OpenAI CEO Sam Altman has projected that the usage of tokens — the fundamental units processed by artificial intelligence models — will increase by a factor of one million. The forecast, which Altman shared in a recent statement, points to an enormous surge in demand for AI services and the staggering computational resources that would be required to support it.
What a million-fold token jump means
Tokens are the building blocks that AI models use to process and generate text, code, and other data. Every query to a system like ChatGPT consumes tokens, and as models become more powerful and widespread, token consumption rises. A 1,000,000x increase would mean that for every token used today, a million would be used in the future. That scale implies a future where AI is deeply embedded into daily workflows, enterprises, and possibly entire industries.
Altman’s projection does not come with a specific timeline. But the sheer magnitude suggests the company expects a world where AI inference — the act of running a trained model — becomes a massive utility, akin to electricity or bandwidth.
Infrastructure and energy implications
Such an increase in token usage would require data centers many times larger than today’s largest facilities. Power consumption would skyrocket. OpenAI and its competitors would need to invest heavily in specialized chips, cooling systems, and renewable energy sources. Companies like Nvidia already supply the graphics processing units that power most AI workloads, and a million-fold jump would strain even the most aggressive manufacturing roadmaps.
Altman has previously spoken about the need for massive capital investment in AI infrastructure, but this projection puts a concrete number on the expected growth. It also raises questions about whether the global semiconductor supply chain can keep up.
Economic scale of the projection
Each token currently carries a small cost when processed through OpenAI’s API. A million-fold increase in total tokens would multiply revenue potential if prices hold, or force dramatic price reductions if competition drives costs down. Either way, the economic footprint of AI would expand far beyond today’s market. Cloud providers, chipmakers, and energy firms all stand to be affected.
Altman did not specify whether this growth would come from more users, more powerful models, or both. But the projection signals that the company is planning for a future where AI is not a niche tool but a ubiquitous infrastructure layer.
For now, the industry is left with an open question: when — and by what path — will token usage reach that million-fold mark?
