Article Info

SFP+ … Yet Another 10G Optical MSA?

This article is an anonymous submission from a dedicated reader and frustrated engineer

A decade or more ago life as an “optical guy”…as grad student or a professional, revolved around fairly simple things…moving photons over here, a little “guvment” money over there, a product or two now and then…good times, good times. And if you were lucky enough to score a (gasp!) postdoc (swoon!) at Bell Labs, why ANYONE would work grueling hours for Dickensian low pay for a chance to work on things like… SONET…. DWDM… FIBER… BANDWIDTH… DATACOM…

Economics, ROI, quarterly growth, P/E, investor guidance and things of that sort, well that simply did not matter. For the truly patient investor, the general idea of a “boutique investment” applied well…a fairly complex science understood by the few in service of fewer still.

That was then…fast forward through the “Optical Boom”, through the maniacal billions spent on optical whatnots, misunderstood complexity, through the “Optical Bust” to the present…and what have we? The same boutique market…only now it’s turning ever so surely into a basket case for lack of profit margins…(yes, Dickensian :) )

The latest in this series of adventures that has industry lemmings headed for the cliff once more is yet another 10G optical MSA (multi-source agreement)…SFP+. MSAs were originally conceived as a means of spurring industry consensus around a common optical module standard. The “Average Joe IT-Buyer” could then easily figure out how to plug a fiber into the spanking new router he just bought and get everyone’s productivity inching forward.

The consensus on form, fit and function would presumably spur volume sales of optical components into the routers and get us all back on the road to profitability…”bye bye boutique, hello Wall Street”…so went the thinking. This worked unarguably well in the first years…GBICs and SFPs have shipped in the millions to get Gigabit Ethernet and Fibre Channel networking going, not to mention the lower rates like 10/100Mb/s as well as SONET/SDH and Fiber to the Home telecom.

But wait…a strange thing happened on the way to 10G. The last 5 years alone have seen FIVE different MSAs for 10G…XENPAK, XPAK, X2, XFP and the 300-pin MSA. That with at least three (3) different distance standards (ratified by the venerable IEEE) the “Average Joe IT-Buyer” now has a choice of FIFTEEN ways of figuring out how to connect to the network.

Now, a sixth MSA for 10G is coming along folks… SFP+. What? What happened to the other five? “Too bulky…” quoth one OEM…”Too expensive” quoth another…”Too much power” quoth a third…the list of failed expectations grows…and more money continues to be spent engineering for yet another standard. Designing each of these new modules requires a whole change to the product PCB, bill of materials, components, supply chain, test, marketing/disty cost, etc. It shows up in the bottom line in the likes of the Finisar (FNSR) and JDSU (JDSU) and Avago as well as their Asian brethren (ok ok…maybe the keiretsu don’t really worry about short term profits)

Having six standards to do just one thing (hook up fiber at 10G) puts the current Blue-Ray/HD-DVD format war to shame. Will SFP+ really set the market straight and enable the IT-spender to save a lot of money putting up 10G in his/her network?

What’s worth thinking about here is how much these optical modules cost to make and ultimately what the IT-buyer pays for it. In the case of the now-humble GBIC or SFP for your gigabit LAN, we hear cost figures anywhere in the $15-$20 range…that is the cost to fabricate, test, market and ship the product.

It is somewhat well known these products are sold by the manufacturers to the OEMs with a 25% to 30% gross margin…we hear of such figures from the likes of RHK et. al (r.i.p). Yet we also hear that your “Average Joe IT-Buyer” (let’s call him AJIT) gets to buy it from an OEM somewhere in the neighbourhood of $150 – $200…a heavenly markup north of 500%.. for a fully tested, commodity status and standards compliant product !

Whoa! Nice ! If you’re a large OEM life is good being an “optical guy” !

Yes yes, there is some value added to OEM reselling…but let’s face it… we’re not talking about DWDM long haul here….this is your basic LAN networking gear. It’s essentially distribution, stocking and servicing customers…is that worth these heavenly markups? Buying enterprise optics direct from the manufacturer vs. buying from the OEM… that’s the crux of it.

It’s reasonable to ask if AJIT could just pick up the phone and call direct. I’m sure folks like Fiberxon or Finisar or any aspiring optical company would love to take good care of him…but alas…we hear nifty OEM tricks get in the way…from voided router and switch equipment warranties to software that looks for a specific ID in the optical module and locks the port if the OEM branded version is not plugged in…neat!

Consider this interesting parallel a world away with printers and print cartridges. In a recent Supreme Court ruling in Lexmark v. Static Control Components (SCC) (see for example here and here). Lexmark lost the right to essentially prevent SCC from making a chip that enables use of non-Lexmark print cartridges in Lexmark printers.

One could argue, in similar vein, allowing AJIT direct access to enterprise optics makers would benefit him. After all, why pay 5x more than they need to? These optical module folks know standards…heck they wrote ’em…they know manufacturing, service and distribution…and they know where to invest their profits…to build better optical products that add real value. Six different MSA standards is not adding value, and 25% gross margins don’t leave much left for real R&D.

We think SFP+ has good objectives in increasing density and reducing power. But is is not clear that reducing cost further with yet another MSA going to improve the economic fortunes of the industry. In the meantime, the boffins at Intel (INTC) (and now Luxtera) are hard at work with well-funded projects on “reducing the cost of optics”… by making everything in silicon. Admirable, but it’s clear fundamental problems remain in how this type of product makes its way into the mass market. Until these are tackled, no amount of thumb sucking or blanket-holding is going to help the optical P&L statement.


Comments are disallowed for this post.

  1. * Too many interfaces always sucks! Agree Agree Agree.
    * margin food chain is always interesting … “value” is a slippery thing.
    * don’t think optical (or IC) networking component vendors are gonna better margins with fewer OEMs. It is gonna get worse before it gets better. This is not the PC industry

    Posted by Iain | June 22, 2006, 12:14 PM
  2. Trackbacks / Pingbacks

  3. Quake - Another Failed AMCC Acquisition | August 3, 2006, 10:09 AM