All markets force companies to make choices: value proposition, how will it be differentiated, does anyone care, how does each investment support the intended value chain/value proposition. A disaggregated Internet forces existing companies to make new choices and creates the opportunity for new ones to disrupt.
The emergence of Broadcom, and companies like Arista that use Broadcom, has led to an ongoing strategic conversation at network equipment companies for many years now. The conversation is especially poignant to those network equipment companies that differentiate in the market with their own silicon, including Cisco, Juniper, and Nokia. Some of these companies have been straddling merchant silicon for some products and their own silicon for other products. It is easy to get into a strategic straddle when going down that path: creating internal confusion about what the company’s value proposition is, and not executing it well on either side of the straddle. In the vernacular of American Football, what is the company’s “identity”.
It is seductive to think you can use your own silicon in one product, and merchant silicon in another. Not every company has the same relationship with Broadcom as Arista does, especially those companies that send mixed signals to Broadcom by developing their own silicon. For those companies that do develop their own silicon, there is that uncomfortable feeling that every time they spend money on Broadcom, they are funding the war chest of a company that is fundamentally opposed to what they do. Rule number one in war, don’t do your enemies any favors. For a company with the heritage and pride of Cisco, let’s multiply that a few times.
The bigger problem though is value chain and value proposition. Compelling value propositions don’t just pop out of nowhere, they are not solely the result of a single product manager with a game-changing product requirements document. Compelling value propositions are the result of aligning the value chain with a value proposition, and when companies have a mix of business models, then the value chain can become mixed as well. Strategic straddle. Is the value in the software, hardware, or services, and how does the culture, value chain, processes, and incentives of the company systematically contribute to the intended value proposition?
Broadcom forces companies like Cisco, Juniper, and Nokia to examine if they want to continue to develop their own silicon. In the cloud era, the issue is volume, the same issue that has impacted the optical transceiver business. Hyperscalers, and some SPs, are individually so large, that their purchases can be of significant enough volume to achieve pricing below what network equipment suppliers can or want to provide components/systems at. This is explicit in the optical transceiver pluggable market, and a little more subtle in the network processor market. The proprietary silicon of a network equipment supplier is limited to the volume of systems sold by that supplier. Broadcom’s volume is not. While Cisco is certainly a high-volume operation, they have to have one eye on Broadcom’s volume and another eye on who is going to lead the industry conversation about what silicon should be. These questions are a little more existential for lower-volume players, but Cisco would not be exempt from such questions.
In a routing TAM that has been flat to declining over nearly a decade, internal R&D efficiencies have also become important. Network equipment companies can no longer afford to maintain hundreds of software trains or port new software to endless variations of silicon. Few people outside of a network equipment supplier probably appreciate how much cost is sometimes involved in supporting new silicon, even if there is a hardware abstraction layer (HAL) in place, something more present in modern designs than legacy ones. Magical when internal drivers meet external demand: simplify network architecture, simplify network equipment, reduce the amount of code that has to be tested and could fail, and shift some of the heavy-lifting to controllers, who can leverage compute clusters, huge databases, and emerging analytical techniques, artificial intelligence, and machine learning.
In a world where silicon, software, and optics are disaggregated, technology suppliers will need to think carefully about what their value proposition is and ensure their culture, value chain, and incentives are aligned to it. Even if Cisco has now provided some strategic clarity as to how they are going to respond to industry forces and the emerging influence of hyperscalers, Cisco is not exempt from examining this as an ongoing strategic issue. Opportunities for strategic straddle still exist, and how far Cisco goes in transforming itself in new directions will be fascinating to watch and telling in terms of outcomes. The strategic trap Cisco will have to navigate is whether they can sell components to only a few hyperscalers, or whether they have to sell broadly to all buyers, and if the latter is true, what will it mean for Cisco’s value chain.
In terms of how good Cisco’s new silicon is, I’ve watched a couple of silicon announcements over the years, and it is usually best to wait until the details come out before making sweeping conclusions. Cisco has created interesting positioning though, one definitely worthy of industry discussion. Is there really a magical silicon architecture that navigates all packet processing and fabric switching tradeoffs with ease? Is the ease of navigation across all use cases for the chip: SOC for a pizza box, line card NPU+fabric, and fabric? I look forward to learning more detail about the Q100 and successive chips. I look forward to the industry conversation in general, it should be interesting.
The volume disruption created by hyperscalers and the need for significant R&D efficiencies driven by hyperscaler price pressure and TAM declines has combined to create fertile ground for disaggregation. Never a dull moment in IT. Never a dull moment in networking. Game on.