Conversations around SDN are starting to pickup and we’re seeing more tire-kicking and trials from mainstream clients. The bulk of the SDN conversations are around agility (i.e., I want to provide cloud-based services to my constituents but the network provisioning takes a long time). SDN can certainly help here, but there are non-SDN alternatives for this. You can use automation tools for this, and Puppet and Chef come to mind. Perhaps the biggest benefit of SDN is that it fosters long-term innovation in networking. With SDN and de-coupling of hardware/software, you can now innovate independently in hardware and software. So software innovation isn’t held back by hardware and vice-versa. Also, you aren’t beholden to what existing network vendors do within their R&D. In other words, six guys in a Baltimore garage can bring a network feature/application to market without having to raise $35M in funding to wrap sheet metal around it and spin ASICS. Then, they can port the application to run on multiple controller platforms assuming we have standard/open interfaces.
Side Note: this is why there is so much importance associated with the Northbound API initiatives that should come out of OpenDaylight and ONF.