cd /news/artificial-intelligence/five-seconds-used-to-count-as-low-la… · home topics artificial-intelligence article
[ARTICLE · art-124504] src=streamingmedia.com ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

Five seconds used to count as low latency. Now only 16% agree.

Bitmovin's 10th Annual Video Developer Report, based on responses from 486 video professionals in 94 countries, finds that only 16% now consider sub-five-second delivery as low latency, down from 60% five years ago. The report also shows 98% of respondents use at least one AI or ML use case, 46% use AI coding tools daily, and 54% use FAST, level with SVOD.

read4 min views1 publishedSep 9, 2026
Five seconds used to count as low latency. Now only 16% agree.
Image: source

#

                Five seconds used to count as low latency. Now only 16% agree.

Five years ago, 60% of video professionals said anything under five seconds qualified as low latency. In this year's Video Developer Report from Bitmovin, that number is 16%.

Nothing about the physics changed. What changed is what viewers treat as normal, and what teams have to build against as a result.

Movement like that only shows up because the survey has put the same questions to the same audience for ten editions. This year's answers, from 486 video professionals in 94 countries, show four parts of the stack shifting at once: AI became an operating layer rather than a feature, advertising overtook subscriptions as the growth engine, quality of experience data started feeding machines instead of dashboards, and the transport layer video has relied on since the early 2000s picked up a credible challenger.

AI moves from feature to operating layer #

The question of whether video teams should use AI has settled. 98% of respondents report at least one AI or ML use case, and 46% of developers now use AI coding tools every day as part of their core workflow.

What limits further adoption is no longer cost or access. 35% cite accuracy and reliability on domain-specific work as the main constraint, against 10% who point to cost. Once budget stops being the objection, the decision shifts from whether to use AI to how much of the production path a team will let it touch, and that is a platform question more than a developer one.

FAST reaches parity with SVOD as CTV advertising scales #

Subscriber growth has matured, so services have spent the past few years building a second revenue engine around advertising. The survey shows how far that has already gone:

  • 69% use an ad-supported model
  • FAST entered the survey in 2024 at 46% and now sits at 54%, level with SVOD after two cycles
  • 67% use linear ad formats
  • 45% use ads, a format that barely existed in most workflows a few years ago

The technical story behind all of it is CTV. Sessions there run two to three times longer than mobile or web, but most CTV devices ship with a single low-power decoder that was never designed with advertising in mind. Delivering a clean ad experience on that hardware falls to the player software on the device far more than to the ad server.

Root cause time is the more honest health metric #

29% of teams need 12 to 24 hours to identify the source of a streaming issue. 9% can do it inside an hour.

That spread tells you more than uptime or backend SLOs do, because a service can look healthy from the server side and still be failing viewers in ways nobody catches until the complaints arrive. The gap between when a problem starts and when a team finds it closes in one place: the player, the only vantage point where a live session is visible as the viewer is experiencing it. Teams are also moving away from static dashboards toward direct access to that data, which changes what can be automated on top of it.

MOQ is on the roadmap well before it is in production #

28% of teams plan to deploy Media over QUIC within 12 months. 2% run it in production today.

Set that against the latency reset and the case for MOQ gets more interesting rather than less. If sub-five-second delivery is no longer what the industry means by low latency, and low latency is no longer the top innovation priority, then the argument for replacing a transport layer that has worked since the early 2000s has to rest on something else. The report spends a full chapter on what that something else is, and how close it is to holding up in production.

Streaming's underlying problems have not changed much across ten editions. What changes each year is which layer of the stack carries the pressure next.

The full 10th Annual Video Developer Report covers AI adoption and agentic workflows, the economics of ad-supported streaming, real-time QoE and observability, and what MOQ means for video delivery, alongside historical trend data and analysis from Jan Ozer and Dan Rayburn.

**Download it at [https://bit.ly/Streaming_Media_Bitmovin_VDR](https://bit.ly/Streaming_Media_Bitmovin_VDR)**

            Companies and Suppliers Mentioned
── more in #artificial-intelligence 4 stories · sorted by recency
── more on @bitmovin 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/five-seconds-used-to…] indexed:0 read:4min 2026-09-09 ·