  So far, there is not enough profit for the entire chain to thrive, which points to either M&A horizontally or vertical integration.

lightmonkey 12/5/2012 | 4:40:54 PM
re: Can Vendors Build Their Optical Components?

I don't think systems vendors should run out and start their own components companies.  However, if you look at optical components in general, very little has changed in 10 years.  Yeah there's some new stuff but there's really not much new in the way of new technology.  It's basically the same old crap, slap together some optics with some epoxy. 

JDSU has made some inroads with their PLC devision but their PLC's don't have much other than a few couplers.  And of course Infinera has made a splash with their PIC.  Other than that, there's not a whole lot of "innovation."  What you get from component vendors are mostly "obvious" ideas.  Even JDSU seems to really be de-emphasizing their components business.  So where's the innovation going to come from.

If you are a systems vendor and you want to do something nobody else can do, then you'll want to have some in-house components anyway.  The DSP chip that Nortel has is a good example.  But even then, you still need a deep in house understanding of components, particularly in coherent systems, so you can design around the impairments of those components.  In some cases, it'll be hard to know what all those impairments are unless you are actually building the stuff. 

Overall, I think the only innovation is going to come from some kind of vertical integration.  That's because if you are a component vendor, you have to appeal to the broadest possible market at the lowest cost.  This seems by definition to make it impossible for them to push the envelope.



rahat.hussain 12/5/2012 | 4:40:53 PM
re: Can Vendors Build Their Optical Components?


quoting brookseven:

"The only way it makes sense for people to make their own components is if they can either:

a) Get enough competitive advantage because of the capability of the component


b) Lower the cost of the end product by more than the R&D cost (adjusted for risk and inflation, etc.)"

... or,

c) there is no other way to survive, with huawei winning network bids on the one hand, and big guys like ericsson and alcatel-lucent winning larger contracts for managing carrier networks.

i don't blame the small ( $2-3B) equipment makers for thinking this way/


paolo.franzoi 12/5/2012 | 4:40:52 PM
re: Can Vendors Build Their Optical Components?



But if they don't get a competitive advantage from features or cost, maybe it is time to realize it is time to get a new business.




Pete Baldwin 12/5/2012 | 4:40:50 PM
re: Can Vendors Build Their Optical Components?

Thanks for a great discussion, guys.

I was thinking about vertical integration as something the big companies would do, but I like Odo's point -- that it's a tactic for the smaller guys to try to survive.  They do have to try something radical.  Maybe it could even help get them acquired-?

rswitch09 12/5/2012 | 4:40:49 PM
re: Can Vendors Build Their Optical Components?

So, one would only go vertical if they want to be Cisco?

But this is exactly the point. The vertical integration makes sense only if you are a Tier 1 and you want to protect your market from a Huawei.

A vertically integrated market makes sense for small volume, high price system markets, such as the 40G/100G core and metro core. In such a case, the Tier 1 system vendors (and they are the only one to compete in this market) must protect their investment and avoid a price war. Let's face it, Cisco/ALU/ERIC will find it very difficult to be a price leader and beat Hauwei/ZTE at this game. Their only way to protect themselves is through technological advantage. Tier 2 TEMs are not players in this field.

From a component vendor perspective it also makes little sense to get into this low volume / high development cost, risky market. And if you don't realize that, the market or the investors will tell you that. The other problem is that Tier 1 TEMs are very reluctant to buy from component vendors that are themselves Tier 1 companies.

Which leaves us with the vertical integration choice.

The alternative would be to create building blocks that still leave a lot of IP for the Tier 1 TEMs to develop. Something along the lines of FPGAs.

Pete Baldwin 12/5/2012 | 4:40:28 PM
re: Can Vendors Build Their Optical Components?

We've posted a post-OFC sequel to this story:

Vertical Integration Takes Its Lumps


Sign In