As we will demonstrate on an OFC 2015 panel, the data center VCSEL market will remain healthy for an extremely long time, especially in terms of steady revenues. The most shocking change is based on recent evidence that Avago Technologies is selling VCSEL components, possibly including as far down the food chain as bare die, to at least one interconnect solutions provider and at the request of a minimum of one very large enterprise. In even reading between the lines of a small portion of the transcript of the latest quarterly report from the vendor, it indicates a move in this direction: “…[W]hen you start talking about 25[G], you’re talking about having to drive short-reach or even longer reach native 25, which makes it very tough to produce laser, VCSELs, laser, so to speak, that does 25 gigabit. We are one of the few guys who can do it. And we happily sold it to guys who do 100 gigabit Infiniband today.”
The reason that such a move by Avago is so stunning is that there has been speculation for about a decade and right up to the present (despite the buyout of CyOptics in 2013) that the firm would exit the optical space entirely, and now for apparently the first time, it is evidently engaging in supplying devices further downstream with even less potential for appealing results for investors. Perhaps with the recent trend of the huge data center users looking to design their own optical hardware (like transceivers) to achieve the lowest cost possible, including the Google-InnoLight partnership, Avago figures it is better to get something rather than nothing.
While the past characterization of the VCSEL transceiver business as a duopoly between Finisar and Avago may have been a bit of a mischaracterization, it has increasingly become more accurate as a description. To whatever extent JDSU still sells VCSELs, our expectation continues to be that Finisar will acquire CCOP. JDSU’s short-reach operation has been evidently folding up with the departure of senior management involved with new product development for 850nm devices, and it will in effect have less overlap with Finisar.
There is also reason to anticipate that Sumitomo will eventually exit the VCSEL business mainly because of an inherent preference to play in higher margin arenas. In 2012, the vendor paid $17 million for Emcore’s VCSEL solutions.
Especially at the lower data rates, there are some Chinese vendors of VCSELs. At times, there are questions involving reliability, as they are newcomers doing R&D in this area. In addition, there are signs that the Chinese government apparently has come to the conclusion that in general, it has become overextended with its investments, as too many companies in the country are not making any money.
There are at least a couple of European participants as well. Nevertheless, it appears that their participation in the communications market is fairly negligible.
Obviously, with the prospects of less competition, volumes on VCSELs for both of the leaders would increase dramatically, which is greatly desirable in a low-margin situation. In particular, the high fixed costs in the production of this gear make volume the critical factor.
Yet, even with a duopoly position, these suppliers need to be careful about any future plans to increase their prices in any meaningful way on VCSELs because of two main aspects:
- nothing else is more associated in data communications with “cheap” than these products and it could open the door to additional competition;
- the need to remain competitive with single-mode fiber alternatives (the biggest advantage for VCSELs is the reluctance of the vast majority of data centers to change out their multimode fiber).
[written by Mark Lutkowitz]
(For further elaboration on Avago, please click here. To read Cisco: Buyer of Potential Finisar-JDSU Combo,? please click here. For insights on Emcore’s sale of its narrow linewidth tunable laser product line to Neophotonics, please click here. To view Google and Cars: Same Revenue Stream, please click here.)
SHARE THIS