In most mobile markets, the key numbers are "three or four," representing the number of sustainable operations. In some fixed markets, the relevant numbers might now be "two or three," likewise representing the number of viable facilities-based competitors.
Technologists, service provider executives and analysts have endlessly debated the “cost” of deploying high-capacity networks of several types for a few decades. Over those decades, the cost parameters have changed.
Fiber to home and digital subscriber line hardware have gotten more affordable. Cable TV DOCSIS platforms have vastly improved. Now there are other potential options.
Internet access by balloon, unmanned aerial vehicles, fifth generation mobile networks, fixed wireless (TV white spaces, for example), Wi-Fi hotspots, municipal networks and new constellations of low earth orbit satellites are some of the reasons to argue that Internet access business models are liable to be redefined over the next decade.
On the other hand, other key business model inputs have not changed very much. The network cost (outside plant) still appears to hover around $1,500 per location, including construction and hardware.
Active elements still cost about $280 per location.
Activated drops and customer premises equipment costs remain substantial. CPE alone might represent costs of $455 or so per active customer location. Installing an actual drop for a customer can cost $300.
Taken all together, the cost of an FTTH install in a medium-sized U.S. city, for example, might be about $1780. The cost to connect a paying customer adds another $755.
So it still can take $2535 in network-related costs to serve a customer. Marketing costs have to be added on top, as well. New attackers likely can figure out ways to spend less than the $850 to $2,000 tier-one competitors typically spend (service discounts plus out of pocket costs) to acquire a new account.
But business models are even more sensitive to take rates, in turn driven in large part by the number of locations served by a competitor that cannot be dislodged.
In other words, in markets where two other competitors--both accomplished--continue to hold about 60 to 66 percent customer share, per-customer costs are substantially higher than per-location costs.
If per-location fixed cost is $2535, and take rates are 33 percent, then network cost per paying customer is $7605. After drop costs and CPE, per-customer cost (without marketing) is $8360.
If per-month revenue is $100, it takes more than seven years to recover network costs. Few, if any, private firms would undertake such an endeavor, given those obstacles.
The business model works better in markets where just one serious competitor operates (a cable operator, for example), allowing the attacker to contemplate take rates more along the lines of 50 percent. That reduces per-customer network cost to $5070, with total connection cost of $5825.
That possible variance is possible because, in some markets, competitors might be very vulnerable to a challenge. The incumbents might not be able to afford to reinvest in their own networks.
In other cases they might prefer taking a big share loss to reinvesting at the level required to blunt the attack (some companies focus on major urban markets and are willing to lose rural or low-density markets).
Higher market share (50 percent rather than 33 percent) reduces the break-even point on network investment by about two years. That still is a tough hurdle, though, with more than five years required just to recover installed network costs.
Boosting average revenue per account therefore is a key strategy for quickening the payback. If per-account revenue is $150, instead of $100, break even times shrink. At 33 percent take rates, and $150 monthly revenue per account, break even on network costs comes in less than five years.
At 50-percent take rates, and $150 monthly revenue per account, break even on the network investment can come in a bit more than three years. That is a workable business model for many firms.
One new approach, in that regard, is to strand fewer assets, building only in some neighborhoods, for example. That also lowers overall capital investment, since a smaller network is built.
A 2016 DIscus Project white paper on high speed access network business models reviewed models where more than one physical network operated, asking the key question: how many such networks are sustainable?
The conclusion, as you would expect, is that few such networks are sustainable in any given region or market: “two or three.” The important number is “three.” Under some conditions, three facilities-based competitors might be sustainable, where today the number is “two.”
Where the contestants are private firms, sustainable operations are possible in dense urban areas, for example, or where one existing incumbent cannot, or will not, upgrade to match the attack.
The point is that, as hard as the business model might yet be, the business case for new gigabit networks--even in markets where telcos and cable TV already operate--might well be improving to the point where, in some markets, three facilities-based competitors can sustain themselves.
