40GbE Modules Expand their Success as ‘Hyperscale’ Becomes Mainstrea
While IT spending continues to be soft, enterprises are looking to outsource what they can to cloud datacenters. eCommerce and Web 2.0 content providers continue their exponential growth and all of these players are building scale-out datacenters. To illustrate that point, the Internet companies tracked by LightCounting spent 19% more in Q1-2014 than in the year-earlier period.
This means large datacenters are taking an increasing share of the Ethernet optical module market, and consuming most of the 40G and 100G Ethernet modules. Their flatter interconnect architectures or ‘fabrics’ now being implemented require lots of cost-sensitive links. Last year, LightCounting doubled its forecast for 40GbE-SR modules as early consumption really took off, though mostly as a 4x10GbE density- play with optical breakouts. The most recent market data confirms this trend, and leads us to increase our forecast even further. LightCounting now forecasts 40GbE module revenue passing $0.75 billion by 2019.
So with 40GbE modules taking off, is 100GbE ready for its ramp? Not quite. While 100GbE gets most of the news today, native 40GbE is just getting started and has almost all of its life cycle still ahead of it. There are even discussions of 40GbE serial modules, perhaps repeating the sequence of 10GbE, and following the same path to success. For some companies (i.e. Microsoft), 40GbE is not enough as they expand by 100,000 new servers per month. They want 100GbE as soon as possible. At Interop in Las Vegas, 12-port 100GbE QSFP28 line cards were on display. Still, many large datacenters are just moving into 40GbE leaving any serious market volume of 100GbE years away. In the meantime, higher module prices are yielding good business and we project 100GbE module revenues to pass $500 million by 2017.
Innovation and Confusion Combine at 100GbE
In 2013 everyone was hoping for an IEEE 802.3 standard to address the “reach gap” at 100GbE for next generation hyperscale datacenters. Short reach has shrunk to 100m and 100GbE-LR is just too expensive when you only need to go 500m or perhaps 2km. But, hope faded when consensus was not reached within the IEEE.
LightCounting presumed that industry would get things done with an MSA. But much of the innovation offered at the IEEE Task Force is now heading to market as four MSAs. And don’t forget there are five different form factors in use or planned for 100GbE. Can they all be successful? Will that accelerate or delay the adoption of 100GbE? Such are the questions product planners and forecasters must address.
In the end, market hype needs to be separated from reality.
