Why AV1 Needs to Win on Mid-Tier Phones

AV1 is royalty-free, technically mature, and offers measurably better compression than the codecs it replaces — but the people who most need it are on the phones least able to run it, and no individual company can fix that alone.

Source

Video Streaming with the AV1 Video Codec in Mobile Devices

By: Meta, Vodafone, Google
Published: September 24, 2025
Type: Engineering article

Saved because it reframes codec adoption as an ecosystem coordination problem rather than a technology problem — and the joint authorship is itself the argument.

The 75 percent problem

Video now accounts for 70 to 80 percent of all mobile data traffic globally. AV1 offers roughly 30 percent better compression than H.264 and VP9 at equivalent visual quality — a number that has been validated independently by Meta, Netflix, Google, and others through separate production deployments. Applying 30 percent better compression to 70-80 percent of mobile traffic would meaningfully reduce network congestion, lower CDN and encoding storage costs, and let the same bandwidth deliver higher-quality video to more users simultaneously.

The constraint is that roughly 75 percent of global handset sales are low- and mid-tier devices. And those devices are where the AV1 bottleneck lives.

AV1’s compression gains come from more sophisticated prediction and entropy coding than H.264 or VP9. More sophisticated coding means more expensive decoding. A premium smartphone has dedicated silicon — a hardware decoder block in the SoC — that can decompress AV1 at the same power consumption as H.264. A low- or mid-tier phone that lacks this block has to decode AV1 in software, running the decoder on the same ARM cores handling the rest of the application. Software AV1 decoding works — it’s functionally correct — but it runs hotter and drains battery faster. Mobile video consumption is a long-session activity. A few percent of extra drain per hour adds up over a two-hour session, and heat throttling under sustained load can turn a watchable stream into a stuttering one.

The irony of this distribution is tight. Low- and mid-tier devices have the weakest network connections, so the 30 percent compression improvement would matter most to their users — better quality video at the same bitrate on a congested network, or the same quality at 30 percent lower bitrate, which matters when a user is on a capped mobile plan. These are also the devices where the hardware to actually use AV1 efficiently is absent. The codec’s benefit is inversely correlated with the hardware’s availability.

There are two paths to closing that gap: hardware and software. Hardware AV1 means a dedicated decoder block in the SoC, which handles decompression with minimal CPU involvement. This is the energy-efficient solution, giving battery and thermal characteristics comparable to H.264 decoding, but it requires a silicon design decision — a SoC manufacturer has to allocate area on the chip for the AV1 block and ship it in a product cycle that typically spans 18 to 24 months. The white paper’s primary recommendation is that SoC vendors evaluate this adoption.

Software AV1 is the bridge. A software decoder ships as a firmware or OS update to devices already in the field. No new silicon needed. The tradeoff is battery efficiency and peak performance under load, but for mid-tier devices in regions with more constrained network bandwidth, a mild battery cost may be worth the quality improvement. The white paper explicitly recommends software decoders where hardware isn’t feasible, acknowledging the transition timeline. Device makers can reach existing installed base without waiting for a chip cycle.

The encoding side is worth noting separately. AV1 encoding is substantially more expensive than H.264 or VP9 encoding — historically, five to ten times the compute per frame, though tooling and hardware encoders continue to improve. That cost is paid by content providers, not by end users or operators. Meta already runs AV1 encoding for Reels. YouTube encodes to AV1 at scale. This is not a theoretical cost — it’s being paid today by the companies signing this white paper. The point of the white paper is precisely that the encoding investment is already made, and the network benefit is already demonstrated in premium device deployments. The missing link is mid-tier SoC adoption.

Codec history is relevant context here. VP9 was technically superior to H.264 and is still not universally hardware-supported on mobile devices, years after its introduction. HEVC (H.265) offered similar compression gains to AV1 but was encumbered by fragmented patent licensing, which slowed adoption and created inconsistent support. AV1’s royalty-free licensing via the Alliance for Open Media removes the licensing obstacle that stalled HEVC. But royalty-free doesn’t automatically mean universal hardware adoption — VP9 also avoided licensing fees and still faced a slow hardware rollout. The licensing structure changes the economics; it doesn’t resolve the SoC investment decision.

What a joint white paper from Meta, Google, and Vodafone does is something a technical specification alone cannot: it signals committed demand. SoC manufacturers invest in codec hardware blocks when they have confidence the codec will matter for their customers’ devices. Three companies representing a significant fraction of global mobile video traffic and mobile network bandwidth saying together that AV1 is their direction is a business signal, not just a technical recommendation. It tells the chip manufacturers that the pipeline is ready and the content is there — the missing piece is the hardware to receive it.

My take

The document is most interesting as a coordination mechanism. All the technical evidence for AV1 has been available for years — the compression benchmarks, the royalty-free licensing, the production deployments at premium tier. The white paper doesn’t add new technical evidence. What it adds is collective commitment by major demand-side players, directed at supply-side decision-makers. This is how open standards get adopted when the bottleneck isn’t the specification but the ecosystem.

The battery/quality tradeoff framing around software vs hardware decode is more nuanced than it appears. Software decode being “viable” covers a wide range. On a mid-range 2023 Snapdragon in a temperate climate, software AV1 is probably fine for short-session use. On a low-tier MediaTek in a hot climate, playing a two-hour video, software AV1 could meaningfully degrade the experience. The recommendation for software as a bridge is right directionally, but the actual user impact depends heavily on device class and usage pattern — which are precisely the segments that vary most in the 75 percent of global sales that aren’t premium.

The 30 percent compression figure should be read as a population average over diverse content, not a per-clip guarantee. Sports and action content with lots of motion may compress less efficiently under AV1’s prediction model than static content. The 30 percent is consistent across large-scale deployments, but individual categories will vary, and that variance matters for encoder profiling decisions.

Where this is useful

• Evaluating codec strategy for a video platform serving a global, device-diverse audience — AV1 makes sense at the infrastructure level today, but the user-visible benefit depends on receiver hardware.
• Understanding open standard adoption dynamics — the pattern of major demand-side players jointly signaling to supply-side manufacturers applies beyond codecs.
• Planning mobile video features where battery efficiency is a first-class constraint; assume software AV1 has meaningful cost on low-tier devices in sustained playback.
• Less relevant for platforms where the delivery target is primarily premium devices in high-bandwidth markets — the AV1 story for that tier is already settled.

Apply this: If your platform serves video to a global user base, check your actual device distribution against AV1 hardware decoder penetration. The gap between your AV1 transcoding investment and the share of users who can use it efficiently is the number worth tracking.

What I want to remember

Remember: AV1’s benefit is largest for the users on the least capable devices — and those are the same users whose hardware is least likely to support it efficiently, which means codec adoption requires coordinated pressure on SoC manufacturers, not just content-side investment.

Related reads

• How Netflix achieved 30 percent streaming savings with AV1
• Meta’s AV1 for real-time communication