Eric Seufert on Apple's coming ad empire, Netflix's fast dilemma, and why OpenAI needs conversion optimization to reach $10B in ad revenue

Eric Seufert on Apple's ad ambitions, Netflix's pricing trap, and OpenAI's path to $10B

Eric Seufert, founder of Heracles Media and Mobile Dev Memo, makes the case that three of the most important ad stories right now are Apple's quiet infrastructure build toward a generalized ad network, Netflix's increasingly uncomfortable position between premium and free, and OpenAI's need to unlock conversion-based bidding before its ads business can scale meaningfully.

---

Netflix's pricing trap

Netflix's premium tier is approaching $30 a month in the US, and Seufert argues the company may have extracted about as much as the market will bear. Password-sharing crackdowns are largely played out. Live events and sports — the Beyoncé Bowl, January NFL games — are expensive growth levers. And the broader streaming market has moved in the opposite direction, toward free ad-supported television, partly because Netflix's own pricing pressure pushed competitors there.

That leaves Netflix threading a difficult needle. Its leadership says it can never go pure FAST because that would damage the premium brand perception. But Seufert thinks the bundling strategy reported by the New York Times — offering subscription add-ons inside the Netflix app — makes that position hard to sustain. A subscriber already paying $30 a month who then sees a prompt to pay more for additional channels is unlikely to stay. The only coherent path through bundling, in his view, is to adopt FAST.

Amazon is the identity backbone across all of FAST CTV, running data deals that give it access to the best impressions across the ecosystem including Roku. Seufert's read is that Amazon never needed to acquire Roku because it was already cherry-picking the most valuable inventory through those partnerships.

“Netflix may have reached a ceiling with what they can charge... Now they're faced with: we just have to bring in a lot of live events and sports, things like the Beyonce Bowl and the January NFL games. It's really expensive. And on OpenAI ads: once you are bidding against a specific outcome, then the growth inflects. That's the gap from one to ten.”

YouTube vs. Netflix: the creator wars

Netflix's summer shopping spree for YouTube creators has forced YouTube to behave in ways it historically resisted. YouTube is now inking exclusivity deals, threatening to deprioritize creators in recommendation systems if they put content on Netflix, and cutting them out of brand revenue if they do. Miss Rachel's season two on Netflix performed substantially worse than season one by view hours per minute of content, which Seufert attributes partly to Netflix's narrower and less deep catalog compared to YouTube's recommendation surface.

YouTube's dominant platform in the US is now the television, making it a direct engagement competitor to Netflix rather than a supplementary one.

---

Apple's ad empire in the making

Apple's ad business is under-monetized, and the commission side of the App Store has nowhere to go but sideways or down. The EU's DMA, plus cases in Japan and Brazil, are steadily eroding Apple's ability to enforce its 30% take rate. Its latest Epic v. Apple proposal — a 15% commission on link-out purchases — is more reasonable than its previous position, but it still includes reporting requirements that may limit uptake. Seufert notes that subscription apps have already largely migrated monetization to the web, sending users through Facebook ads to a website where they register before downloading the app, bypassing in-app purchase entirely.

The growth lever Apple has that can't easily be taken away is engagement — 900 million weekly active users across the App Store. Several moves signal where this is heading. Apple recently added a second ad placement in App Store search results, added ads in Maps, rebuilt its attribution framework (rebranding SKAdNetwork as the Ads Attribution Kit), renamed Apple Search Ads to Apple Ads, and — most significantly — updated its advertiser services agreement to permit serving ads on third-party websites and apps Apple doesn't own. That last change is the clearest structural signal. A unified campaign optimization API built to accommodate Maps and other new placements is already in place and extensible to any new surface.

Seufert also flags the likely Apple AI monetization play: a deal structured like the Google search default arrangement, where a model provider pays to be the default attached to Apple's hardware and services. Apple has already built the infrastructure for it through Private Cloud Compute and the Core AI framework.

---

OpenAI: conversion optimization is the gap between $1B and $10B

ChatGPT's ad revenue is heading toward $1 billion, and OpenAI has expanded its ads product to more than 40 countries, adding 31 in recent weeks. Seufert argues the inflection from $1B to $10B is a single product change: true conversion-based bidding, where advertisers bid against a specific outcome rather than clicks. That's the mechanism that unlocks the auction dynamic — the more value delivered, the more advertisers bid, the more revenue the platform makes. Affiliate monetization and CPC don't get you there because they tend to preference cheap, high-converting goods rather than surfacing real latent value from the advertiser base. Getting from $10B to $100B after that is largely an SMB onboarding problem.

Seufert is also skeptical that ads-to-agents ever works structurally. The incentive conflict is too direct — the agent and the advertiser can't both be trusted at the same time — and ads require a visual component for brand positioning and affinity that voice and agentic interfaces fundamentally lack.

---

Meta's underappreciated AI opportunity

The most commercially durable thread in Meta's AI push isn't Llama model releases or standalone consumer apps. It's the business-facing layer. Meta AI now includes functionality for managing and optimizing ad campaigns — an area where tools like Claude and Codex are being used today but aren't purpose-built for it. No company understands ad campaign optimization better, and even if the use case were limited to Meta's own platform, the commercial opportunity is substantial.

The AI-enabled pixel is another move worth watching: it can self-tune with minimal input from the advertiser, and Seufert sees it as the opening position in a deeper play to handle landing page optimization and conversion personalization for SMBs that lack the resources to run continuous A/B testing. Many of those businesses, he argues, will willingly hand that capability to Meta if it means more sales and more efficient ad spend. Meta's Business AI is also integrated into WhatsApp, enabling direct business-to-consumer communication at scale.

The Robin Media Mix Model, released as open source, fits the same logic. Meta believes it's being under-attributed in multi-channel measurement, and an open measurement framework that surfaces that under-attribution leads advertisers back to increasing Meta spend — not because Meta put its thumb on the scale, but because the optimization actually reflects where the value was.