Amazon DSP Minimum Spend in 2026 What It Actually Costs to Run

Search "Amazon DSP minimum spend" and the results will tell you, on the same page, that the minimum is $50,000, that it is $35,000, that it is $10,000, and that there is no minimum at all. Several of those pages are dated 2026 and confidently contradict each other. Most of them are describing different things and presenting them as if they were the same thing.

This page sorts out which number applies to your situation, shows where Amazon's own documentation disagrees with itself and why, and then answers the question the title actually promises: not what it costs to enter DSP, but what it costs to run it. Because "minimum spend" is media only, and media is a fraction of the number that leaves your business each month.

Three lines, because that is genuinely how simple the top layer is once you stop conflating the categories:

Managed service carries a published minimum of roughly $50,000 USD per month, worded "typically" and varying by country. Amazon's current public documentation does not state a universal minimum for general self-service DSP access, though format-specific guidance still applies in places (self-service Streaming TV packages, for example, carry a $10,000 recommended campaign minimum). And the number that matters more than either is one Amazon does not publish at all: a practical operating budget of about $10,000 to $15,000 per month for 90 days before the data starts to mean anything, which is CMO's agency-practice recommendation rather than an Amazon rule.

Everything below expands those three lines. If you only read the first section, read the one that follows, because the sourcing is where most of the confusion in this category lives and dies.

The reason the internet cannot agree on a number is that Amazon's own pages state different figures in different places, each scoped to a different context. Here is what the source pages say, quoted precisely, with each row attributed to the Amazon page it appears on.

Three interpretation points fall out of that table, and they resolve most of the disagreement you will find elsewhere.

First, the only universal published number is the $50,000 managed-service figure. It appears on more than one Amazon page. But it is hedged with "typically," and it carries an explicit country-variance caveat. That is an indication of what Amazon expects, not a contract you sign. Amazon's FAQ directs readers to contact an account executive for specifics rather than treating the figure as fixed.

Second, and this is the single most misquoted fact in the entire category: the $10,000 self-service figure is scoped to Streaming TV packages, not to DSP in general. Anyone quoting "$10K to start with Amazon DSP" is stretching a Streaming-TV-specific recommended minimum past what the source actually supports. The device-ads guide makes the pattern clear: within a single DSP format, managed service carries the $50,000 minimum and self-service carries none. The number that applies depends on both the access model and the format.

Third, the words "typically" and "may vary per country" are doing real work in that sentence. Practitioners in non-US marketplaces report different thresholds, and Amazon's phrasing is deliberately non-committal. Read the hedging as intentional.

Amazon source pages referenced above: advertising.amazon.com/resources/faq, display-ads, streaming-tv-ads, device-ads guide.

The confusion is not random. It comes from six specific sources of variance, and once you can name them, contradictory articles stop being contradictory. They are simply answering different questions.

The access model is the biggest one: managed service, self-service direct, and self-service through an agency seat each carry a different cost structure. The format matters too, because display, Streaming TV, audio, and device ads all have separate published guidance, and the Streaming TV number in particular gets misapplied to the whole platform. Package type introduces its own confusion, since campaign minimums, monthly commitments, and quarterly engagements are different commitments described with the same vocabulary. Country shifts the managed-service floor, which is exactly why Amazon hedges. Timing compounds everything, because older advice asserting a self-service minimum still ranks alongside newer pages that no longer state one.

The sixth source resolves more confusion than the other five combined, and almost nobody states it plainly: agency minimums masquerade as Amazon minimums. The $10,000 and $35,000 figures circulating in older articles are frequently a partner's own threshold, not a number Amazon imposes. An agency that will not take a DSP client under $35,000 in spend is describing its business model, not Amazon's rules. When you see a number that is not $50,000 and not zero, your first question should be whose minimum it actually is.

One more claim deserves direct handling, because it is spreading and it is not verifiable against Amazon's public pages. A widely shared assertion holds that Amazon dropped its managed-service minimum from $50,000 to $10,000 effective January 2026, sourced to closed-door agency briefings. As of this writing, Amazon's public documentation still shows the $50,000 managed-service figure, and the January 2026 drop is unconfirmed against any first-party page. Treat it as unverified and confirm with an Amazon Ads representative before planning around it. Repeating it as fact would undermine the entire point of a page built on sourcing rigor.

Access model is the variable that governs almost everything about cost, so this is the comparison worth internalizing before any budget conversation.

Managed service buys you an Amazon Ads account team that operates the campaigns and provides consultative support, and Amazon lists a managed services fee, added to media cost, for exactly that. The trade-off is the commitment level. What managed service primarily changes is who operates the campaigns and the level of Amazon support; it should not be treated as a separate premium tier that automatically unlocks "better" inventory. Actual supply and deal availability can still depend on marketplace, format, and account arrangement. The common and expensive misconception is that the $50,000 buys access to placements self-service cannot reach. It buys the account team.

Self-service direct is the lowest-fee route on paper, since Amazon charges no management fee and you pay media plus a percentage-based technology fee. It assumes you have in-house programmatic capability, which most brands under $25 million on Amazon do not, and the platform's complexity punishes teams that treat it like an extension of Sponsored Products.

Self-service through an agency is where most mid-market brands land. The agency holds the seat, brings the expertise the platform assumes, and charges a fee on top of media, commonly in the 10 to 20 percent range or a flat retainer at lower spend levels. There is no Amazon-imposed floor on this route, which is what makes DSP reachable for brands that the $50,000 managed number would have ruled out.

Here is the section that fulfills the title. Every published minimum you will find refers to media spend and only media spend. It is the one line item in a stack of six, and brands that budget only for the media number get blindsided by the assembled total.

Take a worked example at $15,000 per month in media.

The honest conclusion is uncomfortable but necessary: a brand planning around a $10,000 media budget should expect meaningfully more than $10,000 to leave the business each month once the technology fee, management fee, creative, and measurement are assembled.

On the technology fee specifically, Amazon does not publish one universal percentage. Amazon's pricing transparency guide describes the DSP technology fee as set at the entity level, calculated as a percent of media cost, and disclosed at the time of campaign setup. So use the rate attached to your own seat rather than a generic percentage from a third-party article. A brand or agency may legitimately know its contracted rate; what nobody should do is assume a number they read online applies to their account.

CPM context matters here too, because it determines what your media dollars actually buy, but it also resists a single benchmark. Amazon does not publish a universal DSP CPM figure, because pricing varies by format, placement, audience, and auction conditions, with Streaming TV and connected-TV placements generally sitting at the higher end and consuming budget quickly. Before approving a $15,000 test, model expected reach using the CPM assumptions from your actual media plan rather than a generic marketplace average. At a given budget, CPM is what separates "enough impressions to optimize" from "a rounding error in your category," and that math should happen before the spend, not after.

Amazon does not publish revenue or Sponsored Ads spend thresholds for DSP readiness, and it does not publish a budget required for DSP to perform effectively. The framework below is CMO operating guidance for deciding whether we would recommend a DSP test to a mid-market brand. It is agency practice, not an Amazon requirement, and we are labeling it as such because a page that criticizes unsourced DSP numbers cannot quietly introduce its own.

On Budget

In CMO's operating framework, budgets below roughly $10,000 per month are often too constrained for the audience breadth, frequency, and testing depth we want from a meaningful DSP program. That does not mean Amazon DSP cannot technically run below this level; it means we would usually test lower-funnel Amazon Ads options first. The honest minimum viable test we plan around is roughly $15,000 per month in media sustained for about 90 days. That is the range where, in our experience, the bidding model has enough impression density to work with, and 90 days is roughly how long it takes for genuine signal to separate from week-to-week variance. Shorter or smaller than that, and you are usually reading noise.

On Allocation

The most useful way to size DSP is as a share of your total Amazon ad budget, not as a standalone number. If you are spending $100,000 a month on Amazon advertising overall, then $10,000 to $15,000 in DSP is a defensible 10 to 15 percent allocation, and the awareness it builds compounds the return on the rest of your spend. If you are spending $20,000 a month total, DSP is not your next dollar. Sponsored Brands and Sponsored Display are. Adding DSP to an undersized ad budget does not diversify your program; it dilutes it.

On Readiness

Budget is not the gating question. Maturity is. A brand can have the money and still not be ready, and a readiness failure wastes the spend regardless of how much of it there is. The criteria we run a prospective brand against:

• Sponsored Ads already running profitably and stably (we generally want to see a stable program, often around $8,000 to $10,000 per month, before considering DSP)
• Enough Amazon revenue that a failed 90-day test is survivable (we commonly use $1 million or more in annual Amazon revenue as a reference point)
• Creative assets ready, or budget to produce them
• Brand Registry and account standing in order
• An actual person who will read the reporting
• Amazon Marketing Cloud set up, or a partner who already has it

The explicit verdict most articles refuse to give: if a brand fails two or more of those, the answer is not a smaller DSP budget. It is not yet. A reduced budget on a brand that is not ready fails the same way a full budget would, just more slowly.

For readers still deciding whether the channel belongs in their mix at all, the distinction is funnel position. Sponsored Ads primarily capture and convert demand closer to the shopping moment, through keyword, product, and audience targeting. DSP extends the strategy further up the funnel and across programmatic inventory: it retargets the product-page visitors who browsed and left without converting, targets purchasers of competitor ASINs, and runs awareness across Amazon-owned streaming and device inventory that lower-funnel formats cannot touch.

The measurement difference is real, too. DSP is credited partly through view-through attribution, meaning it earns recognition for conversions that happen after an impression without a click. That is why measuring DSP against last-click ROAS understates it, and why total ad spend as a percentage of total revenue, rather than standalone DSP ROAS, is the honest way to evaluate it. If your only goal is to convert buyers already looking for you, Sponsored Ads do that more cheaply. DSP earns its place when the goal is to grow the pool of buyers who look in the first place.

When DSP is the right layer for your brand, the execution is where the return is made or lost, and that is exactly the strategic engagement CMO runs through its Amazon DSP advertising services: audience thesis first, measurement framework defined before launch, and DSP coordinated with Amazon PPC management rather than bolted on as a standalone channel.