The economics of Fibre Customer Data Networks are surprisingly strong, they are a license to print money, unlike any other business opportunity out there now. Exponential growth in revenues are backed by exponentially dropping underlying costs. The most-profitable high-end users willingly pay more and self-identify, allowing NBN Co's Retailers to provide them premium-priced services, maximising retailer margins and profitability. Copper Data Networks fail on all these critical commercial points.
History
At a Unix Conference in 1988 I spoke about how Optical Fibre had changed the rules of the game, today I can tell you a lot more about why. It's another effect of Moore's Law, the headlong rush that has seen CPU performance double even 12-18 months for around four decades.
Optical Fibre has changed very little since the 1980's, all the development happens in the Silicon at either end: the transceivers, the laser transmitters and photodiode receivers.
Moore's Law is driving the speed increase at rates that boggle the mind: a thousand times faster in 15 years. What other commodity has its input prices change that much?
The second part of the equation is the cost of production parts is not related to speed or complexity, but to how many you make. The price halves when you make 10 times more. It approaches the cost of the raw materials when you make 100,000 times as many. Think of flash memory in USB memory sticks or camera SD cards. They are now 4-8GB for $5-$10 at the supermarket. They're "as cheap as chips" because so many are made. Just a decade ago, they were 100+ times more expensive.
The commercial implications of this are profound: it's a license to print money, especially if you can get people to pay by the MegaByte (MB) or now GigaByte (GB), not to be confused with a Gigabit per second (Gbps) [1 Byte equals 8 bits, Big-B and little-b].
This is the commercial underpinning of commercial Optical Fibre networking: install the fibre and upgrade the electronics at either end. The first Fibre link in Australia was put in service in 1988. The Fibre is still in use, the original electronics long gone. They may have been 34Mbps, today they would be using 40Gbps WDM, faster than 1,000Gbps (1Tbps, or Terrabit per second). Now they use volume production standard ethernet parts, not low-volume "Telco grade" systems that were 100 or more times the price for 30,000 times less performance.
The cost-per-bit has gone down 3,000,000 times since my talk. A figure I would not have been game to estimate back then. [In nominal dollars. In real terms another 5 or so times, but who's counting?]
But unless you have customers willing to buy at good price, you have no business. If internal costs go down to almost nothing, but customers purchase no more units, then you can't drop prices!
What we've seen over 6 decades of computing is insatiable consumer demand for every sub-system: CPU's, memory, storage, LAN-speed and now Internet access speeds.
As speed/capacity rises exponentially and per-part price drops, consumer demand has relentlessly increased faster, driving production volumes up, which drives prices down, stimulating more sales in a "virtuous circle".
The first (1983) 4.77Mhz (not GHz) IBM PC with a 5MB drive cost $2-3,000 then, 30 years ago.
A $75 tablet today is 1,000 times faster, with 16,000 more memory and 1,000 times more storage, with much better display and graphics card. It runs off an internal battery, using at most a few watts compared to 130W - and has sound, camera, USB, WiFi networking, bluetooth and more.
While sales do build, they follow the usual early-adopter -- mature adopter -- laggard, formula. They have always been people who will pay a premium to buy the latest and greatest: they can assign a dollar value to the extra utility the device gives them.
The early-adopters, the 5-10% "high-end" consumers, pay for the development of products and let the rest of us ride on their coat tails. Somebody has to start the virtuous circle of demand driving down price, stimulating demand, and it's these people who'll pay more.
While the price of PC's and now smartphones and tablets has come down, while their capacity has increased, the total sales, and hence the total revenue & profits, have steadily increased. The demand for additional computing resources keeps expanding even as prices drop.
It's important to not look at average demand, but what the premium-users (5-10%) are wanting today. Next year, 25% of consumers will want it, the year after 50%, and so on.
Look at the stupendous take-up rates of smartphones since 2007 and tablets since 2010. There is serious money in this. Apple understands these economics and rode them to become, for a time, the big company on the Stock Exchange.
Telcos need to compare the input prices of the competing technologies: Copper and Fibre.
Then the price-per-bit now and in 5- and 10-years: what does the cost curve look like?
Until you get there, the Operational costs determine your profit margin or pricing.
And the crunch is: Customer Demand. How will it grow?
But less obvious: The distribution of consumer demand. What impact do the premium customers have on your income?
To break this down further, some extra concepts are needed.
Management Accounting likes to break product costs into parts: Fixed and Variable.
Telcos and Network Operators have a very large proportion of Fixed Costs and low Variable costs. Fixed costs are those that you have to pay if you sell nothing, like interest and depreciation on all your equipment. The power to run the gear is a variable cost.
Costs are further divided into direct and indirect. Is a particular asset or consumable linked to exactly one product or service, or shared? The Fibre that runs into your house, and all that goes with it, is a direct cost. The transit network and PoI's are shared amongst many subscribers. The fixed fraction could be a direct cost, and shared upgrades, replacements and housekeeping taken as indirect costs.
In Customer Access Networks, engineering is generally simple and cheap, but total costs are high because there are millions of connections. As you move from the edge, closer to the centre of the network, speed, price and complexity of gear required increases because instead of supporting just one subscriber, traffic from hundreds or thousands must be carried.
The majority of the network costs are Fixed, not Variable: the whole network has to be built for it to sell a useful product. Interest and Depreciation dominate the overheads, with I.T. systems and Labour to run the system next. There are Variable Costs: Power, consumables, replacement parts and rental, plus line-related maintenance, dominated by labour costs.
What really makes a difference to input costs is Interest Rate and line-related maintenance.
Any company, like NBN Co, that can get money at 3% p.a. is laughing. One of its biggest Fixed costs becomes quite manageable.
Comparing Copper and Fibre Costs
NBN Co in April released construction figures for direct Fibre: $1100-$1400 per service. These are commercially supported figures, not estimates or guesses based on dissimilar projects.
The Coalition detailed plan suggests they used $900/service passed for VDSL2/FTTN (Copper) with $90/service in line-related maintenance, 10% of the capital price.
NBN Co have not released figures for line-related maintenance. Elsewhere there are suggestions Fibre is 7-8 times cheaper to maintain than FTTN. The Coalition suggested a figure of 1.5-2% of capital cost, this would be $20-$25/service.
The Depreciation rates of the two networks, FTTN and FTTP, Copper and Fibre, are different, but with their different build cost, they end up depreciating around the same amount per year.
The Copper Network will have a service life of 15-20 years, while the Fibre Network will have more than 30 years in service.
Copper depreciation per service, straight-line, will be $900 ÷ 20, or $45/year,
while Fibre depreciation per service, is $1400 ÷ 30, or $45/year.
According to reliable commercial figures, the FTTP (Fibre) network is only 50% more than the price guessed by the Coalition. At 3% p.a. interest, that's $27 for Copper and $42 for Fibre.
Total Costs (Interest, Depreciation, Maintenance) per service are:
Copper = $27 + $45 + $90 = $162
Fibre = $42 + $45 + $25 = $112.
A Fibre Network will be $50/year/service cheaper to own, around 30% less based on sound figures.
Comparing Copper and Fibre Network charges
There is a small variable cost in CAN's related to cost-per-bit.
Consumers value services based on what 'utility' it provide them. For someone that has a need, higher access rates are worth more.
While access-rate barely affects costs, it can be used to differentiate services for users.
For Fibre here are 3 indicative prices: 12/1Mbps: $24, 100/40Mbps: $38, 1000/400: $150.
For users that need higher speeds, they offer exceptional value. To the consumer, it's twenty times cheaper to buy a single 1000/400 Mbps service than 80 of the slowest services.
Currently, the access charges of NBN Co average around $30/user.
That's a 25% increase for exactly the same physical equipment, just for asking. You'll find customers buying the premium service are extremely happy with the deal.
When 1000/400 Mbps access is offered, the average will kick up by another 15%, just for asking.
And underlying the point, with customers happy they are getting an outstanding deal.
For a Copper CAN, without the ability to guarantee per-customer access rate, a single access price of $20-$24 is all that could be charged, the same or less than the cheapest rate for Fibre. It may be possible to increase charges for VDSL2 and VDSL2 Vectoring services, but probably only by 10%.
Right now Fibre Access earns 50% more than Copper, with a 15% kick coming soon.
Comparing Copper and Fibre Upgrades
The current GPON technology used in the Fibre network uses 2.5Gbps transceivers.
Already there are 10Gbps, 40Gbps and 100Gbps transceivers in production. This is not some untested putative "all singing and dancing" laboratory demonstration, but real devices.
We know that volume demand, by the millions as needed for large-scale Fibre Networks, will drive silicon transceiver prices down to commodity prices: the same or less than current transceivers.
The GPON transceiver electronics account for under 10% of the total service cost, under $100 for two ends.
The cost-per-bit of fibre has a proven growth path down to 50-times less than today. That's reassuring to Network Operators. For $100 or less, they can upgrade to 40Gbps, even 100Gbps within 10 years.
Copper, as FTTN, has precisely two upgrades from existing ADSL2 (24Mbps): VDSL2 (to 50Mbps) and VDSL2 with Vectoring (to 100Mbps).
Upgrade costs will be anywhere between 20-100% of the current cost.
$900 buys the upgrade to 24Mbps for everyone.
$50 buys the line-card to upgrade to 50Mbps with VDSL2.
$500-$900/line buys the upgrade to a node for Vectoring. It's noise cancellation that only works if every active line in a node is under control. You can't install it on one or two services on a node.
Note that these are prices only for one end of the line, the Network Operator. The Customer has to bear the cost of the other end. This is not a recipe for happy customers, forcing costs onto them.
Look at the best upgrade, $50 for 24Mbps to 50Mbps. That's $2.00 per 1Mbps, for one end only.
The next copper upgrade, 50Mbps to 100Mpbs (maximum), it's $10 per 1Mpbs for one end only.
Right now, for $100, they get 1Gpbs, or $0.10 per 1Mbps, for both ends.
Looking to the future, this goes down 50-times, or $0.002 per 1Mpbs, or $2.00 per Gpbs: 1,000 times less than the best Copper price.
That's the challenge for Network Operators sticking with Copper: costs per extra 1Mbps have started to rise, and will continue to rise at an increasing rate.
Comparing Copper and Fibre Revenues
While a Fibre CAN (FTTP) is 30% cheaper to run than a Copper CAN (FTTN) and currently gets 50% higher access revenues than Copper by delivering guaranteed access rates, the 1,000-fold less cost-per-bit means nothing, if there is no demand. I contend that Fibre, and only Fibre, unlocks the economic potential of Customer Data Networks because of tiered pricing, with high-end users self-identifying and being charged a premium for access and volume by RSP's.
This is where the Customer Demand Distribution comes into play.
Sandvine data, for the USA, 1H 2013, show the low 50% of consumers account for just 6.4% of traffic. If you dropped them off the network, download volume would barely change and ARPU, Average Revenue Per User (per month), would increase considerably because there's a link between higher line access rates and higher downloads.
People with a need for speed, download and upload more.
The high-end 50% of users account for around 95% of total traffic, more for upload.
The top 1% of users consume 10% of total traffic. These are the people driving demand and traffic growth.
Data Networks for Customer Access are not built for the average user, they account for an insignificant volume of traffic. The customers that drive demand and fill your order books are the top 1-10% of users.
People who base their argument on "I don't need more speed, therefore the whole thing is a waste of time and money", don't understand the economics. They are almost noise in the system.
The long-term average download demand reported by the ABS has grown at around 70% (1.3 year doubling period) for some time, despite the average line access rate being limited to 4.2Mbps. We know that 95% of this traffic is from the high-end users: the ABS data is a good description of the solid, exponential-growth of demand by the early adopters. It is not a measure of ordinary use.
Network Operators will make their money from the top 1-10% of customers who will both buy the highest offered line access speeds and generate the bulk of the traffic. These customers can be charged a premium for both access and volume ($/GB) by RSP's - they have shown a willingness to pay and the NBN Co pricing model encourages this.
Low-end users, the late-maturity and laggards in market-speak, will either continue their current usage patterns and enjoy continually falling prices, or increase their usage for roughly constant ARPU. Some will discover their unique "killer application" and move up the curve above the 50th percentile.
For a Fibre network, the demand for increased volume and access rates will continue, because it is driven by the top 50% of users, especially the top 1%-10%, and the last 15 years of data tells us they are following the six decade old Computing demand curve: the thirst for data keeps growing as more is provided. [See Bell's Law diagram of prices in this previous post.]
Just as the business model of every Telco offering Telephone services is based on the human characteristic, "people love to talk", Computing devices and Data Networks are driven by a similar so-far insatiable human appetite for information and understanding: "people want to know, 'Why?'". This same appetite drives the entertainment and education industries.
For a Copper Network, line access rate cannot be guaranteed. Because of this, the most demanding users, the top 1%-10% cannot be supplied with Copper services they'll pay a premium for: supply and demand cannot be matched, a fundamental market failure.
Because all users will be charged a single price, RSP's will find it very difficult to charge a premium to high-end users, drying up their revenues.
We know from the faster take-up rates of high-end Fibre services and that average monthly downloads are 50% higher on new Fibre services, comparing to the Australian average, that there is significant pent-up demand in the Data Networking market, especially in the important and highly profitable high-end.
This pent-up demand cannot be reliably served or exploited in a Copper network because high-end users cannot reliably be matched to faster services.
As well, there's barely any speed increase options available, and they are increasingly expensive to deploy, versus the exponentially dropping cost-per-bit of Fibre transceivers.
Copper is being pushed from two ends: slower and 1,000 times more expensive!
Comparing Copper and FibreSummary
A Fibre Customer Network will be 30% cheaper to run, already is achieving a 50% higher ARPU than Copper will ever be able to, and has a 10-20% uptick in ARPU coming with 1000/400 Mbps plans. Fibre differentiates the most-profitable high-end users for Retailers, increasing their Gross Margins & Profitability. History tells us, these high-end users will continue to increase traffic and create demand for the next 3 generations of Fibre transceivers already in production.
Input costs per-bit are set to go down 1,000-fold while demand will continue to double every 15-24 months into the foreseeable future. This is a license to print money that can only be attained with a Fibre Customer Network.
The low-end 50% of users get the benefit of price reductions paid for by the high-end users. They can either chose to maintain their current usage and pay less, or pay the same and use a lot more data.
The low-end users get an enormous gift from the high-end. Rather than paying for a network they don't need and don't want, they are getting subsidised by the high-end to a remarkable extent.
No comments:
Post a Comment