Adam Heimlich (representing the buy side) and Gareth Glaser (representing the sell side) dive into a packed earnings season episode examining major structural shifts across digital advertising. The hosts open with a deep dive into the emerging "infrastructure layer" thesis, comparing cloud plays like AWS Fabric and Amazon Publisher Services (APS/TAM) to Prebid setups. They analyze how Amazon gains auction advantages by running proprietary code on publisher pages outside Prebid’s rules, while also highlighting Amazon Bedrock's leadership in multi-LLM model management.The central theme of the show turns to the "Revenge of the SSP". Gareth and Adam break down why supply-side platforms are reinventing themselves as ad tech infrastructure providers through real-time containerization (ARTF) integrations with companies like Chalice across Index Exchange, OpenX, PubMatic, and Magnite. They explore PubMatic's stellar earnings driven by a 30% reduction in infrastructure costs, Magnite's hybrid cloud/on-prem strategy, and how curation accidentally achieved the true goals of Supply Path Optimization (SPO).The hosts also contrast DSP data-hoarding with SSP log-file transparency, dissect Nielsen's acquisition of DoubleVerify alongside Cybids and Rockerbox, and engage in a spirited debate over biddable media versus packaged direct sales using an "omakase vs. buying the whole cow" butcher metaphor. They wrap up with a spicy "Postmortem" segment roasting Jonah Goodhart’s case study on serving ads to AI agents, with Gareth framing prompt-injecting LLMs as nothing more than black-hat SEO.2. Key TopicsThe Infrastructure Play & Amazon APS Advantage: Gareth and Adam examine how scaled ad tech is shifting toward cloud infrastructure layers, noting that Amazon's APS/TAM gains an edge by running outside Prebid's open-source rules and timeouts."Revenge of the SSP" & Real-Time Containerization (ARTF): Adam and Gareth explain that SSPs are thriving by transforming into open architecture and containerized decisioning hubs (ARTF), enabling tools like Chalice to lower infra costs and drive publisher revenue.DSP Data Hoarding vs. SSP Log Transparency: The hosts contrast SSPs' willingness to share raw log files for custom algorithmic optimization against DSPs' historical tendency to withhold logs under the pretense of privacy.Direct Sales ("The Whole Cow") vs. Biddable Programmatic ("Omakase"): Gareth and Adam debate inventory packaging, arguing that direct-sold packages force buyers to pay for unwanted filler ("the whole cow"), whereas impression-level biddable programmatic delivers targeted value ("omakase").Ads for AI Agents & Black-Hat Prompt Injection: In the Postmortem segment, Gareth critiques Jonah Goodhart's case study on context-injecting ads into LLMs, calling it spammy black-hat SEO that models like Claude will actively reject.00:00 Show Intro & Summer Earnings Energy02:26 The Fabric Architecture & Infrastructure vs Application Layer04:43 Amazon APS/TAM & Prebid Rule Bypassing12:15 Amazon Bedrock & Multi-LLM Modular Architecture14:42 Real-Time Containerization (ARTF) Across Index, OpenX & PubMatic16:58 PubMatic Earnings Blowout & Cost Reductions21:05 Magnite's Hybrid Cloud/On-Prem Whitepaper24:18 "Revenge of the SSP" & Accidental SPO27:04 DSP Shifts, Trade Desk & Log File Access34:02 John D'Souza Comment on Amazon Logs & AMC Clean Rooms36:58 Nielsen Acquiring DoubleVerify, Cybids & Rockerbox43:36 Ali Manning on AdExchanger & The Origins of ARTF45:21 Regional Allocation Optimization (YouTube, Meta & DSPs)48:24 The Great Packaging Debate: Omakase vs. The Whole Cow54:51 Agentic Buying & Strategy Selection58:07 Postmortem: Jonah Goodhart vs. Will Kakowski on Ads for AI Agents

Podden och tillhörande omslagsbild på den här sidan tillhör Adam and Gareth. Innehållet i podden är skapat av Adam and Gareth och inte av, eller tillsammans med, Poddtoppen.