Retail Media Ad Spend Is Surging: Where Budgets Are Moving in 2026

US retail media ad spending is forecast to approach $70 billion this year, and that number will be quoted in every deck about the channel between now and Christmas. It is also, on its own, close to useless for deciding where to put your money, because a market can grow enormously while becoming a worse place for you specifically to spend.

Which is roughly what's happening. Read EMARKETER's 2026 forecast carefully and the headline isn't really "surging." Growth is slowing, even as retail media keeps taking share of digital budgets. What's actually happening is concentration.

The number that should change your plan

EMARKETER's key figure isn't the total. It's this: by 2028, Amazon's retail media revenues are projected to exceed $75 billion — more than $65 billion ahead of the next-largest network.

Sit with the shape of that. Not "ahead." Sixty-five billion dollars ahead of the closest competitor. That isn't a growing market with many winners. It's one giant, a small group of viable seconds now emerging with genuine scale, and a long, crowded tail of retail media networks that are drifting further behind every quarter.

The reframe "Retail media is surging" is a fact about the category. "Retail media is concentrating" is a fact about your budget. Only one of them tells you what to do on Monday.

Retail media is not a channel. It's two hundred channels.

Here's the structural point that explains why a $70bn opportunity remains so persistently painful to execute.

Every retailer with a website has launched a network, because selling ads against your own shoppers is spectacularly high-margin revenue for the retailer. That is an excellent reason for them to build one. It is not, by itself, a reason for you to buy.

And each network you add brings its own ad platform, its own product taxonomy, its own creative specs, its own reporting standard, its own attribution rules, and its own account manager who would like a call. So:

The three tiers, and what each is actually for

Stop treating "retail media" as one line in the budget. It's three quite different products that happen to share a name.

That last row deserves candour, because everyone in the industry knows it and nobody writes it down. A meaningful share of long-tail retail media spend isn't bought because the media works. It's bought because the retailer's commercial team made it clear that advertising support and shelf space are related topics. That may well be a rational cost of doing business. It is not a performance channel, and reporting it as one corrupts your entire marketing P&L.

The incrementality trap

Now the uncomfortable question, and the one that should govern the whole budget.

A shopper types your brand name into a retailer's search box. Your sponsored listing appears. They click it and buy. The network reports a conversion, a handsome ROAS, and a job well done.

But they were searching for you. They were going to buy you anyway. You just paid a toll to stand in front of your own customer.

Defensive brand-term spending isn't always irrational — if you vacate, a competitor's ad occupies that space, and losing the shelf costs more than the toll. But it must be understood for what it is: a defensive tax, not growth. Booked as growth, it flatters your numbers, inflates your apparent ROAS, and quietly funds an ever-larger toll every year.

The only honest test is a holdout. Turn it off in a region, or for a subset of products, and see what actually happens to sales. Most brands have never done this, which is precisely why most brands do not know what their retail media is worth. The broader discipline here is the same one we set out in marketing mix modeling: when clean click paths lie to you, model and test rather than believing the dashboard.

The scorekeeper problem

Related, and rarely said out loud: the retail media network sells you the advertising and then grades its own performance.

It sets the attribution window, which tends to be generous. It decides what counts as a view-through. It claims the halo. And it reports the results in its own dashboard, in its own taxonomy, in a format that doesn't reconcile neatly with any of your other channels or with anyone else's.

No one should be surprised that a channel measured entirely by its own vendor reports excellent results. This is a specific, unusually pure instance of the problem we described in why attribution is getting harder, and the fix is the same: get an independent measurement layer, insist on comparable metrics across channels, and treat vendor-reported ROAS as a claim rather than a finding.

Where the money is actually moving

1. Off-site, which is the real story

An underappreciated shift, and the one EMARKETER flags as the arena where non-Amazon networks can genuinely compete. On-site inventory is finite — a retailer's website only has so many slots, and Amazon has vastly more of them than anyone else.

But a retailer's data isn't finite. Knowing who bought nappies last Tuesday is valuable everywhere: on the open web, on connected TV, in social feeds. So the smarter retail media buy increasingly isn't the shelf at all. You're buying purchase data, applied elsewhere. That reframing changes who you should partner with, because now the question isn't "how big is their site?" but "how good and how usable is their data?"

2. In-store, slowly

Screens, retail displays, connected shopper journeys. Widely discussed, still underdeveloped, and precisely because it remains underdeveloped it reinforces the advantage of already-scaled players who can afford to build it. Watch it; don't bet the budget on it yet.

3. Away from the long tail

The clearest instruction in the data. If growth is concentrating among the biggest networks, the eighth-largest grocery chain's ad platform is not where your incremental dollar belongs, however warm the relationship.

How to actually allocate

A workable sequence, in order:

1. Separate the tax from the investment. Defensive brand-term spend and trade-driven spend go in one bucket, labelled honestly. Growth spend goes in another. Never mix them in the same ROAS.
2. Cap the number of networks. Pick the ones where your buyers genuinely shop in volume. Two run well beats seven run badly, every time.
3. Run a holdout before you scale anything. If a network refuses to support a clean incrementality test, that refusal is your answer.
4. Measure it yourself. One independent view across every channel, in your metrics, not theirs.
5. Fix the destination first. Retail media sends people to a product page, and a click you paid for that lands on a weak page is money set on fire, which is the whole argument of our product page optimisation guide.
6. Benchmark against your alternatives. The right comparison isn't "good ROAS," it's "better than the next-best use of this money," a discipline we set out in structuring a profitable Google Ads campaign.

For the broader case on why the channel exists at all and why brands piled in, our earlier piece on why retail media networks are growing so fast covers the ground. This one is about what to do now that everyone has arrived. If you'd rather someone ran the tests properly, that's what performance marketing support is for.

The bottom line

Retail media in 2026 is not a rising tide lifting all boats. It's a market consolidating hard around one dominant player and a handful of scaled challengers, while a long tail of networks quietly stops mattering. So the useful question was never "should we spend on retail media" — it's "which tier, for what job, and how would we know if it worked?" Split the defensive tax from the genuine investment and stop reporting them together. Run a holdout, because a channel that grades its own homework will always pass. Concentrate on fewer networks and run them properly. And follow the data, not the shelf, because the most valuable thing a retailer owns is what it knows about its shoppers, and that's the part it can sell you everywhere else.