By owning the silicon I mean owning the immediate design change and control capability and technology in order to lock a certain design down and get it within a usable ASIC in fairly short order.
For example, Emulex contracts with ServerEngines for the design and development of its Universal CNA (UCNA). Emulex is basically porting its Fibre Channel software stack to ServerEngine's silicon. While QLogic, and others, are taking a more traditional approach by either moving into building and owning the silicon upfront, or acquiring companies that can immediately provide them with the ownership of key intellectual property.
I noted in QLogic's recent earnings announcement that their CEO, H.K. Desai, was quoted as saying, "I think there's going to be some consolidation, and we want to be one of the suppliers. We believe if you don't have your own silicon you should be out of the business. I look back at iSCSI three years ago, and everybody who didn't have their own silicon did not survive." This is more than a causal observance, the company consolidation is real and Mr. Desai is clearly trying to take the long view as he looks ahead at the industry's ongoing shift toward network convergence and consolidation in the data center and cloud. QLogic has its own silicon now for both converged network adapters (for Fibre Channel and Ethernet) and for data center class 10GbE LOMs -- two critical pieces of silicon for convergence. In my opinion, Emulex is missing silicon ownership and control in both of these areas. For its single chip silicon for its CNAs, QLogic was able to leverage the Ethernet-based storage architecture expertise it gained from acquiring iSCSI specialist Little Mountain Group (LMG) in 2001. And with its newly acquired NetXen assets, the company now has the silicon to port its FC software stack to produce converged network on motherboards (CNOMs). Recent design wins with HP and IBM with NetXen technology will position the company well for further integration of its FCoE convergence technologies. Don't get me wrong here, I'm not saying that if a company doesn't own the silicon they will go down in a heap of flames -- I just believe from my experience that time-to-market will be significantly more challenging, and having the opportunities to have products selected in a product, or data center, design phase may be fewer and farther in between. Clearly, the network convergence market is on fire now with many eyes focused on the big name players. The subtle differences between them may make all the difference in the end.