Friday, March 16, 2012

Big money for photonics in data centers

The best news that photonics people could hear came last week at OFC when Cisco announced that it was buying Lightwire for $271million in cash. Lightwire is a start-up making integrated photonics, and Cisco is interested in it for making interconnects in data centers, among other things. I try not to get intoxicated with financial ups and downs that ultimately benefit only a few investors (if that),and I'm not fond of buzzwords like "integrated photonics," but this is good news for anyone with a similar technology.


It makes sense that Cisco needs the expertise developed in Lightwire. Companies have spent millions on this; for Cisco to do it itself would take millions more and years of delays. Meanwhile, its competitors--Huawei for one--are encroaching on Cisco's market with technology of its own. Cisco can't rely on the merchant market for everything.


The data center bottleneck is particularly important. I attended four discussions on the topic, including the OIDA workshop to develop a roadmap (where I was a moderator and am writing the report). The challenge is for the industry to develop new architectures and inexpensive components that can address the many-to-many interconnects necessary in modern data centers, such as those of Google and Facebook. Traditional data centers are not a challenge: a conventional switching hierarchy can store and retrieve data, and that scales predictably. The new data centers don't scale as well.


Both Google and Facebook made the rounds at OFC, and both claim that the technology is available today, it just has to be commercialized. They say there needs to be a whole new sector of components that don't need to meet Telcordia and NEBS standards. It just has to be good enough for the controlled data center environment. Oh, and make it really really cheap, thank you.


The one problem I have will all of this is that the net margins for Google and Facebook are about 25% or so. That's net profit, not gross. The optical components suppliers' net margins are a few percent to negative. So how about giving some of that nice margin back to the components suppliers? Especially as many suppliers don't see the return on this new segment justifying the risk.


That's why Cisco's acquisition is such good news. It's a return for the start-up's investors, but it also means that Cisco is willing to fork over some real money for components.


The OIDA roadmap report on data center interconnects will be coming out sometime in the coming weeks. Look for it at the OIDA web site or here.

No comments:

Post a Comment