The most important pieces of network infrastructure are access and metro. There are three major access questions in that space that 2026 may answer. First is how far fiber can go. Second is the impact of FWA, and finally there’s the impact of satellite. In the metro area, there’s really only one question, but all the access answers could impact metro as well.
Access technology is the most expensive piece of network infrastructure, largely due to the fact that you don’t get economies of scale from it the way you do for deeper pieces of the network. But this lack of potential cost efficiency doesn’t mean you can dismiss it; access technology is usually what limits the speed of the link between users and what they want to use, and so it’s important from a competitive perspective. Finally, even though that’s true, users are reluctant to pay for capacity, and so they want more for less, which gets us back to cost.
Fiber is usually delivered to homes through passive optical networks (PON), which creates a kind of tree feed from the head end to anywhere from a half-dozen (or, sometimes, less) to over a hundred endpoints. PON is great for creating efficient access where the endpoints are fairly concentrated, but as they thin out and the length of the branches increases to reach enough users, the economics suffers. Urban or suburban areas are generally good PON targets, but rural locations far less so. Even where PON works, it’s still physical media, which means you have to install fiber cables and set up termination terminals in customer sites. CATV is the same, but can often serve a larger number of “drops” or endpoints on a span.
Early fiber deployments tended to stay within commute distance from major metro centers, areas where suburban community populations almost run into each other. In rural areas, the towns are smaller and more separated, and today’s fiber focus has been in addressing these areas, which are simply small pockets of suitable endpoints rather than larger ones.
This is enabled by what’s now popularly called “convergence”, meaning the convergence of mobile and wireline. Cell towers now need fiber feeds, and people expect to be able to use cellular broadband even when they travel, so the towers are distributed over wide areas where wireline fiber would not likely offer a good ROI. If the tower and a PON head-end are combined to create a feed point, you might well be able to serve smaller communities with PON. Telcos, and mobile infrastructure players like American Tower, are all exploring how far the concept can be taken profitably.
FWA can also piggyback on cellular tower points, and also could be hybridized with PON. You could feed cellular, FWA, and PON from a common converged head-end point, but you could also drop FWA head-ends on PON to create a more economical way of extending FWA closer to users, increasing its performance and reliability.
Besides creating a bigger fiber footprint, this all plays into the “metro” strategy. The reason is that the more you do at a fiber head-end, the more you may be able to do there, profitably, down the line, because you have more user traffic concentrated at that point. You might think that this is more likely to generate an “edge computing” opportunity than a metro one, but telco experts I’ve chatted with disagree, and what they say makes sense to me.
Edge computing at tower sites might, just might, eventually create some opportunities, but not in the next three years, experts say. What “convergence” does create at tower sites is feature-hosting opportunity, and in particular opportunity to host service analysis for customer care. This, say at least some telco experts, is the sort of thing that Open RAN missed (see my blog from yesterday); if you’re going to have a “RAN intelligent controller” that has generalized resource management capability, for both non- and near-real-time stuff, you need to think about just what sort of stuff you might want to host and manage, beyond RAN-specific elements.
The simple truth is that 5G and 6G advantages, when you come down to the basics, all depend on real-time applications. Yes, these applications need network support if they’re not run local to the real-time elements, but before you worry about that you have to decide where they are run and how they’re managed along with their resources. As I’ve noted before, nearly all real-time applications are best implemented as a hierarchy of control, with some extreme elements hosted locally and others further out. This approach would facilitate remote hosting by creating a model where network edge hosting resource needs were minimized by distributing the logic, and where latency would still be important but not so demanding as to prohibit remote connections to model components.
The experts still believe that metro hosting is more important, because concentration of potential users is far greater at the metro points, and because more potential applications for hosting tend to exist in cities because of industrial and transportation concentration there. However, they do believe that converged edge hosting would encourage the development of the hierarchical modeling approach that would also then empower metro, and that metro hosting could be an attractive way to address resource congestion further out. This would be particularly true if some 5G/6G latency benefits extended inward to metro points, something operators should consider in promoting those services.
What about satellite? Telcos haven’t offered a lot of comments other than fear of competition, but some experts in telco-land tell me that they believe that texting and calling might shift more and more to satellite unless telcos take consistent steps to offer emergency calling services as a part of their plans. They note that unless this is done, satellite providers would be encouraged to offer emergency connectivity themselves, which would open the door to expanding their mobile services to the point where they’d replace traditional cellular.
The fact that FWA, and even mobile wireless, is growing so fast is a sign that access investment is increasingly challenging in terms of ROI, which of course is based on the fact that basic broadband is pretty much what current applications depend on, and that’s a mature market. The question is whether some form of metro hosting could itself help to redeem access infrastructure cost, and perhaps even validate premium handling of some traffic that could earn incremental revenue. That may be the most important question for telcos, and their vendors, in 2026. Otherwise, competition for users and revenue is a zero-sum game that, like them all, creates only losers.
The key point here is that simply pumping up broadband access capacity isn’t working today, and isn’t likely to work in the future. Access is expensive and willingness to pay is limited. We need something different, if we expect different outcomes, and it’s as simple as that.
