Optical components

Vertical Integration Takes Its Lumps

The topic of vertical integration was fair game for discussion at last week's OFC/NFOEC, but lots of people at the show were skeptical about it happening.

We're talking about the idea of a systems company either acquiring an optical components company or building its own components division. The idea of building one's own optical modules out of purchased components counts, too, but on a lesser scale.

Some executives and analysts claim there are OEMs seriously considering that route, as Light Reading reported just before OFC/NFOEC. (See Can Vendors Build Their Optical Components? and OFC/NFOEC: Mergers Haven't Gone Far Enough.) It would reverse the divestment trend that's dominated since 2002.

What we're not talking about is the vertical integration where a components or subsystems company owns its own manufacturing. That's already being tried -- and is working for Finisar Corp. (Nasdaq: FNSR) and Oclaro Inc. (Nasdaq: OCLR), said Paul Bonenfant, an analyst with Morgan Keegan & Company Inc.

The one systems company that everyone can believe would make its own components is Huawei Technologies Co. Ltd. It's just a very Huawei thing to do. (Huawei didn't respond to a brief pre-OFC/NFOEC query about this.)

But the thought of anyone else doing it? Most people at the show were skeptical.

"If you're an OEM, is your best strategy to imitate Huawei? You're not going to out-Huawei Huawei," said Craig Iwata, manager of corporate marketing for JDSU (Nasdaq: JDSU; Toronto: JDU). The only company he could envision trying it would be ZTE Corp. (Shenzhen: 000063; Hong Kong: 0763).

Complete vertical integration seems out of the question, just because the technologies behind optical networking are so widespread.

"What's being underestimated is the cost of doing that," said Terry Unter, CEO of Mintera Corp. "You have micro-optics, you have indium phosphide, you have other compound semiconductors... I doubt any TEM [telecom equipment manufacturer] can build volume to amortize all those technologies."

"It's extremely expensive to make those investments," said Tom Fawcett, director of marketing for JDSU's transmission modules unit. "A middle ground we see is people pulling in at the component level and doing their own integration."

But equipment manufacturers would have a hard time building up that expertise, Unter claimed.

Of course, it's natural that module and subsystems vendors would reject the idea of a vertically integrated systems player, as it could hurt their businesses. And it would dilute their role as the bastions of innovation in an age where the systems vendors have had to skimp on R&D.

But even people without a direct stake in the game didn't like the idea. "Usually, you do it to get control of fixed costs," said Andrew Schmitt, an analyst with Infonetics Research Inc. "For the equipment guys, getting vertically integrated isn't going to help the cost structure."

Why do it, then?

The theory that spurred the original Light Reading article is that systems vendors are running out of ways to stand out. Optical modules, for instance, all conform to multisource agreements (MSAs). That saves the systems vendors some money, but it also means they're all pulling lookalike optics out of the same pile. Maybe, the theory goes, they could find a new spark by building from the components up.

The idea particularly works with new generations of technology. Nortel Networks Ltd. built its own optical modules for 100 Gbit/s, because the pieces weren't available yet, and the company has a headstart and a Verizon Communications Inc. (NYSE: VZ) contract to show for it. (See Verizon Switches On 100G in Europe.)

One fan of vertical integration is Ray Conley, a partner at Palo Alto Investors . He noted there could be systems-design advantages to owning one's own components. (See, again, OFC/NFOEC: Mergers Haven't Gone Far Enough.) And he thinks it's the best way to avoid competing on price alone.

"Huawei and ZTE are taking advantage of commodotization of components and are winning because it's a price game. If the other vendors want to have an advantage in a full value-add game, they're going to have to have full vertical integration," Conley said during the The Optical Society (OSA) Executive Forum, a pre-OFC/NFOEC gathering.

— Craig Matsumoto, West Coast Editor, Light Reading

pinny 12/5/2012 | 4:40:26 PM
re: Vertical Integration Takes Its Lumps

Makes a lot of sense. True, fixed expenses need to be amortized over as many customers as possible to justify the entire investment in generic technology development but ...... there aren't so many customers out there anymore! I would argue that the combination of competitive advantage generated by these new optical technologies that will be developed in house (or in tech JVs between optical equipment vendors) will have enough market potential to justify the investment.

The optical component market, at its current state, is not sustainable - few companies with tons of talent invest a lot of money in developing cutting edge technologies and products only to end up with meager Gross Margins and razor thin Operating Margins (if at all).

steady 12/5/2012 | 4:40:26 PM
re: Vertical Integration Takes Its Lumps Isn't component business also price game? How are the two price games integrated to be an innovation game?
IJD 12/5/2012 | 4:40:21 PM
re: Vertical Integration Takes Its Lumps

The real issue is that the difficulty of developing the technology (e.g. for a 100G coherent transceiver) has suddenly taken a quantum leap over previous generations -- it's no longer a case of developing some relatively cheap small-scale components and bolting them together, it needs a lot of complex system and signal-processing IP, many man-years of effort and many millions of dollars to develop the components and build a module.

The optical market has traditionally had many small players (as well as quite a lot of large ones, of course) but it's difficult to see how they can stay in markets like this where the bar to entry has suddenly been raised so much -- and as demand for more bits/Hz over capacity-limited channels rises inexorably, it's only a matter of time before coherent technology and/or MLS moves down into metro, then short-haul -- then LAN, then backplane, then chip-to-chip...

The number of players must fall as time goes on and the cost of developing components for new systems spirals ever upwards -- how many companies will have the ability and be able to justify the cost (>>$100M?) of developing their own 1Tb/s coherent transceiver, maybe in 11nm technology in 5 years time? They'll need to be pretty damn sure of capturing a very big slice of the worldwide market to pay for this.

In the end the answer surely has to be that this end of the optical ecosystem will end up like CPUs and foundries, with only a couple of big players -- maybe module suppliers who sell to all, maybe merged telecomms companies, maybe industry consortia -- with product differentiation either through software, support, systems, or getting your own "special sauce" hidden in one corner of the chip (next to everybody else's, of course ;-)


Pete Baldwin 12/5/2012 | 4:40:21 PM
re: Vertical Integration Takes Its Lumps

The difference here is that the systems vendors wouldn't be selling the components themselves. Rather, they'd be using them to build newer products ahead of market (100G, for instance) ... or possibly to increase system performance.  Pinny's post explains the thinking.

Ironically, what's spurred the idea seems to be the lookalike nature of MSAs -- which is something the systems vendors (especially Cisco) demanded in the first place.  Apparently they're not getting the differentiation out of software that they expected.

Whether it's a good idea is still up for debate, as the article shows. I'm certain it's being seriously considered, though.

cw.774 12/5/2012 | 4:40:19 PM
re: Vertical Integration Takes Its Lumps


I think getting differentiation out of the software is not a problem for CISCO or anybody with good operational system.  Rather, Ian points to a shear leap in technical challenge in a.) modulation formats and such, that are not core competencies of today's small-to-mid sized (and one big ...) optics manufacturers and b.) the newer optical technology which is very cost intensive to develop as it marries high end processing technology with the optics in a somewhat rare union of expertise today.  This will change of course.  


MSA's based on IETF/OIF input etc... may not seam so necessary for the bench top demo stuff, but they are already used for packaging show demos (CFP, XFP-E).  I agree in thinking low level components and systems companies may not need the MSA "middle-man" for these technologies at 40/100G etc. going forward.  This should be interesting.

Sign In