The telecommunications industry is used to paying for software based on usage. Would it accept paying for hardware based on usage?

The concept of pay-as-you-go software is not new, especially in the telecommunications industry where software is usually licensed rather than purchased. In many cases, this is a recurring license that scales according to the number of instances of the software that a customer has running in their network. It is an advantageous arrangement for both the vendor and the customer as it allows the customer to pay based on what they are actually using and allows the vendor to have an ongoing revenue stream.

With the ascendancy of NFV in particular, but also SDN, software instances within the network will be deployed for an increasing number of functions. Rather than a network that is almost entirely made up of purpose-built hardware running vendor-specific hardware, more and more functions will be migrating to common hardware with purpose-built software. Essentially, the cost of the network is moving from a capital-intensive deployment to an operations-intensive deployment. This has huge benefits for network operators, even beyond the scalability and flexibility advantages of software-designed networks. Vendors who have traditionally sold hardware that performed these functions are scrambling to figure out the best way to deal with the new paradigm, but most have plans in place to confront the new reality. While initial capital income may decrease, the move to a more predictable software model for functionality deployment should lead to a long-term profitable revenue stream.

Recently, I have had chances to talk with several people in the industry about the idea of moving the non-virtualized network hardware to a pay-as-you-go model. The concept was first mentioned widely in discussions at the GEN15 conference, but has been brought up now in a variety of venues. A large percentage of network hardware will always be purpose-built, capex-intensive hardware. Optical transport, physical network interfaces, and other hardware cannot be virtualized and will never run as a virtual application on a white box. Many would claim that much of the switching infrastructure will always be custom-built. Even the functions that can be virtualized will have to run on hardware somewhere in the network. All of this hardware represents an up-front sunk cost that must be justified before the network can be built.

There have been some concepts of pay-as-you-go models for hardware in telecommunications in the past. One example is the optical equipment model from Infinera that allows customers to pay for the high speed bandwidth in 20% chunks (their hardware comes in 500Gbps pieces, and customers can pay for it 100Gbps at a time). Another example is the old DS1/DS3 multiplexers that had software keys to allow more interfaces to be used as they were needed. Pluggable optics have made many WDM devices essentially pay-as-you-go, since the most expensive piece can be added as  needed. All of these models allowed the initial capital expense of the hardware to be spread out over time to make swallowing the initial investment easier for the CFO organization. All of them were still fundamentally capital expenditure models.

The concept of taking hardware to a true software-like model was first brought to my attention by an executive at a major US carrier. His statement was that all purchases for the network would move to the pay-as-you-go model, and that vendors should start pricing their hardware accordingly. Rather than purchasing a 100Gbps packet-optical system based on the hardware installed, customers would purchase a license to run the equipment. That license would be either usage- or time-based. As long as the equipment was in the network, the customer would pay the vendor for the right to use it. However, as soon as the equipment was removed, the payments would stop. The initial outlay would be much lower and the ongoing costs would be stable and predictable.

There are some real benefits to moving to this type of model, but also some significant potential pitfalls.

Benefit to carriers

A reduced up-front expense could allow faster expansion of networks and services to more customers and allow more experimentation with equipment deployment in new areas. The predictable costs would make planning easier, especially as the in-service equipment costs start to dominate any new additions. Also ongoing expenses can always be reduced in future negotiations with vendors, meaning that the future costs could go down but would not be expected to rise. For newer carriers (e.g. the content providers) this model works especially well, as they tend to change out hardware every 2 to 3 years, meaning significantly less sunk cost.

Benefit to vendors

While the initial conversion to pay-as-you-go would likely hit the bottom line for vendors, long term this concept could lead to a predictable ongoing revenue stream. The payback timeframe would have to be carefully studied, as networking equipment is expected to have a lifetime of about five years in normal deployment. The reality is that many network devices stay in the networks for much longer – many for more than ten years. For equipment providers who have designed their licensing plans such that there is a three year payback, all payments from years four onward are pure profit. The low initial cost could entice customers to try equipment with lessened risks, allowing innovative solutions to gain a foothold in networks.

Pitfalls for carriers

For a carrier that keeps equipment in the network for a very long time, the total cost of ownership for pay-as-you-go equipment could greatly exceed the value of the hardware installed. This could be compounded by no longer being able to depreciate hardware for tax purposes. There is also a possibility that moving to this model will reduce competition among vendor solutions, as only those vendors with strong long term financials would be able to participate. Finally, the additional record-keeping requirements – especially if the license is usage based – could impact the ability to deploy solutions just as strongly if not more strongly that current capex pressures.

Pitfalls for vendors

As mentioned above, only the strongest of vendors would be able to participate in this model. New startups often rely on their first purchase orders to give them enough runway to get to the second purchase order. If the first purchase order is for a small percentage of the cost of the hardware, that model breaks down. This is not only true for startups, but also for new projects within larger companies that are reporting revenue on a quarterly basis. Also, vendors to webscale-type customers may not be able to make enough money in licensing before the equipment is replaced in the network.

The bottom line is that the concept of pay-as-you-go hardware is in the early stages. While the larger network providers will likely push such a model, it is not yet clear how that model will work to keep the entire ecosystem healthy long term. As network operators and their vendors adjust to the realities of the new software-defined world, this is just one of the many issues that will need to be addressed.