The critical reason for Cisco Systems’ elaborate dog and pony show focusing on its latest ASIC, Silicon One, appears to be the simplest one: to indirectly start laying the foundation for its eventual exit from the system space across the board. Although fibeReality has discussed the possibility of such a radical change in the supplier’s corporate strategy previously, shifting market dynamics have apparently forced the supplier to move more quickly towards what it perceives to be the greater appeal of underlying, innovative technology solutions: hence, an even greater focus on the mantra: “chips, optics and software.” However, we continue to assert that the second category will remain by far the least vital to Cisco because it simply does not tend to generate the higher margins of the other two sectors, and it has always represented such a tiny percentage of its overall revenue. In addition, we contend in this article that Cisco’s latest acquisition further supports the notion of speeding up the process in targeting valuable IP strictly at the low end of the food chain.
There have been six major factors over time, which have presently culminated in what we believe will result in a speedier, remarkable alteration in direction for Cisco: 1) a long while ago, it apparently determined that the brand name by itself (“nobody gets fired for choosing Cisco”) would allow it to remain in the systems business indefinitely, and on its own terms, resulting in at least some resting on its laurels; 2) ascendency of the hyperscale operators in the network equipment space -- they successfully pushed for open sourced platforms as well as likely been looking for ways to punish the firm for announcing its intent to corner the high-end, coherent DSP merchant space (we were first in publicly pointing out the negative ramifications); 3) severer price competition, especially from Chinese players and white boxes, in the world marketplace; 4) declines in the service provider segment in recent quarterly reports out of Cisco have become the rule rather than the exception; 5) dramatic shrinking of enterprise sales opportunities as these customers have increasingly outsourced to cloud and colocation entities; and 6) the firm arguably has the worst reputation in networking for vendor lock-in (for example, AT&T has made a concerted effort do everything possible to get away from it, resulting in purchasing well over half less gear from the company, and Verizon, historically, has been inclined to place hard caps on annual expenditures with the firm).
Concerning this last point, we think that Cisco has probably determined that it can more effectively maintain it corporate culture based on “captive” customer relationships from the hardware circuit side. With loads of cash available, it can not only keep on making acquirements of vendors with unique capabilities, but there are other imaginative ways to ensure rigid ties to purchasers. For instance, Xilinx has been able to lock in customers by gaining ownership of sockets by financing the development of software, etc.
Actually, Cisco’s pending purchase of a small startup, Exablaze, would otherwise seem rather limited in having a tremendous impact in just the context of an enhancement to its Nexus product line. Yet, from the perspective of a potential backdoor means of helping to enable Cisco’s entry into the FPGA chip game directly, going after the very attractive financial vertical (along with the latency advantages clearly having utility in supporting 5G wireless networks), offering a fuller package downstream that includes ownership of the processor, would result in a more compelling story.
Ever since Altera was purchased by Intel, Xilinx has undeniably become the dominant force with FPGAs, a situation that we foresaw in the spring of 2017. Cisco could theoretically become a solid, second player, or could conceivably just buy out Xilinx, and gain further diversification in the integrated circuit realm.
While Cisco cannot get out of the system business overnight, it would ultimately divest those portions of the company. Just as when it comes to DSPs, other gear manufacturers will resist buying chips from any system entity.
As we have pointed out in our other blog, the mere fact that Cisco has declared itself a direct competitor to Broadcom will in due course, make it practically impossible to remain at the top of the assembly level for very long, even assuming revenue for both options from the hyperscalers. The idea that the former would find it attractive to be an active participant on both extremes of the value chain makes no sense because it fails to both optimize development costs and maximize profitability, and thus, it would have unprecedented, dangerous implications for its future performance.
For our fast-growing, totally separate, quick update blog, which is exclusively on fibeReality’s LinkedIn page, please follow us here.
As always, fibeReality does not recommend any securities, and this writer does not invest in any companies being analyzed by us.
[written by Mark Lutkowitz]