Spending on telco infrastructure is declining. Telcos don’t know how to find new revenue. Enterprise network spending is also under pressure, and in particular spending on new technologies or on new projects that could lead to higher spending down the line. Open technology is more and more an argument rather than a property, and many of the theories on how to reap its benefits, for telcos or for enterprises, seem to be proving false. What’s going on here? A bunch of stuff, some unrelated to others, but all combining to set us in a new trajectory.
The biggest factor in all of networking is consumerization. Up to the 1990s, the only meaningful data service revenue came from business. The dawn of the worldwide web made the IP-based Internet the primary data service, and forever altered the balance of business versus consumer data revenues. Not only that, the wide deployment of IP infrastructure transformed what businesses used to build their WANs from private digital trunks and routers to IP VPNs.
The reason consumerization is a problem, at least now, starts with the fact that consumers want more for less, period. You can’t get them to pay proportionally for capacity, but that’s also true of business. The problem with consumers is that they really don’t want to pay more at all, and really do want to pay less. Since capacity is the only feature that they recognize for broadband services, that means that operators can’t monetize their only recognized differentiator.
Another problem is the impact of consumer Internet on business services, beyond the initial VPN transformation. Access infrastructure designed for business use, meaning things like carrier Ethernet, can increasingly be supplanted by consumer broadband access technology, which is much cheaper. After all, branch locations of businesses rarely need even consumer levels of bandwidth, and increasingly remote sites are supported through the same Internet-and-cloud front-end that supports consumer/B2B access to applications. Thus, consumer technology is continuing to find new ways to undermine business revenues.
The next factor impacting networking is subordination. Networks deliver stuff, they don’t produce it. That makes their growth inherently dependent on the things that do produce things, meaning applications and content. For a big chunk of the past, the period starting in the 1950s and extending into the 21st century, producers of stuff outran delivery capabilities. That meant pent-up demand for networking, but that pent-up demand has now subsided. We need changes in the “stuff”, in volume, delivery requirements, etc. and those changes are beyond what networking itself can deliver.
5G is a classic demonstration of this. There are many things that you could do with 5G that you can’t do as well with 4G, but most of those things are either examples of minimal-advance enhancements to delivery of things we can already deliver, or new applications or content whose production doesn’t yet exist, and whose value is questionable. Yes, 5G might deliver 4K or 8K video to a phone, but on a small screen could you even recognize the content?
The need to foster production has another dimension, breadth, and it’s easiest to offer an example of this in the business space. Could digital twin technology and IoT combine to empower the 40% of workers who aren’t chained to desks? Yes, probably, but to get to that we’d need new hosting, new software, new sensors. An entire multi-part ecosystem is needed, and networking is useful only when we get it all in place. Network vendors don’t have the ability to move those producer pieces, and everyone is conscious of the fact that their investment in their own role in the ecosystem can’t pay off until others also make their own investment in their own pieces.
The biggest challenge in this area is that already-thorny element, business services. To increase business spending on new network services or technology, you need to make a business case. To make a business case you need a project, and that project has to frame both an accurate cost and an accurate benefit that can be expected. Both of these are challenges because there may be little company experience with the cost or benefit elements. That’s especially true in the last decade, where most enterprises have seen fewer and fewer new technology projects and thus may have lost most of or all the personnel on staff with experience in making one work.
That’s not the only issue, either. Of the 412 enterprises who offered me comment on this area, 389 said that it was more difficult to make a new project business case because of a lack of vendor support. This appears to be caused by the fact that the leading IT vendor has become accustomed to selling into a pure budget-driven orderly upgrade cycle, and fears that a new project might introduce a competitor.
That raises our last issue, which is the challenge of undue exposure to a vendor. This used to mean “vendor lock-in” but these days the problem is seen in a different light, one that focuses more on preventing a relationship that could lead to lock-in than to dealing with it after the fact. It’s this shift that has led to enterprises and operators alike favoring “open” strategies.
Open source software offers the buyer a hope that they don’t have to depend on a single vendor to sustain something they buy, but also levels the playing field during an RFI/RFP comparison. With proprietary software, a buyer often has different package features to consider when looking at multiple sources, which means that the comparison involves a kind of sliding scale of capabilities that often means a different benefit set to consider in the business case. Since costs are also different, this makes it much harder to evaluate different vendors.
One thing that AI proved was that sometimes an open software model isn’t as open as it might seem. An AI model is a bit like a programming language; it’s a tool that’s used to do something, but it’s the “something” that creates the benefit, and there may be other steps in getting to it. With AI, the model has to be trained, and so you can have an open model with proprietary training that creates as much dependency on your supplier as a closed model would. Before, and apart from, AI, there’s also the question of support. Over 80% of enterprises who use open source software get it bundled with support and integration and from a commercial supplier. Two-thirds of these firms say that they consider that supplier to be a “vendor”.
OK, there are a lot of reasons why network spending isn’t keeping pace with growth rates of the past. What can be done? At one level, nothing, and at another, perhaps a lot.
Nothing can be done to network services to create a significant boost in operator revenues, other than to raise prices (which in many areas is happening or has already happened). The only possible remedy is to increase the value of what networks deliver, to expand use of networks and network capacity. All of the things that we hear about today from network vendors is at best holding the line for a while in a losing battle. QoS, opex reduction, cost management…it all vanishes to a point. The only thing that can work is a combined initiative to create new producer value, new applications that deliver new stuff. Absent that, we can expect that some form of subsidization by OTTs or by taxpayers will be needed. Sorry, but that’s how it is.
