AMC Is Free Now — Sort Of: What Amazon's Clean-Room Democratization Actually Changes

Amazon Marketing Cloud never had a list price. Amazon’s own product page has said for years that AMC is “available at no cost to eligible advertisers” — the money was never the gate. The gate was the word eligible: a DSP relationship, a managed-service contract, a registered partner, and — per third-party accounts, never an official Amazon figure — a practical self-service floor around $10K+/month in spend. What happened on September 16, 2025 is that the word eligible quietly expanded to mean “runs any sponsored ad.” That is not a price change. It is an access change, and the distinction matters, because almost everything worth knowing about the new AMC lives in the gap between “free” and “open.”

So: AMC is free now — sort of. Free the way it always was, open the way it never was. Here is what changed, what it lets you answer, and the catches that survive democratization.

What actually changed, dated and sourced

On September 16, 2025, Amazon announced instant, self-service AMC access for sponsored ads advertisers, directly inside the ads console (the news-library post and press coverage followed on September 18). Anyone running Sponsored Products, Sponsored Brands, Sponsored Display, or Sponsored TV now finds AMC under “Measurement & Reporting” — no DSP contract, no partner, no application. Live in 27 countries.

What the free tier includes: no-code analysis templates, a low-code use-case library, the full SQL query editor, audience building, automatic ingestion of your sponsored ads campaign signals, and Ads Agent — an AI query assistant, still beta and region-gated. That last inclusion is not decoration: audience activation in sponsored ads (Sponsored Display targeting, bid boosts on Sponsored Products and Sponsored Brands) had already shipped at unBoxed on October 15, 2024, a year before the access change. The activation path for small advertisers existed first; September 2025 handed them the query engine that feeds it.

Two scope notes. DSP advertisers were promised their own direct self-service access “early next year” — i.e., 2026 — in the September announcement; whether that has shipped is not publicly documented, so verify in your console rather than assuming. And the September change did not make everything inside AMC free, which brings us to the table this essay exists to put on the record.

The free / paid / until-when table

Read the third row twice. The richest signals in AMC — total retail purchases rather than ad-attributed conversions, the Full Shopping Insights table with its exposure_type field — were paid subscriptions until June 2, 2026, when Amazon made them free to query through the end of 2026. That announcement arrived via a LinkedIn post from an Amazon Ads employee, reported by PPC Land; there is no official Amazon pricing page to cite, and no public statement about what happens on January 1, 2027. Label that what it is: a promotional window with an undisclosed exit. Build your 2026 analyses on ARP and FSI by all means — that is what the window is for — but do not build standing workflows that assume the meter stays off.

What AMC answers that the console cannot

The reason this matters at all: the sponsored ads console reports last-click attribution per campaign type, in silos. AMC is Amazon’s data clean room — event-level signals from every Amazon ad product joined in one queryable environment, with only aggregated output allowed out. That unlocks the questions the console structurally cannot ask:

• Path to conversion across campaign types. Did the shopper see Sponsored Brands, then Sponsored Display, then convert through Sponsored Products? The console credits the last click; AMC shows the sequence.
• Combined new-to-brand. NTB across all your ad products, deduplicated, rather than per-product NTB that double-counts.
• Overlap and frequency. Which audiences your campaign types share, and at what combined exposure frequency.
• Custom attribution. Your own model — first-touch, position-based, time-decay — instead of Amazon’s default.
• Long-window customer value. Ad-traffic lookback extended from 13 to 25 months on November 11, 2025 (open beta, US/CA first; UK, DE, IT, FR, AT, TR, JP slated Q1 2026), and up to five years of Amazon store purchase signals for measurement use cases — enough runway for genuine CLV and repeat-purchase analysis in long-cycle and seasonal categories.

Before September 2025, this list was the practical moat of agencies and enterprise in-house teams: path-to-conversion decks and NTB overlap analyses were a service line. Now they are console features. The moat did not disappear — it moved. What was an access advantage is now a literacy question, and literacy is a different kind of scarce.

The catches — what democratized access does not democratize

1. The 100-user floor, silently enforced

Every row AMC returns must aggregate at least 100 unique users. Rows that fail the threshold are not flagged, not errored, not footnoted — they are silently dropped from your results. Your query succeeds, your output is simply missing the rows you could not clear.

This is a k-minimum-style privacy control, and it is the honest answer to “is AMC useful for small sellers.” A $5K/month Sponsored Products advertiser can open AMC today — and will find that any granular cut (path by ASIN by week, NTB by audience segment) fails the floor and comes back blank. The threshold does not discriminate by account tier; it discriminates by traffic volume, which amounts to the same thing. Worth noting: no formal seller-versus-vendor distinction in AMC access or signals is publicly documented — third-party blogs asserting one are asserting past Amazon’s paperwork. The real stratification is volume, enforced 100 users at a time.

2. The dialect, and the wall around the room

AMC SQL is a restricted subset: certain standard commands are disallowed, and event-level export does not exist at any price. Raw data stays in; governed aggregate output moves out. That is the clean-room contract working as designed — but it means “knows SQL” does not equal “productive in AMC.” Your team will fight the dialect, the schema, and the threshold simultaneously. The no-code templates soften the first week, not the first quarter.

3. Ad-attributed by default — incrementality is not included

Base AMC conversion tables are ad-attributed. Total-retail views require Amazon Retail Purchases or Full Shopping Insights — the paid features currently in their free window. And even with FSI’s exposure_type field, comparing exposed versus unexposed shoppers is observational: the exposed group saw your ad partly because Amazon’s systems predicted they would convert. That is selection, not randomization. Genuine incrementality still requires designed experiments with holdouts — the same bar the retail and commerce media measurement standards apply to every network’s self-reported lift. Democratized AMC hands you better observational data; it does not hand you a counterfactual.

4. The remaining gates, briefly

The 25-month lookback is US/CA open beta with Europe and Japan slated for Q1 2026; Ads Agent is beta and region-gated; DSP self-service was a promise for “early 2026” whose delivery is not publicly documented. If your plan depends on any of these, check the console, not the press release.

Where AMC sits in the clean-room landscape

One structural fact frames everything above: AMC is a single-publisher clean room. IAB Tech Lab published its Data Clean Room Guidance v1.0 and the OPJA interoperability spec in July 2024; AMC does not implement OPJA, and nothing in the September 2025 change moves it toward multi-party interoperability. The guidance is still useful — as vocabulary. AMC’s 100-user floor is a k-min-style output control; its no-export posture is the “data stays in” pattern; you can grade AMC against the IAB checklist the way you would grade any clean-room deployment. It grades well on governance. It grades null on interoperability, because it was never trying.

Which is the polite way of stating the hard limit: everything AMC tells you is about Amazon. Path to conversion — on Amazon. New-to-brand — Amazon’s definition, Amazon’s shoppers. Five-year purchase history — Amazon’s store. The moment your question crosses a retailer boundary — is my Walmart Connect iROAS comparable to my Amazon iROAS? is this audience incremental for my brand or just for this network? — AMC has nothing to say, and neither does any other single-network clean room. Cross-retailer comparability is the unsolved problem of retail media, and September 2025 did not touch it.

What to actually do

If you run sponsored ads and have never opened AMC: open it this quarter. The console is under Measurement & Reporting; ingestion of your campaigns is automatic. Start with the no-code templates — combined NTB and path-to-conversion first, because those are the two questions the standard console physically cannot answer. Expect blank rows on granular cuts if your traffic is thin; that is the floor, not a bug.

If you have volume: the window matters. ARP and FSI — including exposure_type — are free to query through December 31, 2026. Run the total-retail and long-window CLV analyses now, while the meter is off, and treat the outputs as a calibration benchmark you keep even if the features return to paid. Post-window pricing is not publicly documented; plan for the pessimistic case.

If you are deciding whether AMC deserves a seat in your data architecture: that is a governance question, not a feature question — instance strategy, signal residency, who queries, what leaves the room. The AMC platform-fit checklist exists for exactly that decision.

If you operate across retailers: AMC just became the best-instrumented node in your network — and it is still one node. Every retail media network is converging on the same shape: a proprietary clean room that verifies its own media brilliantly and says nothing about anyone else’s. Democratized AMC raises the floor of what one network will tell you. It does nothing for the question of what to do with N networks’ self-verified answers — that is an orchestration problem, and it stays yours.

The September change was real. A measurement capability that required an enterprise contract in August 2025 requires a Sponsored Products campaign today, and “our agency handles AMC” is no longer a moat — it is a line item to re-examine. But keep the sentence straight: Amazon did not make AMC free. Amazon made AMC reachable, kept the richest signals on a promotional clock, and left the aggregation floor exactly where it was. Free is the headline. Eligible, literate, and above the threshold — that is the product.