FCC recently published an Omnibus Broadband Initiative (OBI) Technical Paper, titled “Mobile Broadband: The Benefits of Additional Spectrum“. Its message can be summarized as follows:
- By 2014 there will be a significant deficit of available spectrum to cope with rapidly growing wireless data traffic. Even if we assume flat voice traffic, with 3500% growth in data traffic between 2009 and 2014, paper estimates there will be a spectrum gap of 275 MHz.
- Cost of delivering data traffic in the absence of additional spectrum is building new cell-sites, over 200,000 of them, each at an estimate cost of $550K (this includes the operational cost).
Following graph from the paper shows the projected deficit spectrum by 2014.
Paper includes sensitivity analysis covering assumption about data growth rate, current amount of spectrum used for data and cell-site deployment cost. Even though these three sets of analysis show significant variation in valuation, ultimately paper concludes that there is a very high value associated with this spectrum primarily to reduce the amount of cell-site builds.
Looking at the numbers in more detail, I consider FCC’s assessment unnecessarily pessimistic. My reasoning is as follows:
- Year-over-year cell-site growth figure is closer to 9% as opposed to 7% figure used in FCC’s model. That number was based on a 5-year averaging from 2009 that included the growth figure of 2% in 2005 in the analysis. 2005 was an anomaly since AT&T and Cingular completed their merger in later 2004 and T-Mobile and AT&T/Cingular entered into network sharing agreements. Another major factor for a yearly growth rate of 9% is the LightSquared’s public plans to build out a 40,000 cell-site network by 2015.
- Spectral efficiency figures projected by FCC are too low. They assume 2*2 MIMO for LTE with average spectral efficiency of 1.36-1.5 bits/s/Hz. On the other hand LTE (3GPP release 8) is designed to provide downlink (that’s typically what we care about when talking about data services) average spectral efficiency of 3.7 bits/s/Hz with 4*4 MIMO (Look at IEEE Comm. Mag. article from Feb. 2009). Considering LTE-Advanced (release 10 and beyond) will be in service around 2014 or so, I believe a much faster increase in average spectral efficiency will be achieved. There are two more factors in favor of this trend: (i) 3G system performance will continue to improve. At least for the GSM family, i.e., HSPA+ will continue to ride the bandwagon of MIMO for some time to come. (ii) Around 2014 LTE networks will carry at least 30% of the total data traffic. Fast adoption of more spectrally efficient technology has a lot of precedents, such as moving from CSD to GPRS or 1xRTT, or moving from EDGE or 1xRTT to HSPA(+) or EV-DO. This adoption will increase the average spectral efficiency to a higher value than what FCC predicts.
- Ultimately even wireless data traffic growth will slow down, at least percentage-wise. Beyond 2015, this may slow down to as low as 20-30% yearly growth figures. Similar slow-downs in growth have been observed and expected in wireline traffic.
Using these assumptions, I took the liberty of updating Exhibit 10 in the FCC OBI paper. I ended up with very different numbers, especially when I stretch the timeline to 2016.
I believe the uncertainty in forecasting spectrum needs is too high. Especially attaching a $120B value to additional 300MHz spectrum seems premature. Considering a significant percentage of this traffic can be off-loaded using WiFi or even a less likely method of femtocells for much lower price levels, I find FCC’s numbers to be quite difficult to accept at face value. Certainly the actual spectrum value will be known only when projected auctions start taking place 2012 onward. By then we will be able to see how traffic growth, cell-site deployments and spectral efficiency improvements are taking shape. Next 12-18 months will help decision makers in attributing a more accurate value to spectrum.