As Ory Okolloh recently put it: if mobile phones are the computers for Africa, why do all the journalists who write about them do so on laptops? I wholeheartedly agree. The smartphone is a great handheld entertainment and communication platform, but it isn’t a good device for the creation of much more than a drunken text message.
No, the truth is that that laptop computer - or some variation of it - is almost certainly the utility computing platform for a connected Africa. It can do pretty much everything a smartphone, TV, desktop or tablet can do, yet costs less than any one of those devices. It isn't perfect, but it is the best utility computing device available today. In my view the device challenge is not to identify the computer for Africa, but to create a supply chain that reduces unit costs to a point that is affordable to the middle class (in the first instance).
So if the computer for Africa probably isn't a phone, why should it'd connection be designed for a phone? Or, put another way: if cellular is the broadband connection for Africa, why do all the commentators who say so submit their stories using fixed lines?
My personal view is that the relatively low cost per POP of installing cellular broadband systems versus fixed alternatives blinds commentators of all kinds from the practical reality of using cellular systems for mass market broadband Internet. Particularly if that Internet is provisioned to devices that are will actually have a meaningful impact on the education and workplace prospects of African consumers.
Simply put, the physics and engineering of cellular systems makes it impossible to service a reasonable level of broadband access demand in a market with no fixed offload.
I’ve written before about the fundamental economics of mobile broadband. So now I’d like to extend the analysis to test whether or not it is feasible for a cellular network to cope with the traffic demands of mass market provision. I’ll do this by showing the absolute minimum number of cells required to support the traffic demands of consumer data in a broadband economy with 35 million people in it... One the size of Kenya, for example.
Now for the science bit...
First, we need make some assumptions about the average volume of traffic a typical user consumes. In the UK, the average citizen “consumed” about 17GB of data in 2011. This takes into account the 25% of the population that does not have a broadband connection and the 50% who now have a smartphone. The UK is by no means an outlier in this – US, Scandinavian and developed Asian economies would post higher numbers. I’ve also removed file sharing from the total, as I presume an African Telco would shape it out of traffic or shift it into off peak.
To be conservative, I’ve also made some assumptions about the nature of consumption in an advanced African economy in 2015. Let’s say for the sake of argument that it’s 60% of the average in the UK in 2011. Very reasonable given the continued growth of consumption of broadband data in advanced economies and the increasing size of the average website.
I’ll also assume (and this is really conservative) that the African country has a universal LTE network. No African country will have this in 2015 - as I say, I’m being conservative. To get to a number of basestations, I'll assume 3-network competition. Most African countries have more operators, but perhaps by 2015 that situation will trend towards the mature market norms.
I won’t bore you with the calculation (leave a comment if you’d like it). The upshot is that in order to supply that amount of data via cellular in a competitive market requires about 115,000 cells and 100MHz of paired, cleared spectrum. Furthermore, this network would have no spare capacity for voice, which is responsible for the majority of ARPU and EBITDA in developed and developing markets. For reference, Kenya has fewer than 8,000 cell sites today and the majority are based on GSM.
...and the economics bit
Installing, provisioning and connecting (with fibre) those base stations and antennas would cost roughly $34Bn. Curiously, that’s about twice as much as it would cost to provide every household with an FTTH connection, not that I’m suggesting that’s a great plan in an economy that will have a GDP per capita of $3,000 if it really runs fast for the next 4 years...
So what’s the answer? Well, the success of Alvarion and other providers of advanced Wifi networks in some African countries provides a strong hint that the future African ISP will a hybrid of fibre to the curb and high-speed (and most importantly – very cheap) non-cellular Wifi access.
My benchmarks suggest that the cost of installing, backhauling and installing a Wifi hub in an urban area is around $7,000 – vastly cheaper than even distributed cellular RANs using cloud baseband technology. Clearly using Wifi is more difficult in rural areas, but cellular also becomes more expensive outside cities – hence the recent interest in satellite backhaul.
Perhaps this is why companies like Jamii persist in building out fibre POPs in countries with tiny fixed broadband penetration. If so, they may be good investments. One further thought is that Wifi chipsets are at least a 5th of the cost of HSPA and about a 20th of the cost of LTE. So there is a user device cost benefit of not using cellular for mass market broadband provision, in addition to the network cost described above.
Don’t get the impression that I’m down on cellular as a communications technology for Africa. For voice, text and mobile services like payment it has the reliability, flexibility and security to be an excellent technology that makes a lot of money for those companies that can grab enough market share in the short term.
But it is not a substitute to fixed broadband, either directly provisioned to the user’s home or extended into that home with Wifi.