Meta rolls out Media over QUIC for Instagram Live ingest, replacing its old stack

Meta has started rolling out Media over QUIC (MoQ) in Instagram Live. It covers the first mile: the creator's phone sends a video stream to the company's servers. MoQ is an open protocol for real-time video that is being developed at the IETF, and it does not yet have a final RFC. A platform with an audience of more than 2 billion people a month is betting real production traffic on it. In its announcement, Meta talks about a simpler protocol stack and a better user experience. The published text contains no figures for latency, packet loss, or the share of traffic that has been migrated.

What exactly changed

Classic live streaming has several layers, and each plays by its own rules. Ingest has been built on RTMP for decades: a protocol from the Flash era that runs over TCP and is familiar to every encoder. Video reaches viewers by a completely different route, most often through HLS or DASH with segments and buffers, and sometimes through WebRTC for low latency. Converters sit between these two worlds. Each one adds latency, a point of failure, and more lines in engineers' backlogs.

MoQ tries to fold all of this into a single design. Media is divided into tracks, groups, and objects. A publisher pushes them into the network, subscribers pull them out, and intermediate relays forward the data without inspecting its contents. The transport is QUIC (RFC 9000) or WebTransport on top of it. According to Meta's description, the stack becomes simpler on this segment, and the company has already gained hands-on operating experience. As far as is known, Meta also has its own open implementation of MoQ, the Moxygen library.

Why QUIC

For mobile broadcasting, QUIC has concrete advantages over TCP:

• No head-of-line blocking. In TCP, losing a single packet stalls the entire stream. In QUIC, independent streams do not interfere with each other.
• Connection migration. A creator walks out of Wi‑Fi range and switches to LTE. The session can continue without a full reconnect.
• Fast connection setup. The handshake is combined with TLS 1.3, so stream start and recovery after a dropout are faster.
• Control over data staleness. When the network is congested, MoQ can drop old groups of frames instead of accumulating delay. For live video, that matters more than delivering every byte.

What is known and what is not

• Protocol: Media over QUIC, IETF drafts, no final standardization.
• Deployment segment: ingest in Instagram Live.
• Platform scale: more than 2 billion users a month.
• Claimed effect: a simpler stack and a better user experience.
• Not disclosed: latency before and after, the share of traffic on MoQ, the impact on stream stability, and infrastructure cost.

The main question here is organizational, not technical. Running a protocol without a final standard in production at this scale means Meta is prepared to maintain a fork of the specification and change its implementation as the drafts evolve. For small teams, that path is usually too expensive. For Meta, it is an investment in unifying its video stack and not depending on the RTMP legacy.

What streaming teams should do

It is too early to rewrite production systems for MoQ. A sensible plan: follow the IETF drafts, deploy a test relay, and run your own traffic through it. Start with open implementations, including Moxygen, and then compare latency and behavior on poor networks against your current RTMP and WebRTC setups. Be sure to check support in encoders and mobile SDKs: its absence is what most often blocks the move. If you have no requirement for sub-second latency, low-latency HLS remains the more predictable choice for now. Real numbers from Meta, once it publishes them, will be the best guide for making the decision.