Altman Outlines OpenAI's Three-Step Strategy at DevDay: Models, Building, and Distribution
nashnova research
OpenAI CEO Sam Altman unveiled a Models → Build → Distribution strategy at DevDay 2026, pushing the company from AI lab toward platform ecosystem. ChatGPT now has 1.2 billion weekly active users — a direct answer to questions about its lofty valuation.
What does the three-step strategy actually mean?
OpenAI slices its business path into three layers: Models (build the best AI) → Build (give developers tools) → Distribution (connect users and partners to make money).
This means → OpenAI no longer wants to be just "the lab with the strongest model." It wants to become a platform company on the scale of a cloud giant.
Altman's own words: "We want to give you the best combination of intelligence and capability at every price point across the entire market."
Models layer: why does GPT-6.1 Sol matter?
OpenAI released GPT-6.1 Sol, positioned as the affordable alternative to its flagship model Astra — similar performance, significantly lower price.
This means → It directly squeezes the mid-market pricing room for Anthropic, Meta, and Google, forcing rivals to cut prices or justify their premium.
OpenAI also launched Ultrafast, a high-speed access tier priced well above the standard layer, targeting enterprise users who demand extreme response speed. In plain terms = grab volume at the low end, grab margin at the high end — a pincer move.
The planned flagship GPT-6.1 Astra was shelved over safety concerns. When it clears safety review remains unknown — a key variable for gauging OpenAI's product cadence.
Build layer: what new tools do developers get?
The coding agent Codex moves to the cloud — developers can now launch and monitor coding tasks remotely across devices, no longer tied to one machine.
Two new APIs ship: a general-purpose Agents API (lets customers build autonomous agents on OpenAI models) and a Decisions API (enterprises pre-set fixed options; AI picks in real time).
In plain terms = Decisions API solves a specific problem: you want AI to help judge, but you don't want to give it unlimited autonomy — give it multiple choice, not an essay question. This drew the loudest applause at the event.
OpenAI also launched ChatGPT Space, a persistent workspace for teams to collaborate with AI agents, blending messaging, document collaboration, and cloud storage.
Distribution layer: what is OpenAI trying to become?
Enterprise customers can now use existing OpenAI spending credits to buy partner products. Users can also spend their own Token quota on third-party AI products via "Sign in with ChatGPT."
This means → OpenAI is trying to build AI's version of "app store + single sign-on" — users stay inside the OpenAI ecosystem, and the money circulates there too.
Altman cited cloud providers as his reference point, saying he wants OpenAI to be the hub that "routes demand and connects customers." This reflects an ultimate ambition not to sell models, but to become the infrastructure middleman of the AI era.
Can 1.2 billion weekly users support the platform dream?
ChatGPT's 1.2 billion weekly active users are the foundational asset of the entire platform strategy — without user scale, the distribution layer has nothing to distribute.
But whether the platform model can truly mobilize a partner ecosystem and convert traffic into sustainable revenue remains the key test for the market to validate OpenAI's valuation.
In plain terms = more users does not automatically mean more money. What OpenAI must prove now is: other businesses are willing to set up shop on your turf — and can actually survive there.
市场有风险，内容仅供研究参考，不构成投资建议。