For roughly 15 years, live video delivery has improved in increments rather than leaps. Codecs got better and HTTP caches got faster, but the basic model of a player pulling files over HTTP has not changed. Media over QUIC (MoQ) is the first serious attempt to change that model at CDN scale, and the industry is now moving on it.
The signal from IBC this year was hard to miss. At IBC 2025 in Amsterdam, I joined a panel on early signs that the video and CDN communities were converging on MoQ. Twelve months later, at IBC 2026, I counted at least 30 stands running MoQ demos. We had one too: you can watch our MoQ demo with Thijs Feryn here.
The standard is still an IETF draft, but vendors are already shipping proprietary implementations, and customer tests are under way around the world. So what value does MoQ bring, and how does it differ from the way streaming is delivered today?
MoQ targets a trade-off that today's live streaming protocols force you to negotiate: you can have scale or sub-second latency, but rarely both at a sensible cost. It does this by changing the model. Instead of a player requesting the next segment over TCP, MoQ runs on QUIC, which sits on UDP and builds in encryption and multiplexed streams, so media isn't held up by packet loss or slow connection setup.
Instead of request/response, it uses publish/subscribe: a player subscribes to a track, such as a 1080p rendition or an audio language, and media is pushed the moment it's produced, with no polling for the next segment. Relays fan that single stream out to many subscribers, which is, in effect, a CDN written into the protocol.
QUIC gives MoQ three properties that are hard to get from TCP:
Faster connection setup. TCP plus TLS 1.3 needs two round trips before the first byte of media. QUIC combines the transport and crypto handshakes into one, reducing round trip time.
No head-of-line blocking. With TCP, a lost packet stalls everything behind it. QUIC streams are independent, so a loss on one stream doesn't hold up another.
HLS and DASH players fetch media in segments of two to six seconds and usually buffer three before playing, so most of the delay comes from waiting, not from the network. MoQ breaks media into much smaller pieces, down to individual frames, and sends each one as soon as the encoder produces it. Latency is then set by encoding, network transit, and the player's jitter buffer, not by segment length. Channel changes get faster too: a new viewer joins at the start of the latest keyframe-led group and can start decoding immediately, so switching streams feels like zapping on an old TV remote rather than waiting for a player to rebuffer.
Today, a stadium goal can reach your phone as a push notification well before it appears in your HLS stream. Closing that gap is the most obvious benefit of MoQ, but a bigger opportunity is interaction. When the video is effectively live, viewers can act on what they see:
In-play betting, where every second of delay is a pricing and integrity risk.
Live auctions and live commerce, where bids and purchases must track what is on screen.
Audience participation, where chat and video need to stay in sync.
Remote production, where operators need near-real-time monitoring over public networks.
There is an economic angle too. If MoQ delivers at HLS-like economics, it lowers the barrier for small, niche, or local broadcasters, plus use cases nobody has designed yet.
MoQ makes distance matter less to the latency budget, but it doesn't make the edge irrelevant. Once a subscription is established, each object travels one way instead of waiting a round trip per segment request, so even London to Sydney leaves plenty of room in a one-second budget. Still, a single Origin / Publisher can't serve a million viewers, and long paths cross more networks, peering points, and congestion. The CDN evolves rather than disappears:
From caching files to relaying tracks. A MoQ relay holds one upstream subscription and fans it out to thousands of viewers, much like a cache collapses many requests into one origin fetch.
Consolidation. Because distance hurts less, operators can serve wider regions from fewer, larger relay sites, provided peering and capacity are in place.
MoQ is built for live first, and HTTP caching remains the best tool for everything that isn't live. A production setup needs both.
The spec isn't strictly live-only. A FETCH mechanism for retrieving past objects is included, and relays can cache, so pause, rewind, and catch-up could in principle be served over MoQ. In practice, today's implementations focus on the live edge, while HTTP delivery of segmented content is mature, cheap, and supported by every device. The near-term architecture is hybrid: MoQ for the live edge, where latency and interactivity matter most, and HLS or DASH from an HTTP cache for VOD.
The operators who win will be the ones who make the handoff between the two invisible to the viewer.
At Varnish, MoQ is a long-term commitment. We are active across the MoQ community and building our own MoQ stack, tracking the current specification closely. We're building it ourselves for two reasons:
Full control and deep integration. Owning the code lets us integrate MoQ natively with the Varnish video CDN software suite, so live MoQ and HTTP delivery run side by side on one platform.
Wire-speed performance at today's economics. We're optimizing the relay to run at wire speed, so MoQ workloads can be offered at the same price point as HLS and DASH, whether as software or as a private CDN as a service.
We've made strong progress, and there's still work to do before it meets our bar. MoQ won't replace everything overnight, but it will change what a CDN is for. We'll share more as we go.
Running live video at scale? See how Ora Streaming delivers it on dedicated infrastructure with predictable pricing, or talk to an expert about where MoQ fits your roadmap.