Article Info

Cornering the Commodity Market

The most under-reported but most significant announcement at OFC2007 was Finisar’s (FNSR) Fiber to the Home (FTTH) product. Most optical vendors are communicating their intent to NOT make a product for FTTH or Passive Optical Networking (PON) applications. When the worlds highest volume optical module supplier decides to go the opposite way, something noteworthy just happened.

Vendors express their caution the same way – “We see this as a commodity market and we aren’t commodity suppliers“. I heard this statement directly from JDSU (JDSU) and several other module vendors. Vendors looked self assured when they said it. Sell side analysts love to hear it.

The problem is that most of the optical component and module market will ultimately be commodities, and if a supplier isn’t building commodities they won’t be building much of anything.

Vendors communicating this message have a noble goal; they have no wish to repeat the over-investment sins of the past. In the year 2000 there were over 30 companies building optical modules; it would be a comical understatement to say the market was oversupplied. This memory triggers a flight response in incumbent module suppliers that encourages them to run from any market that looks marginally profitable or highly competitive. The FTTH market is a great example.

Unfortunately for them, any technology market that is big is also commoditized and highly competitive. The communication semiconductor company that figured this out first was Broadcom ::(“BRCM”)::, which won’t even touch a market that isn’t $500M.

The debate about the FTTH market isn’t whether it will be big but whether it will be profitable. PON is becoming the global access technology of choice for new construction. (self quoted from “FTTH in China Part I“)

Consider the numbers in China alone: the massive migration of labor to urban areas is creating the biggest greenfield telecom installation opportunity in history. Three hundred million Chinese will migrate to the cities in the next 15 years -the equivalent of creating a Los Angeles/Orange County/San Diego urban each year until 2021.

Any component market this big is by definition is a commodity market. Marginal producers that cannot cover large fixed costs are unprofitable, suppliers with scale and leverage on such costs are very profitable. Therefore, only vertically integrated market share leaders have any hope of profitability. The trick is for such supplier to arrive at the market “fastest with the mostest“.

Bookham (BKHM) attempted the same strategy in higher end telecom products and met with disaster. I outlined the risks in combining vertical integration with niche products a year ago (see last paragraphs of “Bookham, China, and the Optical Component Market“). I still stand by my assessment that a vertically integrated strategy makes sense for high volume, commodity module markets.

From Bookham, China, and the Optical Component Market (May 2006)

Finisar’s decision to introduce a FTTH product brings them into direct competition with Luminent (MRVC) and Fiberxon, two companies that once merged will form the biggest incumbent supplier of optical components for the FTTH market.

Luminent is a big supplier into Verizon’s B-PON deployments, and Fiberxon is an on-again/off-again supplier into NTT’s GE-PON deployments (discussion of these technologies can be found here). The biggest competition to-date for Luminent/Fiberxon has been the propensity for FTTH equipment suppliers to design and build their own modules at contract manufacturers. If (an if too big for me) the merger is successful, the resulting vertical integration will make Luminent/Fiberxon a force to be reckoned with.

The product Finisar announced is for G-PON, which Verizon will begin deploying this year as part of it’s transition from B-PON. Finisar is clearly trying to capitalize on the market breakpoint created by this transition.

Ultimately we feel GE-PON volume will outstrip G-PON volume based on our assumption it will be the technology of choice in Asia. According to Finisar, they are not betting on G-PON vs. GE-PON. They simply see G-PON as a good entry point for FTTH optics, given that few domestic suppliers exist to meet Verizon’s demand.

I believe this is only a first step, as Finisar appears to understand the importance of being the low-cost supplier (not necessarily low price) and unlike other vendors, embraces the idea of supplying commodities. This attitude will be required for success in not just the FTTH market, but the optics market in general.

Author is long Finisar


Comments are disallowed for this post.

  1. Andrew,

    Finisar noted that they intend to take advantage of high volume production with the FTTP modules. Do you think that means incorporating PLC technology? If not from the outset, how long before this becomes standardized within GPON modules?

    Posted by jeebe | April 4, 2007, 5:12 PM
  2. Its time to revisit your conclusions on GE-PON and the technology choise in Asia: CNC launched GPON trial

    Posted by synergysense | April 5, 2007, 1:20 AM
  3. Nope. I’m sticking to my guns.

    Posted by Andrew Schmitt | April 5, 2007, 6:25 AM
  4. I would tend to agree with Andrew on his China/GE-PON statement. I believe one of the reasons for the Luminent/Fiberxon pairing was to position Luminent better for a GE-PON opportunity in China. Fiberxon has a solid GE-PON portfolio, as well as some strong relationships there with the FTTH suppliers like Fiberhome and technology partners like silicon vendor Teknovus. Factoring in not only Fiberxon’s GE-PON technology but also their 10gig product line and the added manufacturing assets, I think the acquisition was a good one and will be completed shortly.

    If Luminent’s initial B-PON deployment with VZ is any indication, FNSR may have a rough time at the outset as a new player trying to vie for market share. Luminent’s margins in those early stages were virtually zero. Even now, Luminent’s triplexer margins are considerably lower than what FNSR is used to in the LAN/SAN space, even with 95% of the Verizon market share. Before getting too excited about the big PON subscriber projections put out by Infonetics et al., I think FNSR investors should realize that, at least short term, overall margins could be compromised.

    Great blog, Andrew. Definitely a daily “must read” for any serious telecom/tech investors out there.

    Posted by Niles | April 5, 2007, 12:05 PM
  5. How do u think abou Neophotonics?
    It also announce its FTTH product which include EPON &GPON are good.

    Posted by Ivy | April 6, 2007, 1:40 AM
  6. PON is a relic of the US. It must be given a quick burial, with garlic. Countries with a residential paradigm based on apartment towers are better advised to backhaul via WDM boxes in the basement. Bandwidth is incredibly cheap, it is already cheaper than water in many places. Why throttle it with a dumb idea like PON?

    Posted by Bandgap | April 7, 2007, 10:02 PM
  7. I’m trying to find anything in the above comment that is true.

    1. Bandwidth is not cheap. It’s getting cheaper but thats not the same as cheap.
    2. PON has nothing to do with the US, and was installed in Japan in quantity well before it was here.
    3. Apartment towers are best suited for VDSL and Ethernet backhaul. If fiber is in place WDM never makes sense. So far, WDM only is economical if you have fiber exhaust.
    4. PON has nothing to do with backhaul. BT is looking at WDM PON with backhaul, but it’s still PON.

    Posted by Andrew Schmitt | April 8, 2007, 1:09 PM
  8. There’s something which is still unclear to me about the use of DWM on the access vs. transport side of the network.

    assuming a carrier links a MDU with 60 apartments with FTTH. Can it deploy a limited amount( lets say 2-3 lines) of fibre from the central office to the basement, and from there simply deploy optic fibres, one for each unit equipped with an ONT?
    If yes, the load on those 2-3 cables connecting the splitter to the central office should be pretty big. In that case, WDM makes sense also as an access technology???

    Is that what the term “WDM-PON” refers to ?

    I probably mixed up some technical terms but the basic idea is: deploy less fiber from the central office to the basement, and use WDM to deal with the fiber exhaust.


    Posted by ohad | April 9, 2007, 4:24 AM
  9. I am at odds with the promoters of WDM PON as to it’s usefulness. Google “WDM PON British Telecom” for some interesting papers. I think they have the only WDM scheme that truly makes economic sense, long reach, 10G, WDM PON that cuts the number of central offices in Britain from 5000 to 100. That’s revolutionary.

    Posted by Andrew Schmitt | April 9, 2007, 7:45 AM
  10. I criticized PON as a US relic because of its telco-centric assumptions. Greenfields can be built cheaper & simpler based on the PC economy. For example, 100BASE-T is so absurdly cheap that it should be the distribution scheme of choice for apartment towers. It can be upgraded to 1000BASE-T with no change in wiring, esp since many of the apartment towers are yet to be built.

    CPE power is another hoary chestnut. Apartment complexes already have UPS installed for emergency lighting, and anyway in many countries there’s no point trying to call in the first place :-) WDM reduces patch panel hell on the CO side, & it can be coarse to start with.

    Posted by Bandgap | April 9, 2007, 10:22 AM
  11. You cannot upgrade 100 to 1000 base T over the same topology because they have different reach limitations. The apartment you describe would have equipment on every floor of the building. Who is responsible for maintaining it? You need more equipment in the basement.

    PON eliminates all of the active electronics between the demac and the CO. It is attractive to the Telcos because of low OPEX and appears to be infinitely upgradeable.

    I’ve heard the patch panel hell story for the CO but it sounds like nonsense to me. Going from copper to PON reduces the overall cable count by a factor of 32, and the cables themselves (fiber) are smaller & lighter. GOing to WDM PON to reduce it further seems like a lot of technology for relatively little incremental gain.

    Posted by Andrew Schmitt | April 9, 2007, 11:06 AM
  12. btw,

    in an interview with Verizon representetives they claim that in older buildings ,they would rather use special MDU ONTs and link each apartment via coax. VDSL is their last choice . Any idea why they prefer coax, even though a trial conducted using VDSL2 in the empire state building showed a rate of 75Mbps?
    Perhaps the fact they chose MOCA for home network wiring played a role?

    Posted by ohad | April 9, 2007, 1:30 PM
  13. That’s a great question if it is true – do you have the link to the info?

    I would guess that coax in an MDU would be a lot trickier given it must be shared (it might even be owned by) the cableco. There also may not be home runs of coax to the wiring closet.

    Coax would be preferred because it makes sending the video a whole lot easier. But I bet the infrastructure headaches in an MDU situation offset this.

    Posted by Andrew Schmitt | April 9, 2007, 1:40 PM
  14. The way apartment towers work is they have their own residents association that handles utlities issues, and the bandwidth pipe is just another. For example, the new trend is that cooking gas gets supplied to a central manifold and then metered to each apartment. What happens if the gas leaks and the whole tower blows up? Nobody is worrying about that, why would they worry about a 6 nines telco SLA? The ethernet switch gets put into a locked cabinet and the service provider has a key.

    No TV? Just call the guy and yell – he’s round the street corner, anyway. It works just fine – no ‘customer care’ representative, no waiting all day for the service tech to show up. That’s capitalism in action – in comparison, big telco’s are socialist empires.

    Posted by Bandgap | April 9, 2007, 2:31 PM

    “We realize we can’t pull fiber into the living unit every time,” says Brian Whitton, executive director of Access Network Design and Integration with Verizon. In those cases, Verizon installs a special MDU ONT in the basement of the building that serves multiple units. The last few feet are usually covered by the building’s coax wiring. In instances where this can’t be done, Verizon uses VDSL.

    How much the network performance is compromised depends on the unit’s distance from the ONT and the type of wiring used. “We’ve done testing in the Empire State Building. With VDSL1, we can deliver 30 Mbit/s and with VDSL2 we can deliver 75 Mbit/s,” says Whitton.


    Posted by ohad | April 9, 2007, 3:46 PM
  16. Trackbacks / Pingbacks

  17. MRV, Luminent, and Fiberxon at Nyquist Capital | July 3, 2007, 11:12 AM