by Guido Meardi
Why should you read this?
Do you ever wonder why operators in the media industry, and especially New Media players, are so keen on royalty-free codecs? Do you see little difference between per-device patent royalties and per-use supported SW licensing? Do you believe that VP9 and – when it will be available – AV1 are the only royalty free codecs in sight?
If the answer to any of the above questions is yes, reading this blog should be time well invested.
* * *
It recently dawned on me while talking to our partner Oliver Gunasekara of NGCodec that many of the commentaries about royalty-free codecs are often neglecting the very core of the problem, i.e., the reason why the media industry – and especially New Media operators rather than traditional broadcasters – were so keen to invest significant resources to develop a royalty-free alternative to MPEG codecs.
After all, why would codec royalties be such a big issue now, while at the same time operators had hardly noticed them in their P&L’s for the past decades?
The global video industry is quickly approaching a trillion USD per year of revenues, figure estimated by combining subscription revenues to advertising revenues and not even accounting for ancillary video-rich industries such as social networks, gaming, smart cities, connected vehicles and/or video conferencing. Codec royalties – connected to the effectiveness of the very “core” that makes video delivery work and that drives the large majority of infrastructural CAPEX and OPEX – account for less than 1% of the top line. Why then are cash-flush companies like Google, Netflix, Amazon & Co. so keen on royalty-free codecs? Are they truly fearing the impact of codec royalties on their sizeable margins? Do they believe in the sustainability of a model where valuable innovation resulting from large investments, hard work and ingenuity is given away for free?
In short, no, they are not crazy. They are actually being commercially sensible, and are rationally responding to the way in which the market is evolving. They look at their data, and drive sensible inferences.
Working with large scale innovative customers around the world provided us with valuable insights on the priorities of operators seeking fast revenue growth in OTT video delivery, typically equipped with good analytics, vs. the priorities of the ones that have not yet had a chance to carefully analyse their traffic data.
For instance, many of those data-rich operators violently nod with a large smile when I show them V-Nova’s analyses on the “reality of data networks”, which I introduced in my previous blog “The Codec War: a Battle with Many Winners”. The gist of that data is simple: available bandwidth in the richest countries is possibly more bottlenecked than what most people would expect from developing countries. About 30% of video streaming sessions – especially at prime time – happen at bandwidths that do not allow reliable monetizable services (and hence most users switch off), and for another 50-60% of sessions video quality is way below the quality that today’s users expect, again with impact on watching time, churn or both. These figures are hugely divergent from many official claims and assumptions, as well as from the rightfully proud feeling that “Now it finally works, some of the time”: it is a phenomenally big deal – with lots of second-level implications – for operators that want to make money in digital video with TV-like experiences.
Data network bottlenecks are not the only thing that many operators are noticing. Another element with game-changing implications is that the majority of video viewing is inexorably shifting from Big Telly to mobile devices, creating a clear polarization: large screens are (and will continue being) used for certain types of film and TV content, and the rest is mostly watched on HD-capable mobile devices. Ericsson reckons that by 2020 over 50% of all video watching will be on mobile screens. At YouTube, mobile accounts already today for more than half of total views. Facebook registers more than 1.1 billion mobile daily active users, driving the large majority of their usage. Netflix made several strong statements about the growing importance of mobile, and indeed their watched hours in mobility grew by over 35% in the past two years. Incidentally, Netflix also mentioned multiple times the importance of finding ways in the future to deliver HD video below 400 Kbps, so we are probably looking at traffic data in the same way. In fact, PERSEUS enables operators to do that already today, also for live streaming, compatibly with 100% of devices, browsers and streaming protocols. (yes, I know, apologies, parent’s pride, end of the ad)
So what, you may say? What does this all have to do with codec royalties?
Well, the point is that the large majority of the 2 billion mobile devices sold each year are not iPhone X’s or Samsung S8’s. The mobile market is no longer an early adopter market for tech enthusiasts, but a mass market: most viewers, and especially young people, mid-income people and people living in developing countries – all relevant market segments for social networks and ad-based services – are satisfied with good devices selling at prices between 50-100 USD. It’s a high-volume play, where manufacturers make limited margins on extremely high volumes. Also, the market is extremely elastic: even just a few extra dollars in the price can have material impact on sales volume.
Taking into account what above, it is clearer why a PayTV broadcaster with two-digit monthly ARPUs may not be scared to pay a few dollars of HEVC royalty per set top box, while mobile device manufacturers simply cannot afford to activate the HEVC hardware decoding option of the chips they put in their medium-to-low-end devices, otherwise they would either forfeit their margins or slow down their sales volumes.
In fact, as of 2018 (6 years after the HEVC syntax was defined and frozen) iPhone X’s and S8’s indeed support HEVC hardware decoding, but even several smartphones priced above 150 USD do not activate the HEVC hardware decoding that would otherwise be available in their silicon. On devices without hardware HEVC decoders, operators can implement a software HEVC decoder, but then they will be very limited by the processor, and most budget phones can’t play good quality HEVC videos without seriously taxing the battery.
Google & Co., who make good money from their services and need their market to develop, cannot afford the mobile market to be slowed down: they know that the performance of h.264 is not enough to enable OTT video services at sufficient scale, and at the same time they know that HEVC is not going to be economically viable for the majority of decoder devices. They needed a royalty-free alternative – i.e., a codec that would not slow down device ecosystem growth with per-device royalty costs –, offering at least the performances of HEVC, and if possible better.
In all of this, the interesting thing to notice, which it dawned on me I never stress enough, is that royalty-free does not necessarily mean license free.
Everybody understands that valuable code-optimized, reliable, maintained software solutions (whether closed-source or open-source) must come at a cost: some fee must pay for the salaries of the people working on it, and volunteering work can only go so far. If an operator wants to deploy a codec such as VP9, they will either have to pay a company that packages the codec into a deployable supported solution, or they will have to assemble and maintain an internal team of programmers, dev ops, quality assurance and service professionals dedicated to that effort: either way, it costs money, there’s no free lunch. The point, however, is that software licensing is usage-based paid by the operator, i.e., you pay for it when you use it and thus when you have demonstrable benefits that more than make up for the extra costs, while royalties to enable dedicated hardware acceleration are per device and ultimately paid by the user, before the user even turns the device on(unless it’s a managed payTV service, in which case we would be back to the usual case where the high-ARPU payTV operator can happily absorb the royalties).
So VP9 and AV1 pursued the royalty avoidance route by giving away for free the patents that allow chipmakers to design hardware acceleration for the codec. In so doing, the whole design is free, which allows any solution developer to develop paid commercial solutions based on it. However, as brilliantly articulated by Leonardo Chiariglione, Chairman of MPEG and one of the greatest contributor to the development of digital video, in the long term this model eliminates any incentive to invest the hundreds of millions of dollars and tens of years of life that are necessary to research and invent new encoding formats, since profit making would be left only in the productization and maintenance part of the value chain.
There is another route to fulfil the device-royalty avoidance vision, though.
What if a next gen codec format was developed such that, instead of requiring new dedicated hardware blocks, it leveraged existing general-purpose hardware blocks that are already available in all silicon chips, accessible without requiring OS modifications, even by code as high in the software stack as java script running in a browser? Such a codec would not require any enablement royalty, because all the hardware acceleration and the enablement would come for free, as part of every general purpose video capable device. It would be pure software, with a structure of usage-based royalties & licences in line with that of all the other many software components used – or not used – by digital services (e.g., CMS, packagers, media players, ad insertion, recommendation engines, DRM, audio codecs, User Interface middleware, analytics services, CDNs, etc.). Importantly, such a codec would implement the per-device royalty avoidance vision TODAY, rather than in 4-6 years (minimum time necessary to feature freeze AV1, implement silicon chips including the necessary hardware acceleration, commercialize new devices with said chips, waiting for the majority of the installed base to be replaced with said new devices).
Yes, if you got this far in the blog you probably have already guessed it. Such an intriguing option indeed exists and it’s called PERSEUS. It doesn’t cost anything to enable PERSEUS because video capable devices are already enabled to use it. Of course deploying it and using it does cost a small license fee (like it would be the case – in cash or kind – for any other supported codec), which typically pays back after a few days of operation. It is easy to try out, and because it is a light software it can be disabled at any time or used just for part of the services, although it would make little commercial sense to do so.
* * *
To sum it up, royalties and licenses are often used as synonyms, mainly because in the context of IP licensing they are effectively the same thing. Incidentally, this is exactly the context that people in the media industry associate video codecs with, mainly due to decades of history of codecs being standard-based and purely IP-licensed. Trying to artificially distinguish royalties and licenses within that context may lead to confusion. What we are talking about is a different context, i.e. that of a software-driven solution. In that context, royalties and (software) licenses are two concepts which are related, but not the same. Importantly, there is no need to equip general-purpose hardware with specialized royalty-bearing hardware blocks. In this new context, PERSEUS positions itself differently from any alternative video codec, effectively “bundling” the royalty into the software license, or seen another way, “waiving” the royalty on the IP when using the PERSEUS software.
In particular, these are some of the consequences of such a category-redefining model:
- PERSEUS does not require any device enablement royalty, since hardware acceleration is already available and accessible in all video capable devices, without requiring specific enablement.
- It delivers already today compression benefits unmatched by those achieved or targeted by any current or upcoming alternative, for both VoD and live, with a new S-curve of further improvements in the pipeline.
- It is immediately deployable at scale with minimal integration effort, and does not require duplicate transcodes and/or workflows to independently serve old and new devices, since a single workflow can serve all devices – even old ones such as legacy set top boxes –, as well all HTML5 browsers via plug-in free scripted decoding.
- It generates significant encoding density benefits both in software and in FPGA, unlocking commercial viability of 1080p60, UHDp60, 8K, VR 4K/8K, and in general allowing hyper-scale operators to reach today levels of OPEX efficiency and power consumption that only ASICs could otherwise achieve (although at the cost of tens of millions of investments, 18-36 months of development lead time and, substantially, years of feature freeze).
- It is compatible with a virtuous profit-generating business model that maintains the right incentives to keep investing in research, innovation and format improvements. Like any other IP-rich software component.
OK, I know, that list may look like another ad break. But it’s not, really: it is the best way I could find to illustrate how PERSEUS does deliver on the royalty free vision devised by Google, Netflix, Amazon, Mozilla, Cisco & Co. to support industry growth, while at the same time delivering also on the sensible MPEG guideline of enabling business models that incentivize continuous investments in research and innovation to further advance the state of the art.
We are living exciting times of renewal and growth. I’m sure that there is even more to this issue than I have covered in this blog: feel free to comment below and let me know your thoughts.