Internet Connectivity problem statement
Why is half the world unable to access the Internet?
1. Internet access: coverage ≠ connectivity
Mobile has become the primary means of Internet access in lower and middle income countries. However, as of December 2018 there were 3.3 billion people who live within a mobile Internet coverage area but are not on-line.
The Alliance for Affordable Internet (A4I) considers mobile data as affordable when 1 GB of data costs less than 2% of an individual’s income. A4I data shows that at present ~2 billion people live in areas where even the cost of 1 GB is greater than 2% of income.
Industry trade bodies and research organizations have consensus on the following four as primary connectivity barriers:
- Affordability: two sub-components: a) mobile data prices and b) mobile handset prices.
- Infrastructure*: classified into two types: a) coverage and b) capacity.
- Content and services*: classified into three types a) text, b) audio and c) video.
- Consumer readiness: driven by a) literacy level and b) consumer awareness.
[Note*: intentional over-simplification.]
These connectivity barriers appear distinct but are highly interdependent. To understand this claim, consider the case of Reliance Jio (“RJIO”) in India. RJIO offered a mobile handset for a ~$20 refundable deposit combined with an unlimited data offer for 6 months, thereby making its offering highly affordable.
To make such an offer technically feasible, RJIO built a mobile network in which majority of its cell towers were connected to optical fiber. For context, a single strand of fiber has 10,000x more capacity (or bandwidth) than the traditional alternative, i.e. microwave and a typical fiber cable has 48+ strands.
Thus, fiber infrastructure’s virtually infinite capacity made a 30x price reduction technically feasible for RJIO customers without compromising service quality. This price reduction made it affordable for semi-literate customers to deposit ~$20 in order to stream relevant video content on a mobile handset. RJIO launched its service in September 2016 and by April 2019 it was profitably serving 300 million customers.
Delivering video content from the Internet on mobile at an affordable price point can become feasible only if network capacity increases by several orders of magnitude. If prices are reduced without dramatically increasing network capacity, existing customers will experience decline in service quality.
Considering a single fiber strand has several thousand times more capacity than any other competing communication medium, enhancing mobile network capacity is synonymous with fiberizing more cell towers.
Interesting fact: two fiber strands have more capacity than SpaceX’e entire Lower Earth Orbit (LEO) satellite constellation comprising of 12,000 satellites.
Was it the $20 handset or an unlimited data offer that made RJIO’s offer really compelling?
Consider the case of television (TV) adoption rates relative to income. TV adoption rates show that even when a TV costs as much as an individual’s annual income, adoption is still ~80–90%. This suggests that entertainment is a critical parameter for users to determine value of an appliance. If a mobile phone can provide equal entertainment value to TV, then its price tag should not a limiting factor.
2. Mobile Network Operators: ARPU down, Data per customer up
While RJIO is an important example, it is an exception because it had access to $23 billion of start-up capital, no legacy telecom infrastructure and the support of India’s wealthiest businessman. Traditional Mobile Network Operators (MNOs) face the following challenges:
- Average Revenue per existing User (ARPU) is declining while data consumed per customer is increasing.
- Fiber deployments require high upfront capital commitments whereas an MNO’s cash flow predictability is declining.
- Fiberizing cell towers increases data revenues but cannibalises voice revenues because more customers start using voice over IP (eg WhatsApp calls)
- Majority of MNO towers (90+%) are equipped with capacity constrained microwave equipment and this equipment is not fully depreciated.
As a result, board members of MNOs are reluctant to undertake large scale fiber deployments to enhance network capacity. The situation is even more challenging in rural areas. A typical MNO generates 50+% of revenues from ~10% towers that are located in urban areas. However, majority of an MNO’s towers are located in rural areas, but rural towers account for ~10% of revenues.
Considering urbanization levels will continue to increase, the business case for upgrading network capacity in rural areas for an MNO by itself will almost always result in a negative Net Present Value (NPV). MNOs will fiberize rural towers only if a third party finances rural fiber deployments and ensures the fiber service offered does not increase an MNO’s operating costs.
Thus, barring an RJIO type intervention, making high speed Internet access affordable for 2–3 billion people living in rural areas seems unlikely.
3. Regulations: obsolete definitions and dormant capital
Regulators have attempted to bridge the rural-urban Internet connectivity divide by:
- Imposing download speed targets (in bits per second) in addition to coverage requirements.
- Taxing 1–2% of telecom revenues to create a fund that would ensure universal service known as Universal Service Fund (USF).
- Promoting infrastructure sharing schemes.
Defining broadband in bits per second or Megabits per second (Mbps) is problematic because download speed is a moving target. For example, the Federal Communication Commission (FCC) in the United States defined broadband as 200 Kbps in 1996, 4 Mbps in 2010 and 25 Mbps in 2015. According to Nielsen’s law of Internet bandwidth, user bandwidth doubles every 21 months. Thus, a definition based on infrastructure proximity will be more future proof than one based on bits per second.
Considering optical fiber has several times more bandwidth than any other medium, proximity to fiber networks can serve as a better broadband definition than download speed targets.
Regulators set up Universal Service Funds (USFs) because they knew serving rural customers will not be financially viable for MNOs. While USFs successfully collect funds from MNOs each year around the world, most of have struggled to deploy capital.
A 2014 GSMA study shows ~30% funds collected by USF were deployed globally and $12 billion remained un-disbursed. Subsequent research efforts show that this trend remains unchanged. USF fund utilization data suggests that limited investment opportunities is a bigger constraint than availability of capital.
Since USF is made primarily from MNO contributions, an investment that relies on USF to fiberize rural towers must be able to demonstrate long-term value to all competing telecoms in a given market, otherwise obtaining USF board approval will be challenging. Thus, it is critical that USF supports open access fiber (ie non exclusive) deployments.
While regulators have struggled with broadband definitions and USF deployments, a major success story enabled by suitable regulations is cellular tower sharing. According to TowerXchange, 68.8% of the world’s 4.65 million cell towers are now owned by infrastructure sharing companies (infra cos). Most tower sharing companies achieved scale by acquiring MNO owned cell towers and leasing them back to an MNO through a long-term (~10 year) contract.
In some cases MNOs have attempted to build infra cos as subsidiaries, however, independent infrastructure sharing companies have outperformed those owned by MNOs. According to TowerXchange, tower cos owned by non-MNO shareholders have a tenancy ratio (i.e. number of different customers per tower) of 1.75x whereas tower cos majority owned by MNOs have a tenancy ratio of 1.2x. The superior performance of independent infra cos highlights the competitive distrust that typically exists between MNOs. This suggests that independent infra cos are best positioned to lead rural fiberization initiatives.
4. Infrastructure sharing: tower co business > MNO business
A noteworthy development of tower sharing companies or infra cos is that they have financially outperformed MNOs by a significant margin over the past 15 years. An average tower co today trades at a 2.5x premium to an MNO. This out-performance is even more striking in tower cos that are owned by MNOs. For example, INWIT in Italy was spun out of Telecom Italia. INWIT investors have realized returns whereas Telecom Italia investors have incurred losses.
Driven by the success of tower sharing companies, many tower cos and technology companies have tried to replicate the success of tower sharing business model with fiber. For example, Csquared was spun out of Google, it raised $100 million in investment in May 2017 to create shared fiber infrastructure for MNOs and ISPs. For Csquared to build fiber it first needs to secure a contract from 1 or more MNOs to share the deployed network. Such negotiations became cumbersome because unlike coverage, MNOs still view network capacity as a source of competitive advantage. As a result, Csquared has struggled to scale beyond urban areas.
Important to note that cell tower sharing was not feasible when MNOs were first building networks. When networks were first being deployed, coverage was considered a competitive advantage. An MNO with presence in 60 cities was a more compelling value proposition than one with presence in 40 cities. Tower sharing sharing became feasible only after network coverage became a commodity.
Even though the fiber cable itself is a commodity (costing less than $1/meter), actual deployment costs range from $20–50/meter because of labor and Right of Way fees. Thus, infra cos will have to look beyond MNOs to find an anchor tenant. Furthermore, infra cos will have to innovate to deploy fiber cheaply otherwise offering MNOs an “Opex neutral” deal to fiberize rural towers will be unlikely.
The problem statement
On-boarding the next 4 billion semi literate users to the Internet will require making mobile video streaming feasible at $1 / month. A price drop of this magnitude will require increasing network capacity by at least 1,000x.
Optical fiber has several thousand times more capacity than any other communication medium, which makes it the best technology choice for upgrading network capacity. The fiber cable itself is a commodity costing less than $1/meter, but fiber deployment costs can range from $20–50/meter because of labor and Right of Way fees.
~10% towers (already fiberized) are located in urban areas and account for 50+% of telecom revenues. ~50% towers, equipped with microwave are located in rural areas and account for ~10% of telecom revenues. Fiberizing rural towers is not feasible for telecoms even when an infra co builds a shared network.
Regulators can facilitate rural fiberization initiatives by making USF capital available to independent infra cos. However, infra cos will still have to figure out a way to build fiber without expecting a telecom company to become an anchor tenant. More importantly, infra cos will have to find non telecom revenues to support rural fiber deployments.
If infra cos can deploy fiber with a non-telecom anchor tenant and offer a fiber service to telecoms that does not increase their operating costs, rural tower fiberization will become feasible.
Increasing global tower fiberization from 10% to 80% will increase network capacity by several thousand times. When 3 competing MNOs have access to 10,000x more capacity at each cell tower without added costs, price reduction of mobile data will follow.