Jul 1, 2026
ActionStreamer
Media over QUIC (MoQ): What It Is and Why It Matters
If you've heard someone mention "media over quick" and wondered what on earth that means, they were almost certainly talking about Media over QUIC, usually shortened to MoQ and pronounced "mock." It's one of the most talked-about developments in live video streaming right now, and by mid-2026 it's gone from a niche IETF experiment to something Cloudflare, Google, Meta, Cisco, Akamai, and a dozen streaming vendors are actively building on.
So what is it, why does it exist, and should you actually care?
The trade-off MoQ is trying to break
Live video streaming has lived with an uncomfortable compromise for decades. You could pick two out of three: low latency, massive scale, and architectural simplicity, but rarely all three at once.
RTMP (the old Flash-era protocol) was simple and low-latency, but it doesn't scale to millions of viewers and modern browsers dropped support for it.
HLS and DASH, the workhorses behind most of the video you watch on the internet, scale beautifully through ordinary CDNs, but because they chop video into segments and serve them the way a web page is served, they typically carry several seconds of latency, sometimes much more.
WebRTC gets you sub-second latency and powers tools like Google Meet, but its peer-connection model, NAT traversal, and signaling machinery make it expensive and awkward to scale to broadcast-sized audiences.
Every time someone needs both real-time interactivity and audience scale (think live sports betting, auctions, game streaming with real-time chat, or remote production), they've historically had to stitch together two or three of these systems and eat the operational complexity. Media over QUIC is the IETF's attempt to collapse that entire mess into one protocol.
So what actually is it?
MoQ is a publish/subscribe transport protocol for media, built on top of QUIC, the same modern, encrypted, UDP-based transport that underpins HTTP/3. In browsers it rides on WebTransport, the browser API that exposes QUIC to web apps; outside the browser it can run over raw QUIC directly.
The core idea is refreshingly different from how HLS or WebRTC work. Instead of a viewer repeatedly requesting video segments (HLS) or negotiating a peer connection (WebRTC), a publisher announces named tracks of media, and subscribers simply ask for the tracks they want. Media itself is organized into a small hierarchy:
Tracks: a named stream, like "camera 1 video" or "English audio"
Groups: roughly a chunk of a track, often aligned to something like a keyframe boundary
Subgroups: a subdivision of a group where everything shares one underlying QUIC stream, useful for separating layers of a video codec
Objects: the smallest unit, typically one encoded video or audio frame
In between publishers and subscribers sit relays: lightweight servers that cache and forward these objects without needing to decode or understand the media itself. A relay can take a single upstream subscription from a publisher and fan it out to thousands or millions of downstream subscribers, which is what gives MoQ CDN-like scale without needing a CDN's usual HTTP-segment machinery, and without a WebRTC-style media server having to touch every stream.
QUIC's structure is what makes the low latency possible. Because QUIC streams are independent of each other, a lost packet on one stream (say, an enhancement layer of video) doesn't stall delivery of another (the base layer or the audio track) the way a single dropped TCP packet can stall everything behind it. Publishers can also just discard less important data outright when the network is congested rather than retransmitting it and adding delay, an intentional trade of some reliability for bounded latency.
Where things stand in mid-2026
MoQ isn't finished, but it's no longer just a paper exercise either. The core transport specification, formally called MoQ Transport (MOQT), reached draft revision 17 in March 2026, with editors from Cisco, Google, and Meta steering it. Working Group Last Call (the formal step before a document heads toward becoming an RFC) hasn't happened yet, so the wire format is considered stable enough to build against, even if some details are still being argued over.
The visible industry momentum has been hard to miss. Cloudflare has switched on what it describes as the first MoQ relay network, running across its edge in more than 330 cities, and released its relay implementation as open source. At NAB Show 2026 in Las Vegas, eleven vendors, including Ant Media, AWS, Bitmovin, Broadpeak, CacheFly, Cloudflare, Nomad Media, Norsk, Oracle, Red5, and Synamedia, showed interoperable MoQ demos, and an industry group called the OpenMOQ Software Consortium has formed to push relay and player software forward. Browser support, long the biggest practical gap, has also started closing: Chrome and Firefox already support the WebTransport API MoQ needs, and Safari added it in beta in early 2026.
Notably, MoQ's origins trace back to Twitch, where an engineer named Luke Curley (who has since moved to Discord) built one of the earliest implementations before it became a full IETF working group effort in 2022.
Should you actually use it?
Here's the honest, slightly deflating part: for most products shipping today, probably not yet. The people closest to MoQ are refreshingly candid that it's still early-adopter territory: implementations are pinned to specific draft versions that keep changing, tooling like encoders and analytics is immature, and there's no native adaptive bitrate system the way HLS and DASH have.
It's also worth remembering what QUIC does and doesn't give you. It doesn't move bytes any faster or make a large file transfer quicker; a big video download takes just as long over QUIC as over TCP. What MoQ actually buys you is lower, more predictable latency and simpler fan-out for use cases that genuinely need real-time interactivity at scale: live betting and auctions, real-time chat layered on top of a stream, cloud gaming, or synchronized multi-camera production. If your use case is a straightforward on-demand video library, HLS/DASH over an ordinary CDN is still the pragmatic choice, and if you need universal, battle-tested peer-to-peer conferencing, WebRTC still does that job better.
The bottom line
Media over QUIC isn't a product or a vendor's proprietary trick. It's an open, community-driven IETF standard being shaped by engineers from across the industry, aimed at finally giving the internet one media delivery architecture instead of three incompatible ones. The specification is maturing fast, the infrastructure to run it is already live at real CDN scale, and major platforms are placing serious bets on it. Whether it becomes the universal replacement its boosters imagine, or simply a specialized tool that handles the interactive, low-latency slice of streaming, 2026 is shaping up to be the year MoQ stopped being a curiosity and started being infrastructure.
