The slow fade of comm

When an end market for FPGA design is in its ascendancy, the benchmarks are there for all to see. Expanding video processing and transcoding opportunities, for example, meant that there were more new FPGA-based designs at last month’s National Association of Broadcasters’ conference than might have been anticipated. When a market is in decline, however, one has to discern as much from silences, or what Donald Rumsfeld might have called “the absence of evidence,” as from positive indicators of new designs.

For 20 years, FPGA vendors have seen the data communication and telecommunication markets as two of their richest end applications to exploit. After the first telecom crash of 2001, core routing and Ethernet switching applications were replaced by deep packet inspection and protocol analysis applications at the network edge. But the overall impact of FPGAs remained strong, in part through FPGAs’ displacement of most of the roles assigned to the highly-touted network processor.

But now, a true sunset for the communication market may be in sight. With the exception of a handful of passive optical network processors, the Optical Fibers in Communications conference this year had little to show in FPGA or even microprocessor/microcontroller-based design. This week, the Interop show opens in Las Vegas with almost a solitary focus on software. The role of hardware network elements is disappearing. I would be surprised if we don’t see the same thing happening at October’s Supercomm show (back to its old name and Chicago venue after years with the "Nxtcomm" logo), since the fortunes of wireline service providers are collapsing almost as quickly as the network infrastructure OEMs that serve them.

Now, a few opportunities remain in 4G Long-Term Evolution platforms in the wireless space, to be sure. We saw Xilinx pursue such opportunities in February at the Barcelona Mobile World Congress. But “soft” programmable base stations cannot sustain a multifaceted industry, any more than base stations alone could sustain a now-bankrupt Nortel Networks.

I’m not throwing in the towel on datacom/telecom. New work on exploring and exploiting higher-layer protocols may yet lead to interesting wrinkles in 40G and 100G Ethernet that prove important for FPGA developers. But the glory days are long gone. Communications has moved from bullseye to backwater in the FPGA space, and if it ever comes back, the market may be owned by Huawei, Ericsson, and only one or two other OEM players – somewhat like the automotive industry in the post-recession world.