Amazon reported $19.8 billion in advertising services revenue for the second quarter of 2026 on July 30, a 26% increase over the same period a year earlier. Growth had held at 22% for four consecutive quarters before this one, so the acceleration is real. But the revenue line is the least interesting number in the release. What matters for anyone planning media over the next twelve months is the set of performance figures Amazon chose to volunteer alongside it, because those figures describe a company rewriting how premium video gets priced.

Four properties, no inventory left

Amazon introduced more than 30 new advertisers to the NBA in its first year with the league, and inventory across Thursday Night Football, the NBA, the WNBA and NASCAR sold out in the quarter, chief executive Andy Jassy told analysts on the earnings call.

Scarcity in premium video is not a new story, but the composition of that scarcity is. Amazon’s first full NBA season under the eleven-year rights agreement it signed in 2024 peaked at 6.5 million United States viewers for Game 7 of the Eastern Conference Semifinals, and European viewership of the league more than doubled year over year. NASCAR’s second season on the service averaged 2.3 million viewers and delivered the youngest audience the sport has drawn from any broadcaster since 2017. Amazon did not buy those rights to build a television network. It bought them to attach a purchase record to a live audience.

The numbers that change the negotiation

Two disclosures on the call do more work than the headline. Brands that bought across multiple sports achieved 2.3 times higher unduplicated reach than single-sport buyers. And viewers who watched multiple sports on Prime Video spent 12% more and placed 17% more orders on Amazon.

That second figure upends decades of upfront convention. Linear television has always sold reach and, at its most sophisticated, recall. Amazon is selling reach and then telling advertisers what the reached audience bought afterward, on the same account, in the same session, with no panel or modeled attribution path in between. The measurement argument that broadcast networks have been having for forty years does not exist inside a closed system where the screen and the checkout share a login.

That reframing carries a price consequence. Inventory sold on proven order lift does not compete with other television inventory. It competes with search and retail media budgets, which are larger, less bound to an annual negotiation, and judged on return rather than gross rating points. Brands entering upfront conversations this fall should expect Amazon to argue from exactly that position.

Conversational shopping gets its first receipt

The quarter also produced the first performance data on ads placed inside Amazon’s assistant surfaces. Shoppers who click a sponsored prompt convert to a sale 48% more often and spend 21% more on average than those who do not, according to Jassy. Those placements became billable under cost-per-click bidding only in March, which makes this the first readout since advertisers started paying for them.

The surface underneath grew to match. Amazon merged its Rufus assistant and Alexa+ into a single Alexa for Shopping experience in May, and more than 350 million customers used the assistant over the past twelve months. Customers who shop through it spend over 40% more per order than those who do not.

A caution belongs here. Every one of these figures comes from the company selling the inventory, with no independent verification and no disclosure of how the comparison groups were built. Shoppers who engage an AI assistant are already further along a purchase path than shoppers who do not, so some portion of that 48% is selection rather than lift. The disclosure still matters, because a company defends a format with data on an earnings call when it intends to raise the price of that format.

The competitive picture behind the growth

Alphabet reported advertising revenue of $81.6 billion for the same quarter, up 14%. Amazon operates from roughly a quarter of that base and is compounding at nearly twice the rate, which puts its trailing twelve-month advertising business above $76 billion. Advertising now accounts for just under 10% of Amazon’s $200.6 billion in quarterly net sales.

The pressure runs in two directions at once. Against Google and Meta, Amazon is arguing that purchase data beats intent data. Against Walmart Connect, which reached roughly $6.4 billion in 2025 and has been growing faster than Amazon’s advertising line off a much smaller base, Amazon is arguing that scale plus premium video beats scale alone. The live-sports package is the piece Walmart cannot replicate quickly, and the closed retail loop is the piece traditional broadcasters cannot replicate at all.

What the quarter does not prove

Two caveats deserve room in any planning conversation. Prime Day fell inside the second quarter this year after landing entirely in the third quarter in 2025, which flattered the comparison. Amazon told investors that excluding the event from both years would put guided third-quarter growth nearly 400 basis points higher, an unusually direct admission that the calendar did some of the work. Independent measurement complicates it further. CommerceIQ found that United States advertising spend during the June event fell 8.8% even as conversion rates rose 17.1%, meaning brands spent less and converted better rather than simply spending more.

The sports story has a seasonal hole as well. Prime Video carries no NFL until September, so third-quarter sports monetization steps down before the autumn ramp. A sold-out quarter is evidence of demand, not proof of a permanent supply shortage.

What marketers should take from this

The lesson is not that Amazon is winning. It is that the currency is changing. When a platform can price inventory on orders rather than impressions, every buyer who cannot produce that evidence negotiates from a weaker position, and every seller who cannot produce it finds its inventory repriced downward by comparison.

Three questions follow for brands heading into the upfront. Does your category actually convert on Amazon, because multi-sport packages only pay off for products that sit inside that purchase graph. Are you equipped to audit vendor-reported lift, given that the numbers selling these packages come from the vendor. And what happens to your open-web and broadcast budgets once the comparison set includes a channel that reports orders instead of reach.

The 26% will be forgotten by the next earnings cycle. The 17% will show up in media plans for the next two years.