Photo: TechCrunch / Wikimedia Commons / CC BY 2.0 (https://creativecommons.org/licenses/by/2.0)
The average person could consume 500 billion AI tokens per month by 2033, up from 100,000 just six years ago
OpenAI CEO Sam Altman dropped a number in June 2026 that deserves a second look. The company’s heaviest internal user now processes roughly 100 billion AI tokens every month. For context, the peak monthly consumption back in late 2019 was about 100,000 tokens. That is a one-million-fold increase in roughly six and a half years.
From niche tool to something resembling a utility bill #
AI tokens are the basic units of text and data that large language models process when generating responses, writing code, or analyzing documents. Altman and his peers are increasingly framing AI intelligence as a metered utility, something closer to electricity than software.
That 100,000-token peak from late 2019, which once represented the absolute ceiling of what anyone was doing with AI, now reflects the global per capita average. Looking ahead, Altman projects that individuals could average 500 billion tokens per month by around 2032 to 2033. Leading-edge users might consume quadrillions of tokens monthly.
The $1.3 million monthly tab #
A single external user reportedly burned through 603 billion tokens in one month, racking up a bill of $1.3 million. That figure already exceeds the 100 billion tokens consumed by OpenAI’s own top internal user, suggesting that the most aggressive consumption is happening outside the company’s walls.
In 2026, enterprises across industries are scrambling to implement consumption controls, usage caps, and metering strategies to avoid runaway AI spending.
Why the curve keeps steepening #
Altman pointed to a concept he called “constant running proactive AI” as a key driver of future growth. The idea is that AI won’t just respond when prompted. It will run continuously in the background, monitoring systems, generating insights, and taking actions without waiting for a human to type a question.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our