The US financial community is just starting to see what we have been suggesting for some time, that the seemingly in-exhaustible spectrum that companies like Verizon and AT&T have hoarded for years, is still not enough to sustain a future of data services, where what is delivered is mostly video.
First off this does not mean that video is what the consumers are doing most of the time, but given that a minute’s video takes up about 1,000 times more data than a simple SMS message, the minutes spent on video are about 1,000 times more costly than any other activity. So any video is tough for a cellular network.
There are two ways of preparing your network for such an onslaught – offload like crazy onto WiFi networks, or build out small cells. Both represent methods of extreme spectrum re-use, where a small area, say a single home, is supported by the spectrum in question – in WiFi 2.4 GHz and 5.0 GHz, and in LTE, any licensed spectrum.
The report which raised this comes from New Street Research and it raises concerns that Verizon will run out of spectrum and needs to aggressively buy more. It estimates that it needs to double its spectrum holdings by 2020. We not so sure.
To our minds this issue is more about design and cash than it is about spectrum. If you use the same spectrum in 3,000 Macrocells, they are laid out so as not to interfere with one another by having 3 channels all facing in 120 degree arcs. At the cell edge channel 1 faces channel 3 from the neighboring base cell, so it does not interfere. This needs carefully planned radio networks.
When you put a femtocell in a home, it picks a fraction of the channels, and uses them in all directions at a power level that cannot get far through the walls. It has an artificial cell edge, caused by thick walls, but it can also avoid any neighboring signals at install, in a fashion similar to WiFi. Femto cells are largely useless, but they have the advantage of costing about $50 and taking a free backhaul ride over your home broadband.
Small cells are not the same at all. In LTE Advanced they come with centrally managed dynamic interference management. The central resource that manages them decides how loud and in what spectrum each nearby cell can shout (in radio terms) dynamically, based on the traffic a cluster of small cells needs to deliver.
The problem with small cells is that unlike femtocells they have to be centrally managed, separately backhauled and radio planned. You can certainly tune a small cell network to keep delivering more and more data, by breaking it into smaller and smaller cells, but each time you add a cell, you need another backhaul point, and more processing power in the middle to coordinate all those cells.
We would back Verizon any day to plan a network ahead of New Street Research, which seems to have an imperfect understanding of how the process works. It is quoted in the press this week saying, “While Verizon has touted small cells as an alternative, small cells only carry about one third the capacity of a macro site, and the company would need to roll out over 80,000 small cells in just the top 25 markets to maintain the current QoS.” Now this may be right, and it goes on to talk about the challenge this represents in cost, and a planning challenge.
In Faultline’s view, the cost would be bearable if we were going through a period when ARPU was rising for cellular data. Or at least holding firm. But if a single operator out there falls back, as T-Mobile already has with its Binge On feature, onto WiFi offload, then the pricing available to T-Mobile means it can radically discount the cost of data (in Binge On it charges nothing for certain video data). This is a force driving down the cost of data.
Sprint has lots of spectrum, T-Mobile has original thinking that rests on WiFi offload, and AT&T has a bit more spectrum than Verizon, but also more importantly, more WiFi available to it. It is also perhaps ahead in its plans to use carrier aggregation, which is also available to Verizon to harness the rest of its non-traditional cellular cable spectrum holdings (like 2.3 GHz).
The analysts argued “As usage increases the incumbents will see performance deteriorate first and at a faster pace, leading to a natural shift in subscribers from more congested networks to less congested networks.”
The truth is that Verizon and AT&T will simply risk making less profit if they do not adhere to the lowering data ARPU expectation that WiFi brings, not less customers. And will be forced later into copycat approaches that adopt more existing, already available WiFi, from whichever sources they can muster.
Essentially there is now a capacity gap between the 4 US cellular operators, but video is levelling the playing field. Those with more spectrum (like Sprint) still have the cost of building it out, and those with less spectrum have the cost of adopting smaller and smaller cells. Or both operator types can go for large scale WiFi offload, a move that will lower capex, satisfy its shareholders and allow a more stealthy approach into small cell architectures. And at the same time it will enable ARPU data calls that keep customers happy.
LTE-U is of course another source of spectrum, but it is not here yet, and so we do not know its performance benefits, and it only replaces WiFi spectrum that is already largely built out. By the time the FCC grants it permission to operate in the US, LTE-U may be far less spectrally efficient than demonstrations suggest today.
It may be true that senior executives at Verizon are intent on buying more spectrum,. For instance they may lease or buy Dish’s spectrum, but this doesn’t mean that this approach is right, or the only way to go and it still has to be built out, backhauled and managed and that’s all more cost. New Street has a bit of a point, but the counter-plays are fascinating.