06-reference

stratechery ads in chatgpt

Mon Jan 19 2026 19:00:00 GMT-0500 (Eastern Standard Time) ·reference ·source: Stratechery ·by Ben Thompson
openaiadvertisingbusiness-modelsai-economicsaggregation-theory

“Ads in ChatGPT, Why OpenAI Needs Ads, The Long Road to Instagram” — Ben Thompson

Why this is in the vault

Critical analysis of why AI companies need advertising models — directly relevant to understanding AI business model sustainability and the ad ecosystem RDCO operates in.

The core argument

OpenAI announced ads coming to ChatGPT’s free and $8/mo Go tiers. Thompson argues this was inevitable and overdue. Unlike traditional tech companies where costs are mostly fixed and revenue grows faster, OpenAI’s CFO admitted that compute costs track revenue roughly linearly — more revenue requires proportionally more compute. Advertising is the only model that can bend this curve because it increases both breadth (more free users = more inventory) and depth (unlimited ARPU growth funded by advertisers, not users).

The initial contextual ads (based on current conversation) are a starting point but carry risk of appearing to influence answers. The endgame should look like Meta’s Instagram ads — personalized to the user independent of current context, drawing on deep user understanding. Thompson argues OpenAI waited too long; Google can subsidize Gemini indefinitely with Search revenue, making the competitive window narrower.

Key insight: OpenAI’s compute costs are exponentially increasing (0.2GW in 2023, 0.6GW in 2024, ~1.9GW in 2025), tracking alongside revenue growth — this is fundamentally not a scalable tech business without advertising.

Mapping against Ray Data Co

Directly relevant to RDCO’s advertising and autoinv work. Understanding how AI platforms monetize through ads shapes the competitive landscape for automated ad management. The Meta vs. OpenAI ad model comparison is important strategic context.