Infinera — new component paradigms start in systems

Giga Om and Andrew@Nyquist have good articles on Infinera today. I like Andrew’s comments wrt/ component development.

I’d add that the best vehicle for a “new component paradigm” is a “system” company, not a component company. The “system” company has the means, via the system, to demonstrate the capabilities of the component in the “real” world. The “system” companies future depends on proving this point. It is not easy as Infinera’s 6+ year effort demonstrates ( ie they have yet to be profitable. )

Once a component is proven in the “real” world by a “system” company then the component guys can “kaizen” it and segment it up for mass consumption.

Component guys (like me) often mention that developing components is expensive. But we often play down the point that a box companies revenue opportunity is 5-10x that of the component vendor. Also, the system companies component can be “leaner” because they only need to focus on their feature set, and not the needs of the many. The good news for component guys is that, the surviving box startups of 5+ years have all spent $300M+ which has translated into 1 significant chip development and loads, loads, and loads of software. I love loads of software 🙂

7 thoughts on “Infinera — new component paradigms start in systems

  1. What do you think is so unique about the underlying component? I would think that Luxtera would have a far more valuable piece of real estate as it is in CMOS. I know distance and yields are issues. But what exactly is unique about the components? Could someone wrap their software around a long reach CMOS PIC and do the exact same thing, for far less?

  2. * What is unique about Infinera is that they started down the PIC path 6.5 years ago and had a vision for the problems it would solve in real networks. (Drew has lived thru ATM, LightEra, and Infinera … he’s seen and solved real problems many times )
    * They developed an ecosystem around this PIC to prove its worth in the real world of networks.
    * This means that someone with the correct vision, execution, could take the next generation of PICs and improve upon Infinera. ( I’d say this is highly unlikely … it’s more likely that a company like Infinera would leverage these components for their basic box and extend it with in-house developments)

    * other examples are L2/L3 ethernet devices that were developed for GigE Switches as a means of avoiding Cisco and solving some real network problems. These devices were what made Foundry, Extreme and others. There is no question that the software was 4 – 5x the development effort of the devices, but the devices were the platform that everything else sat on top of. No one did that before and now Broadcom sells much better devices which lots of people wrap software around. But Foundry and Extreme still exist and compete on other dimensions. The original L2/L3 piece is table stakes.

    * Another current example — 2-Wire developed the integrated Residential Gateway ( modem + router + telco_stuff) which component guys like BRCM, Ti, PMCS, and others are now competing for. But 2-wire remains and will likely use these devices so that it can extend its product line.

  3. Hi Iain

    Thanks for the reply. But and it is a big BUT. Luxtera started down the PIC path in late 2001. So this vision thing is overstated. If you look at it the Infinera vision it was actually techynically much easier than the Luxtera vision purely becauae of the material choosen.

    If there are a whole lot of box vendors that have written and tested teh s/w that need the functionality (size, pwer etc) of a PIC then why can they not leverage Luxtera’s work to complete a Infinera competitor in 12 months or so?

    At a 30K foot view that seems fairly obvious to me and what would be scary for Infinera is they have the anchor of a InP fab around their neck

    Thanks again

  4. thanks for the comment — I’ll have to comment later .. I’m supposed to be on holiday 😉
    ** briefly
    * my rule of thumb for telco boxes has been that s/w development effort is a minimum of 4x the hardware development effort.
    * thus the competitive advantage of hardware is over-stated.

  5. Hi Dave,
    Without being glib – At 30K view — You’re missing the value/complexity/defensibility of software.
    — Being a chip guy I frequently discount this aspect of software.

    In infinera’s case — I believe it is easy to “pitch” the value of hardware because it’s simple, but the real value they are building is in their software.

    Here are some of my rules of thumb on Software

    * Software Development > 4x Hardware Development Effort. ( ie s/w is a bigger hurdle than most h/w guys perceive )
    – 4x is the minimum … FFIV-like boxes are even more software intensive
    * Software is what the box customer engages with ( not the hardware )
    – thus software captures what the “end-customer” actually values.
    — a good example is NET Logic who make content processors
    —cool stuff … but they they don’t know how it is “really” used.
    — there is big value is in figuring how to really use it

    * Software complexity and is what allows the “off the shelf” semi/component business to survive.
    – ( semi’s are an enabler … not the product )
    – look at the iPod — standard silicon and customer software
    – look at FFIV boxes — standard silicon and customer software
    – low/mid-range Cisco Routers – standard silicon and customer software.
    – exception — high-end Cisco Routers – custom everything … ’cause they can.

    – optics modules have been pummeled in value by “software” vendors ( ie router vendors like cisco )
    — Andrew@Nyquist is always lamenting this

    ** so I guess I agree with you … why are they continuing to develop their own PIC?

Comments are closed.

Up ↑

%d bloggers like this: