OpenAI moves to token pricing and eyes big ad revenues

OpenAI just changed how one of its coding products charges customers, and the change sounds small until you look at who now pays the bills. On April 2, OpenAI said Codex billing would move from per-message pricing to the same kind of token pricing used in its application programming interface, which counts usage more like a water meter than an all-you-can-eat buffet. (help.openai.com) A token is a small chunk of text, so token billing charges for how much code and chat actually flow through the system instead of how many times someone clicks send. OpenAI said the new Codex rate card applies to new and existing ChatGPT Business plans and to new ChatGPT Enterprise plans, while some older plans stay on legacy pricing for now. (help.openai.com) That shift lines up with the kind of customer OpenAI is chasing. On April 8, OpenAI executive Denise Dresser told CNBC that enterprise now makes up 40% of the company’s revenue and is on track to equal consumer revenue by the end of 2026. (cnbc.com) Big companies usually want bills they can map to departments, projects, and budgets, and token counts are easier to audit than message counts. If one engineer sends three tiny prompts and another runs a 50,000-word code review, per-message pricing treats them as similar events even when one job uses far more computing power. (help.openai.com) OpenAI is making that enterprise push while talking openly about a second business that works very differently: advertising. Investing.com reported this week that OpenAI projects $2.5 billion in ad revenue in 2026 and has set a long-term target of $100 billion by 2030. (investing.com) Those two revenue streams pull the product in opposite directions. Enterprise buyers usually pay for privacy, account controls, and predictable invoices, while advertising works best when a product has huge daily traffic, rich signals about user intent, and places to put sponsored results. (cnbc.com) (investing.com) That is why the Codex pricing change is more than a billing footnote. When a company switches a coding tool from counting messages to counting input and output text, it is building a cleaner yardstick for procurement teams that sign large contracts and ask exactly what they are buying. (help.openai.com) It also gives OpenAI a way to make different parts of the business look more legible as it prepares for public-market scrutiny. In the same CNBC interview, Chief Financial Officer Sarah Friar said it is “good hygiene” for an $852 billion company to look and act like a public company, even though she would not give an initial public offering timeline. (cnbc.com) So the picture now is not just “sell more artificial intelligence.” It is a company trying to prove it can be three things at once: a metered software vendor for businesses, a mass-market consumer product, and an advertising platform large enough to chase a $100 billion annual target by 2030. (help.openai.com) (cnbc.com) (investing.com)