State of the CDN: More Traffic, Stable Prices, More Products, Profits - Not So Much

CDNs (content delivery networks) are the secret shadow super powers behind the web and Dan Rayburn at streamingmedia.com is the go to investigative reporter for quality information on CDNs. Every year Dan has a Content Delivery Summit on all things CDN and those videos are now available. Dan also gives a kind of state of the industry talk where he does something wonderful, he gives real numbers and prices. Dan really knows his stuff and is an excellent speaker, so watch the video, but here’s my gloss on the state of the CDN so far this year:

  • Massive growth. Large customers are expecting 126% growth in video traffic over last year; medium size customers are seeing 48% traffic growth, small sized customer are seeing 73.3% traffic growth.
  • More traffic != More profit. Traffic growth doesn’t lead to more profit because the traffic growth is concentrated in larger customers that can make the best deals.
    • Video takes up the largest amount of traffic on a CDNs network but gives the least amount of profitable revenue.
    • Most vendors do not make money purely on video business because it has become commoditized. YouTube does it for free. What is not commoditized is performance and reach, but the actual process of delivering bits across a network has become well known so is now a commodity.
    • It’s the economics of scale. It costs a lot to build out capacity and then they have to fill it.
    • CDNs are trying to diversify revenue away from commoditized video and like Akamai, are adding more value-add services.
  • Regional CDN vs Global CDN. CDN as a word has lost a lot of meaning. What’s important is to select a CDN based on: delivering the right content to the right user at the right location on the right device.
  • Using multiple vendors. It’s common for large customers like Netflix, Major League Baseball, and Hulu to use multiple CDN vendors.  
  • Bundled vs Unbundled. Content owners increasingly are bundling together contracts for small object, large object, streaming, live content, front-end acceleration, dynamic site acceleration, and other services.
    • You don’t want a price per service where you pay on a per GB delivered model at the end of the month, instead pay on a per Mbps sustained model, or what is called “95 5” or 95 percentile in the industry. So you get one price per meg sustained.
    • This gives vendors a higher margin on average because they don’t have to lowball video, but it’s simpler and easier for the customer.
    • Most unbundled contracts are 12 months in length but bundled contracts are 18 months.
    • As more vendors offer more services we’ll see more bundling across vendors.
  • Expects prices to stabilize. Pricing is down this year compared to last year. Expects pricing to firm up as pricing has nose dived the last few years and CDNs aren’t making money. They can’t go much lower on these deals.
  • Large - prices for customers paying $1M+ per year:
    • These are the largest customers like Netflix, MLB, NHL, CBS, Fox, NBA, and Hulu. Google has built their own CDN.
    • Seeing 126% growth a year and a 18.9% price reduction.
    • 3 petabytes a month pays between 1-3 cents per GB delivered.
    • 400Mbps a month pays between $2-$4 per Mbps sustained.
  • Medium - prices for customers spending between $250-$500K per year:
    • These are the smaller customers and a majority of customers on a CDN's network.
    • Seeing 48% traffic growth a year and a 11.4% price reduction.
    • 500TB-1PB a month pays 2-6 cents per GB delivered.
    • 200Mbps a month pays between $6-$11 per Mbps. Pricing is much more variable.
  • Small - prices for customers spending between $100-$250K per year:
    • Seeing 73.3% traffic growth a year and a 10.6% price reduction.
    • 250TB-500TB a month pays 4-12 cents per GB delivered (6 cents average)
    • Not enough data as too few customers in this range using a Mbps model.
  • Micro - prices for customers spending < $100K per year:
    • Prices are all over the map so there’s no solid data.
    • Customers of this size probably don’t need a CDN. They can use a regional CDN that delivers content in a particular region, like the US or Asia, at a fraction of the cost of a global CDN, especially for services like video.
    • Before Netflix expanded outside they didn’t need a global CDN.
    • Amazon dominates. Expected to put pricing/negotiation pressure on all the other CDNs.
  • Not a big fan of mobile video. Not a lot of adoption because the device market is so fragmented.
  • Online video advertising is over hyped. Most is on paid content like MLB.
  • Everything is moving HTTP. A common infrastructure means cheaper pricing.
  • Too much hype and noise in the CDN market:
    • Terms like edge and core are marketing terms. They don’t mean anything. Most companies never define what the edge is. Just because you are in the edge/local ISP/last mile doesn’t mean there’s better performance.
    • Quantity != quality.
    • Only customers know the real performance. Many measurement services only measure backbone performance, which is like measuring traffic on the highway. That is not the real world. We go to an ISP, not the backbone for internet connectivity, which is like traveling on back streets. True testing involves the last mile. As a customer create your own metric and test the CDN with your own content.