Part 3: Deep Dive into the Apple CDN

Netflix and Comcast built highly scalable content delivery networks that deliver multiple PB’s/month of VOD content to US audiences. They built their CDNs in a short period of time using open source software, and a few dozen engineers. However, if we compare the complexity of Netflix/Comcast CDN build-outs to Apple’s current build-out on a scale from 1 to 10, with 1 being the least challenging, and 10 being the most challenging, Netflix/Comcast are a 1, and Apple is a 10. Apple’s CDN requires a robust feature set that’s on the same level as most pure-play CDNs. With that being said, Apple’s third challenge is generating the highest ROI possible on its highly scalable CDN build-out, getting the most bang for the buck.

Apple has $160B+ of cash in the bank, and can withstand losses of a few hundred million dollars without a hitch, but building a mediocre CDN platform is going to impact the customer experience, the core principal of Apple’s identity. There are various ways to look at ROI outcomes. Apple may spend $500M developing a killer CDN with all the bells and whistles, however, at that price tag, Apple would have been better off buying Limelight for less money, and probably getting a better CDN in the process. In order to assess ROI, the POP cost is required. A CDN POP is comprised of hardware, software and IP transit. At a minimum, Apple will need to partner with three carriers at each POP in their respective regions.

Tier 1 US Carriers

In the US, there are four major Tier 1 Carriers: AT&T, Verizon, CenturyLink and Level 3. If Apple wants to avoid the Netflix type dilemma, than buying transit from AT&T, Verizon and CenturyLink is a must. My guess is that AT&T and Verizon transit cost is probably more than twice the cost of Level 3. Every CDN in the ecosystem has multiple transit providers at each POP. Not only does it provide redundancy, but it empowers CDNs to have more control over the middle mile, selecting those carriers that perform the best on certain network routes in certain regions. Below is an example POP build-out cost for US/Europe, and some of the more expensive markets like Brazil, India and China. The goal here is to get some basic idea of POP cost.

Annual Cost for US/Europe CDN POP
  • 100 Servers x $6k = $600,000
  • 2 Brocade Routers x $15k = $30,000
  • 2 Load Balancers (Appliance) x $15K = $30,000
  • 3 Cabinets x $2k/month x 12 months = $72,000
  • 50Gbps Commit x $2.00/Mbps x 12 months =  $1.2M
  • 50Gbps Commit (AT&T) x $5/Mbps x 12 months = $2.5M
  • 50Gbps Commit (Verizon) x $5/Mbps x 12 months = $2.5M
  • Miscellaneous x $1,000 x 12 months = $12,000
  • Total annual cost = $9.2M
Annual Cost for Non-Competitive Markets (e.g. Brazil)
  • 100 Servers x $6k = $600,000
  • 2 Brocade Routers x $15k = $30,000
  • 2 Load Balancers (Appliance) x $15K = $30,000
  • 3 Cabinets x $2k/month x 12 months = $72,000
  • 25Gbps Commit x $40/Mbps x 12 months =  $12M
  • Miscellaneous x $1,000 x 12 months = $12,000
  • Total annual cost = $12.7M

I understand that some POPs will require more servers and bandwidth than others, and some cabinets are more expensive in certain regions than others, but taking an average provides us with a general idea of POP cost. Based on the estimates above, the average annual cost for a fully functional production CDN POP is about $9M/year in North America/Europe, plus or minus a few hundred thousand, and $12M/year in other markets.

Of course, Apple can always start with a fewer number of servers and lower bandwidth, but eventually overtime, as they ramp up to accommodate demand, the cost will increase dramatically. If Apple builds out 100 POPs, that’s going to cost about $100M. With a 600M+ Apple mobile users, they are probably going to need 5x to 10x the number of servers of what was estimated above. Isn’t just easier to buy an existing CDN at that price tag, or build out a CDN in the US, and continue the partnerships with the existing CDNs. The Apple CDN team definitely has their hands full.