With the launch of the Disney+ streaming service taking place November 12th, there has been a lot of speculation on Wall Street on what the value of that video traffic might be to Akamai and other CDNs. Two weeks ago, PiperJaffray put out a report saying that by 2024, the Disney+ traffic business “could result in $68M in revenue for Akamai.” And when combined with traffic from Hulu and ESPN+, Disney could be a “>$100M Media customer by 2024.” I’ve been getting so many questions about these numbers that I thought a blog post that details some of the costs and deployment details would be helpful. There are also some statements made in the Piper report that are inaccurate and need correction.
No one, including Disney, knows what the company will spend to deliver Disney+ video traffic five years from now. It’s high-level speculation since we don’t how many subscribers they will have, how many hours each subscriber will watch, what the average bitrate will be across all devices, and what percentage of Disney’s traffic will be served by third-party CDNs. Disney could build their own CDN long before 2024 if they wanted to and they could also work directly with ISPs to cache Disney+ content inside last-mile networks, like many of the large OTT platforms do today.
It is inevitable that at some point Disney will do it themselves, especially since they have the in-house expertise to go DIY and Disney+ content is all on-demand. With the expectations for the success of the Disney+ service, scale and economics will dictate they go DIY. Some OTT providers have suggested that when a service gets to 20M subscribers, that’s the threshold from a size and scale standpoint of when it makes sense to consider building your own CDN. Piper’s report says the probability of Disney building their own CDN is “low” suggesting “it would take >5 yrs for Disney to get full distribution.” Of course, that’s not the case at all and Piper doesn’t understand the costs or time involved in deploying a purpose-built CDN.
One of Piper’s arguments in their note against Disney going DIY is the argument that, “Akamai has been transparent that it would take a new “do-it-yourself” (DIY) service at least five years and over $2B in capex to come close to being able to deliver static content,” but they are taking what Akamai has said out of context. At no time has Akamai suggested it would take Disney $2B to build out their own CDN. That reference Akamai has made in the past is their belief on what it would cost a company to do a DIY offering that would compete with Akamai from a services standpoint, with Akamai’s scale, which of course isn’t what Disney would be building. Disney could build out a CDN for the delivery of Disney+ content for under $100M in initial CAPEX costs. It’s not costly or difficult for Disney to do, especially for VOD content, hence why companies like Blizzard, Apple, Netflix and others have done it.
Disney Streaming Services are like the Special Forces of the video industry and are the best at what they do. They built the first OTT video service in the market 17 years ago with MLB.TV and if they wanted to build their own CDN, they could do it quickly and cost-effectively with a great quality of service. Piper also suggests that [third-party] “CDNs now provide a lower-cost option compared to the pre-DIY wave“, but that’s not what Disney cares about. Cost isn’t their focus, quality is. Disney isn’t looking for “lower cost options”, they want the best user experience they can provide.
Disney will go DIY with their video delivery, the question is when. For Piper to suggest the probability is “low”, that’s a bad bet to take. What Disney’s DIY deployment will look like from an architecture standpoint is unknown, but come 2024, third-party CDNs will not be delivering the majority of Disney+ traffic. It should also be noted that the senior technical team at Disney Streaming Services are behind an open caching initiative via the Streaming Video Alliance and not only helped write the technical spec, but Disney’s CTO for their Streaming Services group sits on the board of the SVA and is the President. There is a reason why Disney Streaming Services are helping to design and push an open caching initiative, because it would help themselves and others in the industry.
When Disney+ launches multiple CDNs will be used to stream and download the on-demand videos. Of all the CDNs, I’m told CenturyLink (formerly Level 3), has the largest percentage of Disney’s traffic as it stands today, based on volume. It’s widely known that Disney Streaming Services and CenturyLink have had a great working relationship for a very long time. Akamai is also one of Disney Streaming Services primary CDNs and I expect when Disney+ launches internationally, Disney will use at least four CDNs including CenturyLink, Akamai, Limelight Networks, and Fastly, with the latter two being critical for Disney’s international launch. I would not be surprised if Disney also uses Fastly’s Media Shield product, which would allow Disney to optimize their multi-CDN deployment across all providers.
What percentage of traffic each CDN will get for Disney+ traffic is unknown, because for Disney, service quality is paramount and they will shape traffic across a multi-CDN strategy based on how CDNs perform, just like they have always done. So traffic share will shift amongst CDNs based on performance, potentially as often as each day. The PiperJaffray report says that Akamai’s management team expects to receive the “lion’s share” of Disney+ traffic, however their report doesn’t quote an Akamai person by name and it contradicts what I’ve been told directly. I have personally heard Akamai say they will work hard to try and get as much of Disney’s traffic as possible, but at no time have they implied or suggested that they “expect” to get the “lion’s share”. If Piper is going to quote Akamai on that, I suggest they give attribution, otherwise it’s hearsay.
With all that said, let’s look at the numbers in the PiperJaffray report and break those down based on the estimates they used. Disney has said they expect between 60M-90M subscribers for Disney+ in five years. The report also suggests users will watch two hours of video per day and that by 2024, 9% of the traffic will be in SD, 61% in HD, and 30% UHD. Based on that blend and hours of viewing, PiperJaffray expects each user to consume 2.4Tbps per year, or 200GB per month. But nowhere in the report does it say what bitrates they used to come up with the total bits delivered number or what percentage will be on a smaller screen. The bitrate for HD on an iPad versus a TV is very different in size so just saying 61% of viewing in HD isn’t detailed enough. They do say that “HD and UHD generate ~3.6x and ~8.4x more bytes compared to SD”, but they don’t break out what bitrates they are comparing that to. 3.6x more than what?
But using their numbers and the average bitrate per SD, HD and UHD, that most would default to in the market today, (SD 800Kbps, HD 4Mbps, UHD 12Mbps) the total number of bits delivered would equal 158GB per month, or 1.9TB per user, per year. That’s 20% fewer less bits than they suggest. And on top of the lower number, videos delivered to mobile typically take up 70-75% fewer bits than to a TV screen. So you have to factor in the percentage of hours each month that a user will watch on a mobile device, something the Piper reports doesn’t include.
Even if you estimate that only 10% of total viewing time per user, per month, is on mobile, that drops the total number of GBs delivered per month to 120GB, per user. That’s 40% fewer bits than the Piper report suggests. And that’s without factoring in the possibility that five years from now, Disney+ content could be encoded using Av1, which could reduce the current number of bits delivered by at least 30% or more.
On the pricing front, the Piper report is way off on what Disney would pay on per GB delivered model. The report says, “we believe in today’s pricing world, Akamai would be charging ~$0.006 globally per GB per month to deliver this volume of traffic.” That number is too high. The report goes on to say that, “Given typical annual pricing contraction in the space is 20-30%, our base case assumes Akamai would be charging $0.0011 globally on average per GB per month in 2024. Our downside and upside cases imply 30% and 20% annual pricing declines, resulting in $0.0007 and $0.0016 per GB per month, respectively.” This is way off. Piper is using incorrect numbers to make their estimates both on total GB delivered per month, per user, and the cost per GB Disney would pay today, and five years from now.
Based on Piper’s numbers, they say Disney would spend $196M in total in 2024, to deliver Disney+ traffic to 75M subscribers and that Akamai would get 35% of that traffic, making the business worth $68M in revenue to Akamai in 2024. There’s also an odd reference in the report to Akamai’s pricing saying that Akamai is “charging ~$0.006 globally per GB per month to deliver this volume of traffic“, saying that pricing “would represent an over 80% discount to what lower-quality competitors are offering at scale today.” Akamai isn’t at an 80% discount on any of their services and Piper doesn’t define what they mean when they say “lower-quality competitor“? There is also another instance where Piper says, “with Akamai expected to receive the “lion’s share” [of Disney+ traffic] (not necessarily the majority share.)” Lion’s share means, the largest part of something and majority share means, the greater part, or more than half, of the total. So how is the lion’s share not the majority share?
The Piper report also makes multiple references to Netflix trying to compare pricing to their delivery before they went DIY, which is simply incorrect. In one instance they say, “When Netflix chose Akamai to deliver its streaming traffic in 2010, IHS reported that for the first few months, Akamai would charge Netflix $0.015 per GB per month. After the promotional period, the price would increase to $0.06 per GB per month.” In 2009 I wrote a detailed blog post on Netflix’s cost to stream their videos and the average price point from third-party CDNs was $0.03, or half of what Piper suggests. [Detailing Netflix’s Streaming Costs: Average Movie Costs Five Cents To Deliver] The average bitrate in 2010 wasn’t anything near what it is today, nine years later. Piper is using data that can’t be correlated or compared to today’s market conditions in any capacity.
The report also brings Disney property Hulu into the mix saying, Hulu could be a ~$33M customer for Akamai by 2024. All-in, Disney’s properties of Disney+, Hulu, and ESPN+ could result in Disney being a well-over $100M Media customer by 2024.” But nowhere do they break out how they come up with the $33M revenue figure, especially considering some of Hulu’s traffic is live, which has very different pricing methodology behind it and very different data consumption.
Piper also suggests that, “one trend we could see over the next few years is the shift of Akamai’s Media business for OTT and online gaming specifically shift towards subscriber-based pricing.” Their rational for that is that cloud and email security companies have historically been successful with this pricing model. What does email security have to do with OTT video delivery? Beats me, but OTT providers are not going to move to a per-user pricing model for purchasing video delivery from third-party CDNs as they won’t want to pay for something a user might not use. It’s not an efficient model at all for SVOD OTT providers.
So what is the value of Disney+ video delivery to third-party CDNs in 2024? No one knows since we don’t have enough details on the service. But looking at real numbers like average viewing hours per month, a realistic split of viewing on mobile versus large screens and the fact that the traffic will be split out amongst multiple providers, in different regions of the world, we can take a more realistic guess at the numbers for next year. Based on 15M subscribers, with 30% of viewing on mobile, and 60 hours of viewing a month, with 10% of usage being in UHD (all of which are high estimates for viewing and UHD split) the total spend for Disney in 2020, to deliver just the video bits to 15M Disney+ subs, would be about $4M in total. [Note this is per month, so $24M for the year] [90GB per month, per user, at $0.0025 per GB delivered]
Note that the $4M per month figure would be just for pure bit delivery and Disney could have some additional charges on top of that for other functionality or services tied to the delivery of video.
If you have further questions on this topic, feel free to reach out to me at any time. firstname.lastname@example.org or 917-523-4562.
Disclaimer: Outside of Netflix, Apple and Facebook, I have never bought, sold or traded any shares in any other public company and have never owned shares in any CDN.