HEVC Advance Royalties: Onerous or Not?
With all the brouhaha about HEVC Advance’s content royalties, you would think that they make HEVC commercially unusable. This week, while responding to several consulting client inquiries about the cost per hour of streaming, and the impact of HEVC content royalties throughout the value chain, I had the opportunity to put that assumption to the test.
Specifically, I calculated the economic impact of the content royalties in the decision to switch from H.264 to HEVC for 1080p video and smaller, which still comprises the vast majority of premium and free video distributed on the internet today. So why not look at the 4K market?
First, HEVC is an absolute enabler in the premium 4K content to connected TV market; if you don’t license HEVC, you’re out of the market. Second, while every expense looks small when you’re not paying it, the 0.5% royalty seems minimal for most services, like $0.06/month per UHD subscriber for Netflix’s $11.99 UHD service, or about $0.50 per year for Amazon Prime customers paying $99/year. Should these additional costs upset the finely-tuned profit model of these companies, they can just raise their prices. At the normal 4X cost gross up to recalculate retail price, this would boost Netflix to $12.24/month, while Amazon Prime would skyrocket to $101. Hey, you wanna watch 4K? You gotta pay.
So instead, I wanted to focus on HD and lower-resolution content, specifically analyzing whether the bandwidth efficiencies enabled by HEVC over H.264 would outweigh the costs of the content royalty. Interestingly, in some situations they clearly do, and in others they clearly don’t.
Assumptions and Costs
Here are my major assumptions.
- Royalty costs. There are two high-level costs associated with HEVC playback; the content royalty and the player cost. As I discussed in a recent article, the content royalty is .5% of attributable revenues, which I’ll cover in my calculation. Since there’s no generally available HEVC player, a service looking to deploy HEVC might also have to supply an HEVC decoder, which would be an “other device” in the HEVC Advance taxonomy that costs $1.10 per decoder. Or, the service might send HEVC only when a decoder already exists, like in a future generation Roku, Chromecast, or Apple TV device.
I calculate the impact of the content royalty under both assumptions. The first assumption, which includes only the content royalty, essentially asks, “Does the bandwidth savings delivered by HEVC more than compensate for the content royalty?” The second, which includes the cost of the decoder royalty, asks, “How many units of video do I have to ship to pay for the costs of the decoder?”
- Bandwidth savings. When computing the bandwidth savings delivered by HEVC, I compared a 1080p stream encoded at 4Mbps for H.264 and 2.5Mbps for HEVC. The savings will be somewhat less than this because not all viewers will be able to retrieve the 1080p stream at all times. That said, according to the Netflix ISP index, the average delivery speed in the U.S. for July 2015 was 3.2Mbps, so most viewers should be able to retrieve a fairly high data rate stream.
- CDN delivery charges. For the M-GO scenario, I assumed a delivery charge of $0.01/GB and $0.02/GB. The latter is the lowest price published on Amazon Cloudfront which applies to publishers pushing over 5 PB per month. The former is obviously half that. For Netflix, I assumed delivery charges of $0.01 and $0.005 per GB.
- Player development costs. I don’t include player development costs in the calculation, because those vary widely.
- Encoding costs. I’ll mention encoding costs in all the Netflix example, but don’t include them in the modeling for simplicity. The lowest cloud encoding pricing for H.264 that I’ve seen is Zencoder at 1.25 cents per minute, twice that for HD. I’ll assume Netflix encodes each video into nine streams, five in HD, for a retail cost for H.264 encode 17 cents per minute. I’ll cut that in half to account for Netflix’s encoding scale, then multiple by eight to account for HEVC’s increased encoding complexity, which comes to $.70/minute, or $42/hour.
Let’s look at two models and then touch on some others.
M-GO Model (TVOD)
Under this model, the viewer rents an HD movie for $4.00; rather than delivering with H.264, the service provider uses HEVC. Working through Table 1:
- The three columns on the left show the calculation at 1 cent/GB delivery cost, on the right shows 2 cents per GB.
- In both cases, as discussed, the data rate for the 1080p movies is 4Mbps for H.264, and 2.5 mbps for HEVC.
- The data per 2-hour movies is the average bitrate divided by eight to convert from bits to bytes, then multiplied by 60 to convert seconds to minutes, and then by 120 to calculate the 2-hour data load (4,000Mbps/8 x 60 x 120 = 3,600,000).
- The bandwidth cost applies the per-GB charge to the total data load.
- The rental fee is $4.00, the HEVC Advance content royalty .5% of that, or $0.02. The royalty for H.264, payable to MPEG-LA, would be the lower of 2%, which would be $.08, or $.02 (http://www.mpegla.com/main/programs/avc/Documents/avcweb.pdf). In this case, the royalty would be the same $.02 as HEVC.
Table 1. Application of HEVC Royalty in pay-per-view scenario.
Because of the H.264 royalty, the model is positive from the start, but at 1 cent per GB, it will take 96 rentals to the same client to pay for the player. Throw in the development and encoding costs, and it all feels unrealistic. Let’s see how it looks with the Netflix model.
Netflix Model (SVOD)
Netflix is a monthly subscription service that charges $8.99/month for SD/HD (non-4K) content. The first set of three columns represents a delivery cost of .5 cents per GB; the second 1 cent per GB. (I have no idea how much Netflix pays for bandwidth, but if my numbers are low, obviously the case for HEVC is stronger.)
Working through the table:
- The average Netflix viewer watches 90 minutes of video a day, a very scary number.
- In both cases, as discussed, the data rate for the 1080p movies is 4Mbps for H.264, and 2.5Mbps for HEVC.
- The average data per day is the average bitrate divided by eight to convert from bits to bytes, then multiplied by 60 to convert seconds to minutes, and then by 90 to calculate the 90-minute data load (4,000Mbps/8 x 60 x 90 = 2,700,000).
- I multiplied this by 30 to calculate data per month.
- The bandwidth cost applies the per-GB charge to the monthly data load.
- The subscription fee is $8.99, and I’m assuming that 100% of content is delivered via HEVC. At .5%, the royalty would be $0.045. Note that for subscription revenue, Netflix would pay MPEG LA the princely sum of $100,000/year, which I'm not including in the model.
Table 2. Working through the HEVC Advance royalty in a Netflix scenario.
At $0.005/GB for delivery, HEVC still saves Netflix $0.15 per month for each subscriber, from which you have to subtract the HEVC Advance royalty of $0.045, netting about 10.7 cents per subscriber per month, a very big number when you have 65.5 million subscribers. Even after paying the royalty to HEVC Advance, Netflix nets $7,008,934 per month. Obviously, the numbers are even more favorable at the $0.01/GB delivery cost.
What about encoding cost? According to my colleague Dan Rayburn, Netflix has about 60,000 individual titles, including movies and television show episodes. Assuming an average length of 60 minutes, it would cost about $42/ title to convert these to HEVC, for a total of around $2.52 million. Even if I’m off by a factor of two, it really doesn’t change the result all that much.
If Netflix had to pay the player royalty, breakeven at $0.005/GB delivery charge would be about 12 months; after which Netflix will earn $3.85 per subscriber for the following 36 months. At $0.01/GB, breakeven occurs after about 5 months, and the 3-year take nearly triples to $9.32.
If the savings are so grand, why isn’t Netflix jumping on this model? Several reasons, starting with the installed base. Many Netflix subscribers watch on devices that can’t be upgraded to HEVC, like the Roku box in my living room and my daughter’s iPhone 5.
On computers and notebooks, which could be upgraded, the playback architecture is in transition. Adobe Primetime will support HEVC, but adds another cost to the mix. HTML5 could support HEVC playback, but HEVC playback isn’t currently available in any browser, and likely won’t be for a long time to come. Perhaps Netflix could create its own downloadable player, but that model is also on the wane.
Anyway, the point of this analysis wasn’t that Netflix was ignoring some cost savings, it was to analyze the impact of the HEVC Advance content royalty on different types of deployments. While it does preclude some uses, it absolutely doesn’t preclude others.
Why is the Netflix scenario so much brighter than M-GO? Because Netflix is achieving more than 23 times the bandwidth savings, while only paying about $02.5/unit more in royalty costs.
What does this mean? Basically, that using HEVC to save bandwidth costs doesn’t work in pay-per-view applications if the service has to pay the decoder royalty. Even at 4 cents per GB, which is probably unrealistic for any large commercial service, the average savings per movie is only about $0.054, for a breakeven of about 24 movies if the service has to pay player-related royalties.
What about advertising-funded sites like CNN and ESPN? Using HEVC for the bandwidth savings seems doubtful; most distribute lower-resolution video than 1080p so the bandwidth savings would be less, while the royalty percentage is the same. Ditto for Facebook and Twitter, which use even lower resolution videos. Plus, where there is no H.264 royalty, these entities would presumably owe HEVC Advance .5% of attributable advertising revenue, which will be very difficult to recoup via bandwidth savings in low data rate videos.
Basically, as the cost per GB decreases, the bandwidth savings delivered by HEVC drop. When you incorporate even a very low royalty rate like .5%, the model looks doubtful except for very high-bandwidth, low-royalty applications like Netflix.
[Editor's Note: Netflix’s encoding costs have been revised upwards to reflect additional information provided by industry sources. Table 1 has been revised to include the H.264 royalty which was excluded in the original analysis. Table 2 and Netflix's net savings per month have been revised to remove double-counting of HEVC Advance royalty in the original calculations.]
After widespread criticism of its initial royalty proposal, HEVC Advance has reduced royalty rates and decoupled them from attributable revenue, as well as provided incentives for early adoption and a royalty credit for small publishers.
This year, HEVC has gone from the heir apparent to an unmanageable mess. Meanwhile, open source codec efforts are focused and progressing. What happened?
MLB Advanced Media says the proposed royalty isn't reasonable. With pressure mounting, the new patent pool promises an adjustment.