XNews

OpenAI Agent Milestones and Strategy

OpenAI Agent Milestones and Strategy

Sam Altman announced OpenAI's next phase: transitioning from a product company to a platform company. ChatGPT and Codex have merged, providing a single entry point for users to access personal or enterprise AGI, while developers build products via API. OpenAI aims to cover the full price-performance curve, targeting 100 million new enterprises and 8 billion global users. Separately, OpenAI's Agent active users reached 20 million, with quarter-to-date annualized revenue up 35% and enterprise revenue growing over 50%. Agent user penetration is only 2% compared to ChatGPT's approximately 1 billion active users. Additionally, ChatGPT launched a $100 Business team subscription with 5x usage, no 5-hour limit, dedicated company workspace, data access controls, and default opt-out from model training.

Key insights

OpenAI's Platform Strategy Aims for Universal Reach

Sam Altman explicitly stated OpenAI will not compete in every product category. Instead, the merged ChatGPT and Codex platform will serve as a single entry point for both consumers and enterprises, while developers build on top via API. The stated goal of serving 100 million new enterprises and 8 billion people indicates an ambition to become the underlying infrastructure layer for AI, similar to how cloud providers became the infrastructure for the internet.

Agent Adoption Shows Strong Growth but Low Penetration

OpenAI's Agent active users reached 20 million, with enterprise revenue growing over 50%. However, compared to ChatGPT's 1 billion active users, Agent penetration is only 2%. This suggests significant room for growth as more users transition from conversational AI to task-executing agents. The 35% increase in annualized revenue indicates strong monetization momentum.

Business Subscription Targets Team Collaboration

The $100 Business subscription introduces enterprise-oriented features: 5x usage limits, removal of the 5-hour restriction, dedicated company workspaces, integration with company data and software, role-based access controls, and default data privacy from model training. The minimum two-seat requirement positions this as a team collaboration product rather than an individual power-user tier.

Source posts