The header bidding wrapper war is over. Prebid won.
Five years of fragmentation collapsed into one open-source standard. Here is what changed.
For most of the past five years, every meaningful publisher had to make the same uncomfortable choice about how to run their ad auction. That choice is now gone, and the way it disappeared tells you more about the next five years than any vendor roadmap will.
The state of play
Run Prebid, the open-source wrapper, and you accepted that you were supporting it largely on your own. Run a managed wrapper and you accepted that someone else owned the auction logic on your pages. Most publishers ran both, paid the latency cost, and lived with the operational mess.
Three quiet changes in late 2025 collapsed that trade-off. Prebid did not win a fight so much as the fight stopped being worth having.
How we got here
The wrapper war started in 2018 as an honest disagreement. The open camp argued for transparency and community-maintained adapters. The managed camp argued for performance and a phone number to call at 3am. Both were right at the time.
The interesting story is not that Prebid won. It is that nobody fought back.
Between 2021 and 2024 the gap closed from both directions. Prebid got serious about latency budgets; managed wrappers got pushed into more transparency by publishers burned in the 2022 downturn.
What actually changed in 2025
The 7.4 release was the inflection point. Three things mattered:
- A twin-config setup that lets an old wrapper and a new Prebid build run side by side without double-counting.
- A rewritten floor module that does in Prebid what publishers were paying third parties to do.
- Real governance teeth — three SSP adapters were delisted in Q1 for behaviour the community decided crossed a line.
What publishers should do now
Migrate, but on your own timeline. If you are on a managed wrapper, run Prebid 7.4 in shadow mode for thirty days and compare auction outcomes before you flip anything. If you are on an old Prebid build, upgrade in stages — the floor module alone pays for the work. If you are not running header bidding at all, you are leaving real money on the table and the tooling has never been more forgiving.