In it’s New Year’s Eve email blast, SdxCentral’s tag subject was “Cisco’s eventful 2024”. I don’t know if it was intentional or not, but the phrase connects to a historical-philosophical concept “The Hero in History” by Sidney Hook. Hook differentiated between people who made things happen (“event-making”) and those things happened to (“eventful”), and I think that pretty well characterizes Cisco’s 2025 challenge.
I’ve worked with Cisco and Cisco customers through a lot of my career as a network consultant/analyst, and one thing Cisco insiders always worked to make clear to me was that Cisco wasn’t interested in being a leader in technology advance, but a “fast follower”. That, said my contacts, was because a major incumbent in any tech area has too much to risk exercising game-changing behavior. Wait till others have taken the risks and found the right path, then jump in through acquisition or simple exercise of market power, when it’s clear that the technology shift will succeed. I think Cisco has succeeded through that approach for decades.
But what happens when nobody leads, or when none of the leads play out? What happens if what the market needs is a major transformation, something that only a giant could really present to buyers? What happens if whoever does lead manages to keep the lead? Those are the questions Cisco has to answer in 2025, because the technology focuses it says it believes in for the future are defensive, speculative. Does Cisco even believe in them, or are they just washing with the tech-soap-de-jure to make Wall Street happy?
Networking a la Cisco had it easy for the first couple of decades after its founding in 1984. Enterprises at the time tended to adopt network architectures presented by their computer vendors, for the simple reason that the goal of networks was to connect workers to applications. The problem was that the “minicomputer revolution” of the 1970s and 1980s was breaking the lock that giant IBM had on IT, and proprietary network technology was a potential barrier to exploiting the new IT freedom. Most of us forget that Cisco’s first product was a NIC for DEC minicomputers, and much of Cisco’s early success was due to the fact that IBM’s System Network Architecture (SNA) was an expensive and “closed” model of networking, not well-suited to the more open Internet-centric model that included customers, prospects, partners, and so forth in the information processing chain. We had applications, and we had information. What was needed then was populist connectivity.
What is needed now? That’s the question Cisco needs answered in 2025. Past history says they’ll wait for someone else to answer it, then sweep in with power and majesty to take over. Past history, though, may be too far in the past, and too broadly rooted.
The proverbial Dutch boy who stuck his finger in a hole in the dike was “eventful”, and the outcome of his behavior depended on faults developing in the dike system manifesting themselves in a local, boy-finger-sized, hole. One of my friends from Amsterdam commented “What would we have called the boy if there had been a widespread dike collapse instead? Drowned.” We have to look at Cisco’s situation in 2025 in this light. When does security or AI, the dual touchstones of Cisco magic in 2025, create an opportunity like the DEC NIC card?
Internet security stinks, and you get all the proof you need of that point when you look at tech news (“Major Incident: China-backed hackers breached US Treasury workstations”) or at enterprise budgets, where for many security spending is beating all the rest of network capex combined. IBM’s SNA was highly secure, but the model wasn’t economically viable for open network use. Could Cisco be looking for a populist security model?
Maybe, but not in the near term. Nothing Cisco has done or acquired is really more than another layer of protection added to a cake with too many layers already. Some fundamental change is needed. Cisco may see this, and may realize that if they’re to seize it when it emerges, they’ll need to be credible security incumbents. So they play the traditional security game in the traditional game. Or, they may not see any good solution to security in the long term, but see short-term dollar signs in those layers. Security may be stagnating, commoditizing, but the rest of Cisco’s portfolio is doing both, faster.
I don’t think that there’s a security silver bullet for networking, simply because of the populist roots of networking today. When IBM SNA was king, a big enterprise network had about 85 thousand users. Today, it might well have a thousand times that. The total number of connected users /devices was roughly 50 million, when today there are billions on the Internet. Cisco didn’t displace SNA, it carried it on IP, because even displacing 50 million SNA devices was totally unreasonable. Try displacing billions of devices. No radical change to the Internet as we know it is economically possible.
How about AI? Everyone knows that AI is a major driver—nay, the major driver—of network traffic growth, right? Well, everyone says that, but do they know it? Enterprises tell me that they don’t see any significant impact of AI on network traffic except within the AI cluster or during training. Still, what Cisco rival isn’t touting AI as a source of new network demand? Cisco would be crazy not to do the same, but crazy if they mistake shared delusion for truth. Do they see something else?
What is network traffic? It’s data. Networks move data, between sources, storage and process points, and users. Data is traffic, and the value of data is the driver of business cases, the “R” in “ROI”. If you can get more data, and more data value, you can justify more network spending to move it around. If AI could create data and data value, even indirectly, it could drive network opportunity upward.
Can human intelligence create data? Yes, but it’s important to recognize the difference between creating data and getting or delivering answers. In the movie “2001: A Space Odyssey”, HAL is consumed by the command to calculate, to the last digit, the value of Pi. Since that’s a transcendental number, delivering the result would literally take forever and an infinite number of digits, but that level of precision isn’t needed in the real world. What’s actually needed in terms of data to perform the calculation is minimal, and yet in the real world the truth is that complex answers, meaning complex problems, are typically framed by complex data sources. AI is valuable in data creation if it can create useful business answers from complex data, and if that data is “new” then both getting the data to AI and getting useful results to decision-makers could generate our traffic.
We actually have, in our history of IT, some insight into where that data probably comes from. In the first age of IT, businesses recorded the commercial results of their activity via computers. You sold or bought something, and punched the record of the transaction onto an 80-column card, which the computer would read and transfer to magnetic tape. Applications analyzed the tape to analyze business operations, but the view they had of “operation” was limited to the financial/commercial impacts. In the second age, we did online transaction processing, and that brought IT into the processes more directly. The PC and the Internet took that process integration further, and by that we mean that IT entered into the real world of what we did and how we did it, not just the world of the outcomes of what we did. The ultimate step in this is to have IT look out into the real world and capture what’s there.
IoT sensors of various kinds, and real-time video analysis, could already allow us to analyze movements, activities, risks, and opportunities in the real world. We’re starting, via things like digital twins, to model systems in a way that facilitates simulation, monitoring, and control. Bringing humans into this, ideally using video analysis, could allow applications to manage even complex systems of workers and their supporting technology elements, in the real world. We have already started to apply this approach in some industries, such as health care, and it’s likely that AI/ML, perhaps particularly in the form of task-specific agents linked through digital-twin-like models, could expand this considerably.
Could this sort of thing be what Cisco hoped for from AI, a salvation of network spending created by massive fusions of AI and IoT to link applications better with the real world? Even if Cisco were inclined to abandon its fast-follower strategy, could they drive this sort of thing? In a way, Cisco is like its telecom customers, trapped in a culture. So if they don’t drive the change, can they fast-follow it? It depends on who leads it, and how fast they go.
The biggest threats to Cisco in 2025 come from Broadcom and HPE, because the direction these two would likely take would make fast-following difficult. Broadcom’s logical goal would be to develop a new application set and commoditize networking it, so chips would win. HPE’s logical goal would be to create, via its Juniper acquisition, the perfect network companion to that new application set, thus locking up the network side before Cisco can mount a pursuit.
Another threat pairing is Ericsson and Nokia. Both companies are very interested in IoT, particularly interested in massive and even public deployments, because of the link to 5G. Both companies have been pushing IoT-related applications through sales and marketing channels, and both are directing their initiatives to network operators, telcos. While Cisco also sells to telcos, its AI and security initiatives are more directed at enterprises.
All of this, IMHO, means that Cisco may have to take a different path to growth than the one it seems to be touting now, the one that’s worked for it in the past. Following others to dominance relies on overtaking them, a challenge in itself with large players like the ones Cisco is now competing with. Whether Cisco has something in mind already is something I can’t claim to know, but I’m pretty confident that if they don’t, there will be other reorgs in 2025.
