As the results of companies like Bharti Airtel and Idea make clear, Indias telecom sector continues to be in trouble, though there has been a small let up over the past couple of quarters. Problem is, a modest return to pricing power is quite irrelevantvoice revenues have been slowing for a long time, more so if you discount them, as you should, with inflation (see graph 1).
But its not just India, voice is a decreasing business globally, and data has increased dramatically to take up the slack (see graphs 2 and 3). Indeed, thats why Indias broadband internet targets are 75 million by 2012, 175 million by 2017 and 600 million by 2020. Right now, if you include 3G subscribers, which you should because mobile broadband is really the way forward (by 2011, smartphones outstripped both PCs and laptops), India has about 40 million broadband subscribers, or around half last years target.
Getting even near the targets requires two thingscheaper smartphones and enough spectrum. From R25,000 or so for smartphones in March 2010, we are now down to smartphones that cost around R5,000, but this probably needs to fall to R2,500 or so to get really mass market. Indias real problem, however, doesnt lie here.
It lies in the huge spectrum scarcity. Look at graph 4 to see how data guzzles spectrum. From 80 petabytes of spectrum in Q1 2007, major global voice networks used up around 190 petabytes in Q3FY12. In contrast, data networks, which barely consumed anything then, consume around 900 petabytes of data now (1 peta = 1,000 tera).
Where does India stand in all of this? If you include BWA which you really shouldnt given that it will take 2-3 years before it really matters (3G auction took place in 2010 and there are 30 million subscribers today), India has 2x50 MHz of spectrum as compared to 2x197 in the US and 2x154 in Germany.
To put this in perspective, lets look at how many subscribers India can possibly host on its 3G networks which are up and runningwe can do the same for BWA after that. If each subscriber uses 300 MB of data per month at the moment, based on an assumption of network utilisation being 60-70% and only a tenth of users being online at any point in time (if this increases, the number of users who can be serviced falls), we can have a maximum of