OpenAI: Top token user consumes 100 billion tokens a month, Altman says

OpenAI CEO Sam Altman told a livestream audience that the company's top internal token user is consuming roughly 100 billion tokens per month and that the company has even found an external user spending more. Altman said this rapid rise in token usage has pushed AI budgeting to the forefront for many businesses.

According to Altman, the change is stark: six years ago OpenAI's leading token consumer used around 100,000 tokens per month, making the current figure roughly a million-fold increase. Altman made the remarks during an event focused on enterprise adoption, highlighting both the scale-up in usage and the operational surprises it has produced for developers and customers alike. Independent reports and developer disclosures have surfaced similar high-consumption examples, with some projects reportedly burning through hundreds of billions of tokens in a single month.

Such extraordinary token consumption raises immediate questions about compute, cloud and energy costs. Providers and AI labs face an infrastructure challenge: meeting explosive demand requires large investments in GPUs, data centre capacity and power, which can compress margins unless usage is made more efficient or prices adjust. Executives have recently warned that rising compute bills will be a defining operational issue for AI businesses.

In the broader context, token-based billing is now a material line item in enterprise IT budgets as firms move from pilots to production. Organizations must implement cost governance, routing and usage controls to prevent surprise invoices, and they are increasingly weighing on-premise or reserved-capacity approaches versus pure cloud consumption. This shift creates opportunities for cloud and hardware vendors while forcing CFOs and CIOs to collaborate more closely on AI procurement and budgeting.

Market observers say the immediate winners are likely to be GPU suppliers and large cloud hosts, but the long-term picture depends on whether token costs decline through model efficiency gains or whether customers demand lower prices or alternative deployment models. Analysts expect token-spend scrutiny to intensify in earnings calls and corporate budgets over the coming quarters, making cost management a core competitive factor for AI vendors.