Let’s forget, for a moment, whether telcos have any reasonable path forward with their traditional service offerings. Let’s forget the questions of whether such a position is even rational, or why it’s held by telcos in the first place. Let’s focus on the raw question of what telcos think they need to do with regard to infrastructure and technology evolution. I chat with 88 telcos globally, and I’ve worked to analyze their comments over the last 18 months. At the highest level, telcos present five main priorities, all with weights within the same range of support, in infrastructure/technology evolution. Here’s what I see on each.
By a slight margin, the first on the priority list is seek infrastructure harmony. Operators say that their networks are made up of too many different devices. First and foremost, that means a much larger spares inventory is needed, particularly if they cover a large geography and can’t get access to a central parts depot from remote parts of their service area. Next, it means needing specialty training in too many different devices, and often diverse operations tools and practices.
One result of this priority is a reduced interest in openness, if we define it as support of multi-vendor infrastructure. That particular goal has been denigrated, in favor of a goal of simply offering vendor alternatives from which a small number (preferably one) will be selected. While so far no telco says they have achieved single-vendor status, many say that if you divide networks into functional layers (RAN and core, for example), they have a single vendor per layer.
Then we have a seeming contradiction, embrace functional layer diversity but manage device count. Operators universally believe that their networks are made up of three or even four layers, and they expect that devices in each layer will have features needed to support their specific-layer missions. By implication, that means that they’d expect little or no device-sharing across layer boundaries. But at the same time, they acknowledge that the fewer devices there are, the better, because device number increases mean operations complexity increases too.
The issue here seems to be that the access portion of a network is very different from the core, and most operators believe that they need specialist vendors for the former and traditional “router” vendors for the latter. In each layer, operator desires to avoid finger-pointing and integration issues align to discourage slopping things over layer boundaries. So unity within layers, isolation between them.
Our next priority is apply virtualization techniques to infrastructure. Here, we have a kind of wishing in play, because while operators seem to think that virtualization in their infrastructure could be a tool to manage costs, as it is in computing, they’re imprecise with regard to exactly how that would work at the network level.
Virtual network strategies may be a bit of a hostage to the layer issues just covered. We have virtualization in the IP layer of the network, obviously. We also have it, via slicing, in the RAN, and RAN infrastructure uses virtual-network tunnels to manage mobility. Given the operator conviction that one layer’s vendor can’t handle other layers optimally, any confident virtualization of the network overall remains a divided effort. Even SD-WAN is a kind of service overlay, applied within the IP piece.
Hosting virtualization? Here, operators tend to think about virtualization a la containers, via NFV, or based on cloud technology. In particular, they don’t tend to think about any network-infrastructure-specific sort of virtualization, but their goals to me seem to imply that’s at least a part of what they’d like.
The fourth priority is support AI deployment. This could likely be generalized a bit to “support hosted elements of services”, but right now AI is the particular focus because operators (like enterprises) believe that AI hosting is different from traditional hosting, more demanding and requiring more meticulous planning. However, I’ve already noted in the virtualization priority notes that hosting has technical questions associated with it overall, and AI complicates that not only with its own uncertainties, but because its application is linked both to offering AI services (a stretch for many telcos) and to OSS/BSS, which has been a priority minefield for decades, as we’ll see below.
The big specific issue with AI deployment seems to be the question of LLM/chatbot versus agent. Most of the operators who see AI-as-a-service opportunities see the former, while most who see agents as the goal also see the applications of AI as internal, specifically related to operations. Obviously, there are big business-model differences here, but it’s also true that AI agents in operations missions are distributed rather than cluster-concentrated, and that’s particularly true for operators whose infrastructure is inherently highly distributed geographically.
The final priority? Modernize both netops and OSS/BSS. I can tell you that operators have had this issue on their list for at least two decades, and that most would agree that they’ve made little progress to date. A big part of the problem, say operators themselves, is the traditional separation of OSS/BSS into the office of the CIO, which is a direct report to the CEO, as are the CTO and network operations (COO) teams. About two thirds of all the operators I’ve chatted with, worked with on standards, or consulted with say that this division results in a combination of rivalry and division of responsibility that hampers efficient operations, but like much of the historical issues in telecom, it seems intractable.
Who or what can offer this? There’s no strong consensus among the 88 operators, but the vendor most often mentioned when any vendor is cited is DriveNets, whose “network cloud” cluster model of device composition was based in part on a telco initiative (AT&T, to be specific) to harmonize infrastructure. What’s particularly interesting is that every operator who mentions DriveNets has at least deployed them in a trial, and some have committed to them in production. None of these mentions were unfavorable, which is good. That they haven’t swept the market, and don’t automatically come to mind even for operators who’ve not committed to them to some degree, shows the challenge vendors face in meeting all those priority goals.
I’ve cited the telco priorities most often mentioned, but not gone into the budgets committed to them. Why not? Because there’s no significant commitment of incremental budget noted for any, at least not in 2026. Operators see their priorities as things to consider when they commit the funds they’ve already budgeted, almost all for things like orderly modernization of equipment, completion of 5G rollouts, and other projects that range from APIs to wireline edge enhancement. None of these are extensive enough to impact a big chunk of infrastructure, and without broad impact it’s hard to make meaningful progress on these priorities.
Progress, when it comes, seems to me to be most likely near the edge of the network, from the metro centers where it’s reasonable to expect feature and edge-application hosting outward to the customer. You can’t personalize services in the bowels of the network. But many boxes deep in the network have aged, so there is budget for some replacement there. Enough to actually transform infrastructure? In some cases, but not for the majority.
Here’s the problem telcos face in dealing with their priority issues, and the one their vendors may face as a result. It’s inertia. Stay the course. A network is enormous, and the sunk cost it represents is likewise. It’s very rare for any network operator, and telcos in particular, to be able to fork-lift any significant piece of their network out, because they’d need to write down a lot of cost and incur a lot of new cost. Wall Street would likely hate it. They don’t mind spending on AI, because it supports their bubble-of-the-moment, but do that for old crap like network infrastructure, and they see nothing but an admission of guilt. That’s why new “G” standards are important; they’re a source of hype too, of bubble fuel, but 5G has fallen from grace already and 6G is half-a-decade away.
If you can’t displace a lot of existing gear, then anything new you buy is isolated inside a network largely made up of old stuff. How can that new device be expected to contribute to new revenue or reduce cost in novel ways, as one star in a vast galaxy of depreciating gear? There’s a sad negative feedback pressure at work. Meeting the telco’s own priorities would require major changes in infrastructure. Major changes in infrastructure require major increases in spending, which mandate major improvements in business case, which can be achieved only by a massive increase in revenue or a massive decline in the spending we just said had to increase to fund infrastructure change. No easy answer.
