Warning: Something in AdTech Is Being Switched Off… Most of You Haven’t Noticed
A critical piece of video ad infrastructure dies on April 30. Most of the industry won’t notice until May. By then it’ll just look like a bad week.
What’s Actually Happening (Plain English First)
Before we dive into the AdTech weeds, this post is a bit more technical than usual but it is IMPORTANT… so here’s what’s actually happening.
There’s a system that helps video ads appear on websites.
Think of it like a magic box that sits between the moment an ad is bought and the moment it shows up on screen.
When someone wins an auction for a video ad slot, the ad gets temporarily stored in that magic box. The video player then goes and fetches it.
That magic box is being shut down on April 30.
Most companies don’t realise they’re using it.
When it closes, the auction still runs.
The bid still wins.
Every system still says everything is fine.
But when the player goes to fetch the ad, the magic box is empty.
Nothing shows.
No ad, no impression, no revenue.
And because nothing throws an obvious error, it just looks like a dip in numbers.
That’s it… but it’s a much bigger deal than it sounds.
It Always Starts With a Slack Message
“Anyone else seeing video fill down this morning?”
No panic. No urgency. Just a vague feeling something’s off.
Someone checks GAM. Someone else pulls SSP reporting. A few screenshots get shared. Theories start flying. Seasonality. Buyer pullback. Maybe a deal dropped.
Nobody says “infrastructure failure” like they’re responding to “what coffee do you want?”
Because nothing looks broken. That’s the risk.
The Endpoint Nobody Talks About
April 30th 2026 looks like a completely normal date.
No big industry event. No loudly announced deadline. Just a quiet switch off of something most people have never heard of.
prebid.adnxs.com/pbcIf you recognise that string of characters, you’re already slightly concerned. If you don’t, there’s a decent chance it’s sitting in your setup right now, doing a job you didn’t know it was doing.
An endpoint is essentially a specific web address that one system sends requests to in order to talk to another system… in this case, it’s the address your video player pings to go and fetch the ad after the auction is won.
Over the last few years, this piece of infrastructure became quietly standard across the industry.
It came out of AppNexus, got absorbed into Xandr, and eventually ended up inside Microsoft’s stack. It was fast, reliable, and it was free.
So naturally, the entire industry started pointing at it.
Worryingly, more than 60% of Prebid Cache endpoints are still routing through it.
Not a niche tool.
Not an edge case.
It has been the default path for video ads in header bidding.
Now it’s being turned OFF.
Why This Failure Hits Different
Infrastructure gets deprecated all the time. That part isn’t unusual. What is unusual is how this particular failure behaves.
When this stops working, nothing obvious breaks.
Auctions still run.
Bids still come back.
Winners still get picked.
Every dashboard says everything is fine.
Then the player tries to fetch the ad from cache. That cache no longer exists. So it gets nothing back. And so quietly moves on.
No creative. No impression. No revenue. Just an empty slot where money used to be.
Everything worked. Except the part where you actually get paid.
If you’re only monitoring auction metrics, you won’t see it.
If you’re checking revenue a day later, you might, but by then you’re guessing.
Maybe seasonality.
Maybe demand softened.
Maybe it’s just a bad week.
It’s almost never “that endpoint we forgot existed got turned off three weeks ago.”
This is where it breaks. You have been warned.
This Isn’t Really a Publisher Problem
Most publishers didn’t choose this setup deliberately.
It was added years ago through a wrapper, an SSP integration, or a vendor template.
It worked, so nobody touched it.
The person who originally set it up was probably let go when COVID happened and now lives off grid in Norway dealing with their PTSD.
They are not coming back to fix this.
Whatsmore… there’s no invoice for it. No contract renewal. Nothing to remind you or anyone it exists.
Until it doesn’t.
SSPs can actually see this across all their supply, at scale.
They can scan configurations, identify which publishers are still pointing at this endpoint, and understand exactly who’s exposed.
Publishers can’t easily do that across everything they run simultaneously.
So when revenue drops and buyers start asking questions, guess who the call goes to first…
The Fix Isn’t the Hard Part
There are three options. None perfect, all workable.
Local caching stores the ad directly in the browser and removes the external dependency entirely. Clean solution, but it doesn’t fully work with Google AdX demand yet, which is a real tradeoff for a lot of publishers.
Self hosted cache means running your own version of the service. Full control, proper independence. But you need engineering resource, which immediately rules out a significant chunk of the market.
Managed vendor solutions from companies like Shinka, Aniview and JWX offer hosted options. Some free at scale, some paid, all requiring setup, testing and validation before you go anywhere near live traffic.
And this is where people underestimate the actual problem. Not the decision. The execution.
Even if you choose the right solution today, you still need to:
Update your Prebid configuration
Test across desktop, mobile and CTV
Validate that ads actually render end to end
Update GAM creatives
Wait for trafficking cycles to catch up
That takes time. Time a lot of people in AdTech no longer have.
If you’re starting this on April 25th, you are already in a whole load of shit.
AdTech rarely fails because there’s no solution. It fails because no one finished the job in time.
The Second Wave Nobody’s Talking About
The first impact is obvious. Publishers lose video revenue.
The second impact is messier.
Buyers see campaigns underdeliver.
Frequency data goes strange.
Viewability shifts.
Performance looks slightly off across multiple sites at once.
Nothing dramatic enough to trigger an investigation immediately.
Just enough noise to cause confusion.
Because it’s spread across dozens of publishers, each slightly broken in the same quiet way, it’s almost impossible to trace in real time.
Weeks later Jeanette in finance joins the dots. By then it’s already been filed away as “market conditions”.
This is how money disappears in this industry.
Not in big public crashes.
In small, silent failures that look like normal variance until someone does the maths properly in a quarterly review and realises they’re about to lose their job.
Oh… if you actually want to understand how this machine works under the hood, I literally wrote the book on it — WTF IS PROGRAMMATIC — because this exact situation is what happens when something important becomes invisible.
Send This to Your Team ASAP This Week (you’re welcome)
Hi all, I came across something that I don’t fully understand but it seemed important enough to flag in case it’s relevant to us.
Apparently there’s a “Prebid Cache endpoint” (prebid.adnxs.com/pbc) that’s being switched off on April 30.
From what I can gather, a lot of publishers are still using it without realising, and when it goes it could affect video ads loading, but in a way that doesn’t show up as an obvious error. Revenue just quietly drops.
I might be completely off base on how this affects our setup, but wanted to flag it just in case it’s something we need to look at before the end of the month.
There’s a full post about it here:
https://www.adtechdiary.com/p/warning-something-in-adtech-is-being-switched-off
April 30 isn’t when things explode... It’s when they start quietly not working
And a decent chunk of the industry will only find out in May. When the revenue report looks a bit… off.
This took me ages to research and write. All I ask is that you hit subscribe if you found value.
k, thanks, bye