Why Early 5G Capex Forecasts Were Wrong

One reason for scary-high 5G capital investment requirements--and an argument that used to be made relatively frequently--is the assumption that millimeter wave spectrum would require network investments similar to what has been made in the macrocell era, using substantial towers and real estate.

So many took existing tower site metrics and multiplied by the number of new small cells to replicate present coverage. Early experience in South Korea, with most of its coverage in urban areas, found that, using a mix of 3.5 GHz for coverage and 28-GHz spectrum for capacity, the number of new cell sites increased by four times.

That might sound like quite a lot more sites. It actually represents a shrinking of macrocell signal radius by 50 percent. That shows the importance of mixing coverage spectrum and capacity spectrum, as always has been the case in the 4G era.

That is turning out not to be correct, at least in many markets. Small cells are an order of magnitude cheaper than macrocells, though the backhaul requirements arguably are not so different, unless deep fiber networks have been deployed in the areas where most of the small cells will be placed.

source: Informa

But it is a mistaken notion that millimeter wave spectrum will be used for coverage requirements, instead of capacity. As this 2010 analysis of U.K. cell coverage requirements indicates, small cells in urban areas are a fraction of total coverage needs. Urban areas represent just 0.2 percent of the total U.K. landmass.

Suburban areas represent seven percent of the landmass. And those are the areas that most likely have data demand high enough to require small cells. But that still leaves 93 percent of the landmass better suited to lower-frequency spectrum that will not match urban bandwidth levels, but also can be built at incremental cost, often reusing existing assets.

source: Science Direct

The obvious solution is to use lower-frequency spectrum for coverage, at the expense of capacity, and then use millimeter wave for capacity in the dense urban areas. As the U.K. data suggests, that requires dense small cell networks in a very-small percent of the coverage area.

The other issue is deployment tempo. It often will make sense to use dynamic spectrum aggregation to combine 4G and 5G capacity assets. It will make sense to use spectrum aggregation to bond licensed spectrum to unlicensed. In some markets, spectrum sharing will be possible. All those techniques means that existing licensed and unlicensed spectrum becomes a part of the 5G experience.

The practical implication is that new 5G small cell capacity is added incrementally, to reinforce urban area capacity. In most areas, 5G will use lower-frequency spectrum with much-better coverage characteristics. In many cases, that means the need for new macrocells is sharply limited.

In the interim, boosting 4G performance will be a wiser use of capital.

The point is that early fears about 5G infrastructure cost were based on overly-simplistic and incorrect notions about use of small cells, and the cost of a small cell. Early estimates also ignored new ways of using spectrum that build on and incorporate 4G and unlicensed spectrum, spectrum sharing and the ability to gradually introduce 5G small cells where most needed.