Oracle and Cloudflare use Media over QUIC for real-time AI metadata
Oracle, Cloudflare, and Ateme are demonstrating the use of Media over QUIC (MoQ) to treat AI metadata as a first-class subscriber and publisher within a live video pipeline. This architecture aims to reduce the need for custom transport layers and protocol translation by integrating AI analysis directly into the streaming session.
Key Takeaways
- AI models act as 'machine participants' that subscribe to video tracks and publish event metadata into the same MoQ session.
- Oracle reports time to first frame is roughly eleven times faster than standard HLS at the median when using MoQ.
- The demonstration integrated C2PA provenance signing and verification directly into the live flow via partnerships with Qualabs and EZDRM.
- Cloudflare has deployed MoQ relay infrastructure across more than 330 cities, enabling global-scale pub/sub for live media tracks.
- IETF MoQ Transport spec is at draft 19 as of July 2026, with a target publication request date of December 2026.
Why It Matters
This shift from AI as an external 'observer' to a native 'publisher' solves the long-standing problem of metadata being stripped during protocol translation between contribution and distribution. By using MoQ as a single fabric from camera to edge, operators can deliver AI-generated replays, captions, and provenance data without building bespoke synchronization bridges for every new signal. For the broader ecosystem, this collapses the trade-off between CDN-scale economics and sub-second latency, providing a standardized way for automated agents to interact with live streams. Watch for early adoption in contribution and mezzanine feeds, where single-operator control allows for immediate efficiency gains before the protocol reaches fragmented consumer devices.
Additional Context
The push for Media over QUIC (MoQ) coincides with a broader infrastructure pivot toward what Cloudflare calls the "Agentic Internet." In July 2026, Cloudflare introduced new tools to manage how AI agents interact with web content, including a provisioning API that adds isolation and access controls to MoQ relays. Per Cloudflare, every server across its 330-city network now functions as an MoQ relay, providing the global footprint necessary for 200–300ms glass-to-glass latency at CDN scale. This infrastructure is increasingly being paired with AI-specific routing layers; for example, swXtch.io is scheduled to debut its AI Router at IBC 2026, which is designed to route live feeds to multiple inference models in parallel while maintaining broadcast-grade timing.
Simultaneously, Oracle is positioning its Cloud Infrastructure (OCI) as the backbone for these high-demand AI workloads. In June 2026, Oracle reported that its cloud infrastructure revenue grew 93% year-over-year, driven largely by AI adoption. The company has integrated over 1,000 AI agents across its application suites, emphasizing that these models must now "reason, decide, and execute" within operational workflows rather than sitting on the periphery. This enterprise demand for integrated intelligence matches the technical momentum of MoQ, as major browsers like Safari shipped WebTransport support in March 2026, effectively removing the final client-side hurdle for MoQ-based distribution to end users.
Read full article at enginesofchange.ai
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source