Rethink's Alex Davies Talks New Transcoding Report and Future of VVC, AV1, and LCEVC
Contributing Editor Jan Ozer recently spoke with Alex Davies, senior analyst at Rethink Technology Research, about Rethink’s new report, “The Media and Entertainment Transcoding Workload and Device Royalty Forecast for 2020 to 2030,” and what it says about the future of VVC, AV1, and LCEVC for streaming publishers.
The conversation opened with Ozer reminding Davies that “back in 2021, you were pretty bullish on BBC,” and asked what had changed. Davies was blunt. “It's been very disappointing, I think for the industry,” he said. When Rethink first built the model, “the methodology was to look at HEVC, look at AVC, track the historical patterns, and predict forward from there.”
Figure 1. Once bullish on VVC, Rethink now shows it falling behind AV1.
That logic assumed VVC would follow the same rough trajectory as the previous generations, which hasn't happened. Davies noted, “We're five years past VVC being sort of finished, six years past when it was essentially finalized. And I think it's very telling that Qualcomm still hasn't put it into silicon.”
Why? In Davies's view, “There's no real market pull for VVC. So, companies like Qualcomm that bear the additional royalty costs won't integrate VVC into products if they don't have to. And so it's a Catch-22, or chicken-or-egg classic kind of scenario.”
Figure 2. Device CPU horsepower may obviate the need for software decoders.
He tied that stalled adoption to two bigger shifts in the delivery environment. Since “2015-ish,” he said, “we've seen fixed line and mobile broadband bandwidth rockets upwards, like well past the minimum requirements for streaming. At the same time, “the on-device processing power for enabling software decode is also shooting upwards.”
That combination of more bandwidth and client-side compute means VVC no longer depends on silicon support to be technically possible, at least on high-end devices. As Davies put it, “if you really needed software VVC decode, you can probably do it on flagship devices now and probably last year's flagships too. That's what Qualcomm's IBC pitch was: "Hey, look at these cool software decodes that we're doing with Tencent.”
Figure 3. The bandwidth available for mobile and home delivery may reduce the need for more advanced codecs.
Methodology, Data Sources, and Report Audience
Ozer then asked about data collection for the report. Davies explained, “The underlying data set is video devices,” he said. “Based on the kind of platform they're on, how many devices are sold each year, and how many devices are in use. You can begin to model an estimate of usage, because you can say there's this many hours of content being watched on this kind of device, and work your way up from there.”
Research input is not limited to numbers. “I spend a lot of time interviewing industry professionals at shows. I'm just back from DeMux, which was all engineers.” Public sources contribute as well. “There's useful public data like Bitmovin reports that capture responses from more industry professionals.”
In terms of who buys the report, Davies explained that Rethink TV is a subscription service and that subscribers include “technology licensors and streaming services, as well as OEMs and ODMs.”
Key Takeaways for Streaming Publishers
Ozer then asked what the report means for streaming services directly. Davies put it plainly. “If you're a streaming publisher, you should be aware that there are new streaming content pools that are going to start knocking on the door and asking you to pay them.” These include Access Advance's Video Distribution Pool and Avanci Video. In terms of patent composition, Davies noted that "both have considerable overlaps and members, and they are monetizing the same patents in many cases. There’s a mechanism that prevents you from paying twice.”
Then he introduced a third element, LCEVC. “There's also LCEVC, which isn't a separate codec but rather an enhancement layer that adds additional compression in software.” Though initially focused on content providers, Davies noted that LCEVC's licensing scope is expanding. “The other LCEVC piece is that they have also just decided to target devices, TVs, and set-tops. So that's new.”
Access Advance has published its royalty pricing, but as Davies explained, "if you're an AVOD or FAST service, there are three different components which determine the pricing. It’s definitely worth finding out which tier of fee you're going to fall in.” (Note: Ozer interviewed Access Advance about pricing in a video you can access here).
Avanci pricing remains unpublished. “Avanci hasn’t published its pricing yet. They’ve said we’re very close to announcing the first licenses, and when we do that, we’ll have the pricing. And when they do that, I’ll be able to go back and update.”
Is AV1 Next?
The discussion turned to AV1, with Ozer asking, “You're saying VVC isn’t coming. Does that mean AV1 is?” Davies noted that AV1 device support has grown since the last report. “AV1 availability on devices has improved significantly, making AV1 software decode feasible for many playback scenarios.” Hardware decode is also available in multiple ecosystems, including all iPhones after 15. Davies also noted that software or hardware decode was available on all Android devices sold after 2022. [Author's note: during our fact check, Davies added that "Netflix has stated that 88% of large screen devices like game consoles and smart TVs support AV1 hardware decode."]
Nonetheless, Davies framed AV1 as an option, not a replacement. “If you're delivering AVC and don't want to implement HEVC, AV1 might be more of an option now than it was a couple of years ago. But again, if you're already streaming HEVC, there's not a strong case for moving to AV1. It depends on the devices you're delivering to.”
Ozer pressed on usage. “What's your sense of HEVC usage? I know it's big for premium content producers distributing to television sets, but what about mobile? Is it the codec of choice, or is AVC still it for mobile?”
Davies replied, “There's a surprising amount of AV1 and VP9 happening, and that's essentially because of Google and Meta being an advocate for it.” He added his own experience. “Anecdotally, trying to make any of my devices deliver an AVC stream is difficult. If it’s a large screen, it tends to go to HEVC, and if it's mobile or web, it always seems to be AV1 or VP9.”
He recognized both sides of the trend. “But the availability of AV1 is surprising, I think. And sure, it's an uncomfortable sort of royalty scenario with hostility between Sisvel and AOMedia. But we haven't had that big legal showdown that I've been hoping for since 2022.”
Ozer then asked which services are actually using AV1 beyond YouTube, Meta, and Netflix, and Davies didn't know of any. “Outside of that group, I don't think it's much. There are a lot of viewing and device-type considerations you can make. If you are a video streaming service that has budget, you're almost certainly using HEVC now to deliver to smart TVs and dumb TVs with a connected TV device.”
VVC in China and the Myth of Software Decode
Ozer shifted the discussion toward China. “What do you know about companies using VVC with software decode in China?”
Davies was cautious. “There’s been some pretty compelling demos, but I don't think it really extends much beyond that. This ties back to what I was saying about the bandwidth, where if you need to deliver millions of streams, then being able to get a double-digit reduction in size is good.” The tradeoff depends on the use case. “But the bigger question, the more existential question, is if you're trying to deliver short-form video, where immediacy of playback is your main concern rather than stream size, then that's a different set of priorities. Moving from HEVC to VVC doesn't do much for you in terms of playback starting.”
The real bottleneck is not compression. “It's more the packaging side. To be honest, that's a little bit beyond my skill set, but that's a longer term question as well. How quickly can you get something going?”
Ozer pressed for clarity. “I had always taken the fact that there were Chinese companies using VVC with software decoders as a reality. Are you saying that may not be a reality, or you don't know at this point?”
Davies responded. “Just that in relative terms, it's fractional. It's a small amount of mobile devices. It's still a fractional rounding error in large screen. Sure, Huawei is probably going to be the first flagship smartphone with VVC, but it's going to be sold predominantly in China. And if it's not sold in China, it's going into markets that generally have lower ARPUs, so I don’t see much lift for it. It might be different once Qualcomm adds VVC, but I've been saying ‘once Qualcomm adds VVC’ for more than four years now, and Qualcomm still hasn't done it.”
Ozer then pushed on the issue of software decoding. “Are there any mainstream publishers you're aware of using software decode for either AV1 or VVC in any jurisdiction? Or are they basically saying, look, we want hardware because it preserves battery life and ensures quality of experience?”
Davies did not hesitate. “I think it's the latter. There’s a need to be flexible and to show that you can move if you need to, especially if you want to throw your weight around in license contract negotiations. But if the hardware decoder is available, you are essentially a fool if you don't make use of it. Retooling a workflow to use a more inefficient software decoder doesn't really make commercial sense.”
Davies posited that smaller services are different. “A very small niche service might not be able to take such a grand stand, but for the big services, hardware is the way to go.”
Device context matters. “If it’s a phone, you can hope to have a bit more flexibility through the app. But if it's a browser, it's slower. And if it's a smart TV, you can't deploy software to it. So do you trust what's there? Do you trust a software implementation that got pushed out three quarters past its target? Or do you stick with the devil you know, which is the hardware one?” His conclusion was blunt. “So, currently, no. There's subtle interest, but no.”
Bitmovin Forecasts vs Market Reality
The discussion then turned to the validity of codec-related data compiled in Bitmovin's highly respected Video Developer Reports, which Rethink cited in its report. Ozer statued, "I love the Bitmovin reports, and it's great data, but their predictions for codec usage just seem incredibly optimistic. Davies acknowledged the disconnect. “You have to dig around and ascribe a value or trust to these things. But I think the more important thing in the Bitmovin codecs-used-in-production question is the fact that AVC is continuing to fall, and as long as AVC is falling, it implies that other usage must be increasing.”
Then he tied data trends back to software behavior. “But what’s interesting is that VVC usage declines through 23, 24, 25. And so does EVC, which is basically nothing now. AVC is flat. VP8 is falling. VP9 and AV1 are increasing. I think the best way to join those up is that as a video service provider, you are now focusing more on the codecs you prefer. It’s not that you need to look at six. You need to look at four.”
AV1 Royalties, Stalemate, and Limited Visibility Ahead
Ozer returned to royalties, noting that in 2022 Davies predicted there would be clarity on AV1's royalty-free status, which hasn't yet happened. Ozer asked, "When do we expect to see clarity on whether AV1 is royalty-free?”
Davies was unsure. “I’m not sure we’re going to get it. There’s been hope that you would have a clear legal showdown. If you speak to those more on the AV1 media side, they are more confident now that if anyone brought AV1 patents against them, those patents would be found unenforceable.”
Davies continued. “We are at a stalemate. Sisvel exists and is charging and collecting VP9 and AV1 royalties. But the royalty-free claim from AOMedia is frustrating because it's not true. I do not think we are getting the giant indemnification sweep we might have hoped for, because AOMedia has very deep pockets and could do indemnification if it wanted to.”
He did not expect real movement going forward. “It's a stalemate really. I'm not sure there will be much of a change. There are lots of irons in the fire, but I don't think any patent owner wants to bring a suit because there is too much risk.”
Ozer then asked, "If royalties are uncertain on devices, would content licensing be any clearer?" Davies responded, “Content is more difficult because those streaming service–focused pools are charging for four codecs and MPEG-DASH as well. I have had those described as very good patents. I have had other people say they are the dregs that could not make it into the main hardware decoder pools. This is the first year of them trying to sign services up.”
He noted the economic imbalance of the hardware vs the content patents. “The relative cost of being compliant with the content patents is very low for the very big services. I'm not sure the amount they are charging is enough to warrant a huge legal showdown. If you are being asked for single-digit millions and you turn over a couple of billion, it is probably wise to pay the content royalty even if you are affronted by it.”
AI Codecs and Broadcast: Research-Active, Deployment-Distant
The conversation then shifted to AI codecs, with Ozer asking when AI-based codecs might matter for broadcast. Davies outlined early work. “There is a project between Nokia, Ericsson and Fraunhofer for a 6G native codec. That means doing encoding, decoding, and transcoding within the radio access network itself, maybe as a way to prop up the mobile infrastructure business away from the on-device AI side.”
He described model-based approaches. “Some approaches want to push a model to the device, so you send inputs rather than frames. The model uses a smaller bitstream to create the video. Sounds interesting, but you face the same silicon inflexibility problems as hardware decoders.”
The gating factor is variability in neural decoders. “All the AI-based approaches require stability in the ecosystem when it comes to shipping consistent neural processing units, NPUs. If they change every year and you are trying to build a complicated model, that is a head-scratcher. These things will never arrive in an operator set-top box or smart TV. That is extra cost. They do not want that.”
VVC, H.267, and Whether Standards Still Matter
Ozer pressed the future of standards. “It seems almost ridiculous. VVC was finalized in 2020 and has gone nowhere. Qualcomm is one of the biggest VVC patent owners, but it has not implemented it in hardware. The context of the VVC standard seems irrelevant to most publishers. Why should we care about H.267 as a service provider?”
Davies did not hedge. “I do not think you should. There are many standards groups. If you spend time around them, you see the vision. But many professionals will chew your ear off about inefficiency.”
Brazil’s LCEVC/VVC Mandate and Why It Likely Stays Local
As a final forward looking issue, Ozer asked about VVC and LCEVC in Brazil's national TV 3.0 standard, which Davies called an outlier. “Brazil has always been an edge case. I spoke on a panel for the Colombian Ministry of Communication, and I got to pick a lot of brains there. The rest of the region looks at Brazil and says ‘You crack on. We will stick with DVB or ISDB.’”
He saw no ripple effect. “ATSC 3.0 could learn from the pace of Brazil, but VVC plus HEVC hardware mandates mean waiting for decoders to refresh. In Brazil, that is feasible because import tariffs restrict devices. In the US or Europe, you wait for decoders to trickle in. It is good, but it does not move the needle. Brazil is isolated.”
Related Articles
In this exclusive interview, Streaming Media's Jan Ozer speaks with Ville-Veikko Mattila, Head of Multimedia Technologies at Nokia, about Nokia's collaboration with Ericsson and Fraunhofer Institute on a new codec development and its implications for the next decade of video compression.
31 Oct 2025
HEVC rode the 4K/HDR wave to success, but AV1, VVC, and LCEVC lack an equivalent killer app. With CDNs racing to the bottom on pricing and new royalty pools threatening to increase costs, codec adoption is no longer just about technical merit—it's about survival. These are the issues I'll explore in this article.
28 Mar 2025
H.267 should be finalized between July and October 2028. If history holds, this means H.267 won't see meaningful deployment until 2034-2036, long after I hang up my keyboard. Here's a brief description of what the standard is designed to deliver and my free and totally unsolicited advice to the committees and technology developers that will create it.
04 Feb 2025
Advance testing shows the codec producing strong bitrate savings over its competitors, but a tiered licensing model could prevent all users from enjoying the best performance.
14 Nov 2019