Over the last week, those individuals following our second blog read that we were at least inclined to dismiss several narratives making the rounds, including: 1) Amazon’s purchase of Whole Foods (WF) being akin to the telecom bubble; 2) other suitors making counteroffers for WF; 3) Wal-Mart’s purported unreasonableness in having outside tech firms using Amazon’s cloud service; and 4) Verizon really liking its new strategy of being a landlord. All of our responses to these four matters are related to Amazon’s growing retail position, including its need to branch out to physical locations, which will embrace new technology solutions (including sensors) to significantly lower the costs of such operations. In addition, they are all connected to how Amazon’s competitive pressure has been instrumental in creating a commercial real estate bubble, as retail stores increasingly shut down, resulting in cheaper expansion when it comes to bricks and mortar for itself. One of the subjects that may be getting inadequate attention in the press and in financial circles is the impact of these new developments on Amazon (already one of the biggest hyperscale operators) in the Data Center Interconnect (DCI) space.
Certainly, the requirements of the AWS cloud operation have been the dominant factor for Amazon, when it has come to either leasing or building its optical network, whether it be for the long haul, regional or metro portions. In the past, we are aware of a least one occasion in which the company released an RFP without necessarily the intention of purchasing new equipment, but it seemed to be a means to drive its incumbent vendors to lower price points, as well as determine whether there was a greater justification to switch in a more substantial way toward construction.
Interestingly, while Facebook has tended to use its own network in the long haul and lease from the co-location facilities for shorter haul traffic, Amazon has been inclined to do it in the opposite way. This dichotomy can be partially explained by Amazon’s strategy of clustering data centers together in a particular geographic region.
So, even in the metro portion, the minimum amount of capability for Amazon (at least in some locations) is at 100G. It has bought gear from Cisco Systems and ADVA Optical Networking for metro applications.
The WF purchase may already be having an effect on intra-data center capacity at Amazon. A note from a financial institution was reported yesterday that said, “…Amazon/AWS has partnered with Fabrinet to produce 100G CWDM-4 optical modules based on a reference design from (Macom unit) FiBest and based on Macom components.” (Such a win makes Macom’s over-the-top rhetoric appear even more unnecessary.)
Obviously, the eventual shift of WF from Microsoft Azure to AWS will result in a noticeable increase in bandwidth demand for the latter. The number of Amazon’s distribution centers in the US will rise by about 350 percent along with the addition of over 460 stores (the vast majority in the States).
With Amazon’s “last mile” grocery retail plan, which is about physically getting closer to the customer, especially in order to provide quicker and more specialized fulfillment, we would logically expect that Data Center Interconnect (DCI) vendors will continue to benefit in the metro segment. Also, the loads on Amazon’s data centers just on inventory management should climb dramatically, as WF has been well-known as a laggard in digitizing all types of functionality in its stores. In addition, just the sheer amount of new data to be processed on customers regarding their preferences for both internal use and for the suppliers of products will be impressive.
On a side note, in contrast to Google, Microsoft, and Facebook, we find it refreshing that Amazon does not feel the need to have a bunch of engineers essentially dedicated to public relations efforts, heavily pushing new ideas on optical products, especially ones that they are quite doubtful will work out, so that they have the widest array of options available to them in the future. This kind of activity, whether it is just confiscatory fees to join a standards-setting organization, is just another example of the lack of sufficient regard by these large users for the long-term survivability of the optical ecosystem.
[written by Mark Lutkowitz]
SHARE THIS