AV1 — Now Powering 30% of Netflix Streaming
Netflix reached 30% of all streaming on AV1 through a decade of codec co-development, hardware certification, and ecosystem coordination — and the real lesson is that superior compression alone doesn't ship; device ecosystems do.
A codec reaching 30% of global streaming share is not a technical milestone — it’s a supply chain milestone. AV1’s compression numbers were settled years before most people could actually watch it.
Source
AV1 — Now Powering 30% of Netflix Streaming
By: Netflix Published: Dec 01 2025 Type: Engineering article
Saved because it traces the full arc from codec design to production scale, and makes visible the parts of video infrastructure that normally stay hidden: certification pipelines, hardware decoder timelines, and the coordinated effort required to turn a better algorithm into a better product.
From compression efficiency to ecosystem deployment
Netflix has been using H.264/AVC as its primary streaming format since 2007. By 2015, the limits of that bet were clear enough that Netflix co-founded the Alliance for Open Media specifically to develop what would become AV1. That is worth sitting with: a streaming company committed substantial engineering resources to a codec that would not reach meaningful production share for another decade. The reasoning is not hard to follow — a more efficient codec benefits every downstream system, from CDN egress to mobile data bills to buffer interruption rates — but the timeline is a reminder that infrastructure investments of this kind do not have quarterly payback periods.
AV1’s compression numbers are genuinely large. Netflix streaming sessions on AV1 average VMAF scores 4.3 points higher than AVC and 0.9 points higher than HEVC, while using one-third less bandwidth than both. That bandwidth reduction translates directly to a 45% reduction in buffering interruptions. On a platform with over 300 million members, that last number is not just a quality metric — it is network load, CDN cost, and ISP partnership friction, all at once. Better compression is not a feature; at that scale it is an operational necessity.
The path from “AV1 is better” to “AV1 works on the device in your living room” is where most of the actual engineering lived. Netflix launched AV1 on Android in 2020 using the dav1d software decoder, which was already optimised for ARM chipsets and small enough to ship on mobile without meaningful overhead. Smart TVs depend on hardware decoders, so expanding to large screens required working directly with device manufacturers and SoC vendors through a formal certification process to verify conformance and performance. Netflix added AV1 to its certification requirements in 2019, and over the five years from 2021 to 2025, 88% of large-screen devices submitted for certification supported AV1 — with the vast majority capable of 4K at 60fps. Since 2023, nearly every device submitted is AV1-capable. The ecosystem moved faster than the historical baseline for codec adoption would have predicted, and the coordinated AOMedia membership structure is a plausible reason why.
Two features in the AV1 spec that go beyond raw compression efficiency deserve attention. The first is HDR10+ streaming, launched by Netflix in March 2025, which pairs AV1 with dynamic metadata that adapts tone mapping per scene and per device. The combination delivers detail and colour accuracy that static HDR formats cannot match, and 85% of Netflix’s HDR catalogue had AV1-HDR10+ coverage by the time this post was published. The second is Film Grain Synthesis. Film grain is expensive to encode faithfully — it is random noise, and entropy coders cannot compress random data. AV1 solves this by stripping the grain before encoding and transmitting only the parameters needed to resynthesize it at the decoder. The result is striking: an AV1 FGS stream at 2804 kbps can deliver cinematic film grain quality that a regular AV1 stream at 8274 kbps cannot match. A 66% bitrate reduction with higher perceptual quality is not a minor improvement; it shifts what is deliverable over a typical household connection.
The work is not complete. Live streaming and cloud gaming introduce constraints that VOD does not have. Live content arrives at tens of millions of viewers simultaneously, so AV1’s bandwidth efficiency is attractive, but the encoding side runs under real-time constraints that allow less optimisation than offline VOD encodes. Cloud gaming is harder still: every rendered frame must reach the player with low enough latency that button inputs feel responsive, which makes frame size and decode overhead more important than compression ratio alone. AV1’s layered coding — base content layer plus a swappable enhancement layer for graphics overlays — is relevant for live sports, where on-screen graphics change per region and per sponsor deal. These are active evaluations, not shipped products, and the article is appropriately precise about that distinction.
The easy misread of this story is to treat it as evidence that open codecs win because they are technically superior. They are technically superior, and they win, but the mechanism is not automatic. dav1d is a sponsored open-source library. Hardware support required direct partnership engineering. Certification is a formal process that Netflix maintains and enforces. AV1 at 30% is the result of sustained, coordinated industry effort over a decade, not a natural consequence of a better specification.
My take
The device certification pipeline is the most underappreciated part of this story. Netflix’s ability to enforce AV1 conformance as a precondition for certification is a structural advantage that accelerated hardware adoption in a way that codec quality alone could not. The lesson for systems design is that the deployment environment is part of the architecture — and controlling certification is a way to shape that environment.
Where this is useful
- Understanding why codec adoption timelines are measured in years, not months
- Evaluating the real cost of streaming infrastructure: egress, CDN, ISP load, not just bitrate
- Film Grain Synthesis as a case study in separating signal from noise at the codec level
- Reasoning about trade-offs between VOD, live, and interactive (gaming) streaming pipelines
- Not applicable to real-time communication where latency constraints dominate all other metrics
Apply this: When evaluating a new encoding format, measure ecosystem readiness — hardware decoder availability and certification coverage — as a first-class constraint alongside compression efficiency.
What I want to remember
Remember: A better codec ships when the ecosystem is ready, not when the specification is final — device certification, hardware decoder availability, and coordinated industry investment are what close the gap between algorithm and adoption.