The DSP minimum spend is gone.

Amazon removed the self-service minimum spend for its DSP in November 2025, announced at unBoxed. There is no official floor anymore: a self-service seat can technically start at any budget, the practical working entry is $5,000 to $10,000 per month, and Amazon's managed service still expects roughly $35,000 to $50,000 per month (Marketplace Ad Pros). If a page told you the minimum is $50,000, it was written before November 2025, and most of what ranks for this question still says exactly that.

This guide covers what actually changed, what entry realistically costs by route, and why a practical floor survives the official one.

What changed in November 2025?

Amazon dropped the minimum spend requirement for self-service DSP accounts. Before the change, self-service access carried real floors and managed service was the assumed path, which kept the DSP an enterprise and large-agency product. After it, any advertiser who can operate the console (or work with an agency that can) can run programmatic through Amazon at whatever budget makes sense.

The context makes the move legible: Amazon DSP grew from under 10 percent to roughly 20 percent of global programmatic spend in about 15 months (ppc.land), and non-endemic advertisers, brands that do not sell on Amazon, reached 23.8 percent of Amazon ad spend in 2026, up from 9.1 percent in 2023 (Amra and Elma). Amazon is converting the DSP from an enterprise product into a volume product, and removing the floor is how you mint new advertisers.

What does Amazon DSP cost to start in 2026?

The self-service row deserves the caveat baked into it. "No minimum" means Amazon will not stop you; it does not mean $500 per month buys a functioning programmatic program. Media cost is also not the whole bill: you are operating a real console with a real learning curve, or paying someone who already has.

Why is there still a practical floor?

Three reasons the $5,000 to $10,000 working entry survives the official floor's removal:

Delivery systems need data. Programmatic optimization learns from conversions and engagement signals, and tiny budgets generate too few events to learn from. Below a workable spend level, you are buying impressions without buying optimization.

Fixed effort does not scale down. Campaign setup, creative production, audience strategy, and weekly management cost roughly the same at $3,000 per month as at $30,000. At very low spend, the operational overhead per delivered dollar becomes the worst line in the plan.

Flights need room to pace. DSP buying runs on flights with budgets and dates, and a flight too small to pace evenly cannot be diagnosed when it underdelivers. The mechanics of that are covered in the DSP pacing guide.

Is entering at the new lower floor worth it?

For a growing set of advertisers, yes, and the inventory is the reason. A self-service seat buys access to Prime Video's ad-supported audience, 315 million average monthly viewers across 16 countries (ppc.land), plus Netflix inventory through Amazon DSP since late 2025 and Spotify since October 2025. For a DTC brand or a non-endemic advertiser, that is CTV and streaming-audio reach that used to require enterprise commitments, now available at a test-and-learn budget.

Two eyes-open notes for anyone entering at the new floor. First, the reporting is its own discipline: DSP attribution is 14-day only, report data retains for roughly 60 days, and the console takes learning, so read how DSP reporting actually works before your first flight, and note that Amazon retires its legacy reporting tools on December 31, 2026 as everything moves into Campaign Manager. Second, delivery data is not analysis: knowing your flights delivered is not the same as knowing which creatives earned it, which is the layer Peachblue adds for DSP advertisers, from creative-level analysis through flight pacing you can query conversationally.

That 60-day retention has a timing consequence new entrants tend to discover too late: the console discards your performance history on a rolling basis, so whoever starts capturing it on day one keeps it, and whoever connects analytics in month six has permanently lost months one through four. Your first flights are the most expensive learning you will ever buy on the platform; syncing them somewhere that keeps them is how the learning compounds instead of expiring. Connect the analysis layer when you launch, not when you are big enough to feel like you need it.

The floor's removal also changes the math for agencies: DSP clients too small to serve at a $50,000 floor are now viable accounts, and running a book of them is a portfolio problem, per-client pacing, margin-aware reporting, and creative analysis across every client's flights. If that is the move you are evaluating, talk to us about agency plans.

The floor is gone, the inventory got dramatically better, and the operational bar is the honest remaining cost of entry. Budget for the bar, not just the media.