Components, Anyone?

The last time I felt moved to write a column, I’d just had a couple of pints of Boddingtons with Kevin Kalkhoven (see Kalkhoven's Five-Year Plan). This time around, inspiration struck while I was browsing our message boards after chomping Thai fish cakes with Steve Georgis, the president and CEO of Network Photonics Inc.

The message that got me going relates to a story we ran earlier this week – BrightLink Brings Home the Bacon. The correspondent suggests that BrightLink Networks Inc. would have been better off sticking to its original idea: selling its innovative switching fabric to OEMs (original equipment manufacturers), rather than using it to build its own switch.

That way, says the message, Brightlink might have been selling its silicon to the likes of Ciena Corp. (Nasdaq: CIEN), Nortel Networks Corp. (NYSE/Toronto: NT), and Sycamore Networks Inc. (Nasdaq: SCMR), rather than struggling to compete with them.

It just so happens that Network Photonics faced a similar issue in its early days and reached the same conclusion as Brightlink – and Georgis told me why, while we were chomping on our fish cakes in the Blue Elephant restaurant in Fulham, London.

Georgis’s comments suggest that component startups are in for a tough time – tougher than those already being experienced by systems startups.

First off, a bit of background. Network Photonics is developing the metro equivalent of the all-optical network idea promoted by Corvis Corp. (Nasdaq: CORV). In other words, its switch does wavelength multiplexing and wavelength switching in the same box. This eliminates the need for separate DWDM gear, which typically bumps up costs a lot and requires a dreaded optical-electrical conversion.

Network Photonics won’t say how it achieves this feat, although Georgis says that, unlike Corvis, he will reveal all, once the patents are watertight. However, it all comes down to a single widget that’s based on MEMS (micro-electro-mechanical systems) technology but is more than just the usual array of tiny tilting mirrors. It also includes some secret stuff that splits light into different wavelengths as well as switching it.

This widget gives Network Photonics its edge, according to Georgis. Most of its potential competitors are buying in MEMS-based switching subsystems from OMM Inc. and integrating them with other off-the-shelf components. They don’t own the fundamental technology themselves: They're just doing systems integration, he contends.

Georgis points out that Ciena got an edge in DWDM by developing its own fiber Bragg gratings. Ciena has achieved the same thing with grooming switches by acquiring Lightera, a startup that was developing its own ASICs (application-specific integrated circuits). That enabled Ciena to jump ahead of Sycamore, which has relied on off-the-shelf switching fabric.

All the same, I asked, why doesn’t Network Photonics follow in OMM’s footsteps? That is, sell its widget to OEMs and avoid competing with the folk indulging in these systems integration projects – which include heavyweights like Alcatel SA (NYSE: ALA; Paris: CGEP:PA) and Siemens AG (NYSE: SI; Frankfurt: SIE).

It all comes down to the sales cycle, according to Georgis. If Network Photonics had tried to sell its widget to OEMs, it wouldn’t have gotten very far until it could ship samples and prove that it could deliver on its promises.

Then it would have had to wait until OEMs designed a system incorporating its widgets. Then it would have to wait another long, long period of time while the OEMs developed their products and got them into lab- and then field-trials with service providers. Then, and only then, the service provider might buy a decent number of switches, resulting in Network Photonics getting a decently sized order.

Georgis reckons this whole process can take one or two years, during which time the component vendor only sees a tiny trickle of revenues from the handful of components it sells OEMs for their prototype switches.

OMM is in this position, Georgis contends. It’s got an agonizing wait ahead of it to see whether any of its customers manage to sell a significant number of optical switches. At least one of them, ilotron Ltd., doesn’t look as though it will make it (see Ilotron Hits Hard Times).

And if OMM faces a worrying future, then imagine what all the other developers of MEMS-based switching subsystems must be thinking – those that haven't even got to the point of shipping commercial products.

Georgis hopes that making his own switch means that Network Photonics will reap its first revenues at least six months earlier than if it had tried to sell subsystems.

That’s consonant with what Brightlink once told me: It shifted its strategy to building a switch, because it was much easier to raise money for such a project.

— Peter Heywood, Founding Editor, Light Reading
byteme 12/4/2012 | 8:12:22 PM
re: Components, Anyone? The Blue Elephant eh - you're so passe Peter. Check out La Port-desindes - refined franco-indian cuisine. Owned by the same people. It's incredible. (Marble Arch tube, i think). Interesting column though. But if a company sells components/subsystems and then builds a switch with those components, what market is it playing in? Doesn't it annoy both its potential sets of customers who then go to someone else?
BlueWater66 12/4/2012 | 8:12:20 PM
re: Components, Anyone? I fundamentally disagree with these comments. The start-ups being described are extremely vertically integrated (phy up to software). They require a large number of engineers from vastly different backgrounds to build product. The number of critical failure points increase exponentially. The cash burn is tremendous. In 1999 and 2000, a number of venture groups with BILLION dollar funds started to worry about how much money was "at work" with specific investments. The venture partners wanted to limit the number of investments and increase the amount per investment. This is an insane approach to venture financing. It mirrored the stock market bubble ... and we're going to see a lot of these types of start-ups crash because of their burn rate and lack of focused differentiation. It also demonstrates the myopic investment approach of some of the larger VCs. TheyGÇÖd rather do this kind of deal, than bring in additional partners to reduce the number of board seats per VC.

wdog 12/4/2012 | 8:12:16 PM
re: Components, Anyone? Good article peter. What I like about their approach is the fact that they have intellectual property to build a long term company on. If they've gotten the architecture working, they can have a significant advantage over competitors. It's the only packet switch matrix I know of that can implement a true non-blocking switch. For grooming and other applications nothing should be able to touch it. Moster crossbar chips aren't the real solution to building big grooming switches for their competitors, software is. Nobodys figured out how to make the software a commodity, and doubt anybody will.
red1969a 12/4/2012 | 8:12:15 PM
re: Components, Anyone? I too disagree, The Blue Elephant is far better

Pat Mudge 12/4/2012 | 8:12:05 PM
re: Components, Anyone? Peter --

I disagree with your underlying premise in today's editorial. I do venture funding and know that systems start-ups are bleeding money like there's no tomorrow. Their burn-rates cannot be sustained. Whereas some components developers require very little funding and basically take the same 2 to 3 years to get product qualified.

Along with Siemens' Mustang Ventures and Azure Capital, I've funded a start-up in the 980nm space. They buy their diodes from qualified vendors and package them with proprietary processes that reduce manufacturing costs
by 35 to 50%. They were funded in Dec. '00 and will be profitable within two years. Their first prototype will be complete at the end of July and they'll launch their pilot production this October. Their next round will be in the $4M range and after that they'll fund their own development. Total funding from inception to profitability: approximately $10M.

I'm one of Brightlinks investors and know how difficult it is to bring a system to market. There was a time systems companies got funding but the current downturn will make future funding far more difficult.

Network Photonics may one day be a profitable company but not without burning through several hundred million.

sntwk 12/4/2012 | 8:12:03 PM
re: Components, Anyone? Well that depends on what kind of OEM partners you work with.

To me it sounds good idea that you provide the fabric silicon or just core logic to partners. I would go further and say provide that silicon or logic to complementary partners. For instance for a long-haul switch the complementary partners could be router vendors like Juniper, Cisco and Data Switch vendors like Extreme and Foundry.

If you supply to partners and you have dual business strategy of supplying components and building systems then obviously there will be difficulties. Still you could work out deals. For instance supplying fabric silicon for an european partner or for a metro partner or even to one of the two dominant long-haul players. I was surprised when Corvis bought a grooming switch BayLight (at last!) at the end of the show I was wondering why they bought it when they could have bought BrightLink much earlier and have product shipping.

One of the engineers working for large carrier told me that they view pure OXC as patch panel!
Back in the beginning of 2000 it was widely believed (were made to believe) that there would be no need for grooming switches. Even with all the hype around Supercomm 2000 exhibit halls last year, it was very clear to me that they will need grooming. I didn't even bother to visit some of the all optical vendors booths. I felt it is a wast of time. I wish I shorted all optical vendors last March!. Now what happened , not only carriers don't want to buy all optical long-haul switching gear , they are buying little of anything and whoever is spending is buying/bought cerent (25000 last count) and ciean core-director gear.

It is a good idea to do parallel business. Take a look at MIPS. They license their technology to a number of partners. Take Quantum Effect Devices that PMC Sierra bought. QED was essentially given license by MIPS to work on the core MIPS architecture and QED was a fab-less vendor who was both selling the chips as well as licensing the core in a way complimentary and competing with MIPS itself.
ownstock 12/4/2012 | 8:08:51 PM
re: Components, Anyone? The examples inside it are a bit old these days, but I would strongly suggest you read:

Simplicity Wins, by Rommel, Kluge, Kempis, Diederichs, and Bruck HBS Press 1995.

It is the best summary of how to run a high performance business I have ever read, period.

Summary: Focus, focus, focus...stay small to mid sized...often that means one (very good, high quality) component for example...

BTW, no VC I am aware of wants to fund a box level company, let alone a systems level company these days. It simply takes too much money and time to get them rolling...and in the end they succumb to being human...way over budget, and way under performing. I recently sat at a dinner table and had one of the valleys bigger VCs tell the systems guy next to me that he was damn lucky he got his money when he did...all his future investments will be in simple, extrmely well focussed business models...like components...

Look at Intel...basically one component...or GE, basically a collective of mid-sized narrowly focussed companies...

I think you better re-think your thesis and talk to some more VCs...and not over drinks...

ownstock 12/4/2012 | 8:08:43 PM
re: Components, Anyone? Peter:

Georgis is right about one thing:

A good systems play has to have a competitive advantage, and one grounded at the component level is very hard to beat...


It's also nearly impossible to carry off.

Georgis obviously thinks he can short circuit the time (and cost) by doing the component development and the systems development concurrently...

Building three companies (component, box, and system) at the same time is pretty hard work...

Having carried it off at least once in another life, all I can say is good luck...

This is the mentality / task of most large government funded projects...none of which have to answer to payback...and all of which have Uncle Sams pocketbook to fall back on.

The vast majority of which go way over budget, are late, and barely meet specifications...in spite of very able management, controls, latest technology, brightest engineers and scientists, etc...

The issue is the complexity of the project...or in this case, the business model...

The serial approach (component to box to system) is more logical from the risk reduction (and business) perspective...but then that would focus Georgis' investors on the true problem: classical innovators dilemma.

How can you sell a new and unique component without the box and system that use it?

The answer is to sell the component during development to a few already established box level or systems players...if they buy into it, then you probably have something worth investing in...

They get an enabling component, and you get an established customer base...

And you only have to build one company...not three...

But then you can only raise a few Million, not hundreds...

So what is Georgis' objective...building a real company and making a product or raising a lot of money???

Brattain 12/4/2012 | 8:08:08 PM
re: Components, Anyone? Why do they build systems and not components?

Because they (CEO, VCs) want a mega million
dollar pay day. Building systems means potentially bigger valuations.

Most guys running these startups and their VCs
don't give a hoot about building a sustainable
business. They just want to get rich quick.

So the goal is: Get noticed. Get acquired.
Get rich and retire to eating expensive
fatty foods and beer any day you want.

Oh forgot, CEO's, VC's, and columnists do that already.

Sign In