OpenAI's Sam Altman says the world should get ready for a steep climb in AI token usage. He frames intelligence itself as a utility — something to be consumed like electricity or water — and argues that shift will force new ways of managing how much AI people use and what it costs.
Why token usage is set to explode
Altman's prediction is blunt: token consumption will grow at an exponential rate. Tokens are the basic units of text and code that AI models process. Every query, every generated paragraph, every line of code burns through them. As AI tools get woven into daily work and life, the sheer volume of tokens processed could dwarf anything seen so far.
He didn't offer a specific timeline or numbers. But the logic is straightforward. The more useful AI becomes, the more people use it. And the more it's used, the faster token demand multiplies. That isn't a modest uptick — it's a curve that bends sharply upward.
Intelligence as a utility
Altman has long compared AI to a utility. In this view, intelligence isn't a product you buy once; it's a service that flows on demand. You don't own the water company — you turn on the tap. Similarly, users would tap into AI when they need it, paying for what they consume.
That framing matters because it changes expectations. Utilities are regulated, metered, and priced per unit. They're also assumed to be always available. If AI is heading that way, it raises questions about reliability, access, and who gets to set the meter.
Economic models under pressure
Treating intelligence as a utility doesn't just change how individuals interact with AI — it could reshape economic models. Entire industries built around selling static software or one-time knowledge products may need to rethink their foundations. When intelligence is metered like power, the value shifts from owning the resource to distributing it efficiently.
That could ripple through pricing, competition, and even how companies measure productivity. If a business's output depends on how many tokens it burns, token costs become a core operating expense. That's a different economics than paying for a license or a server.
Managing consumption and costs
With exponential growth comes a pressing problem: keeping consumption and costs under control. Altman's prediction implies that organizations and individuals will need new strategies to manage their AI usage. That might mean smarter routing, caching, or tiered access — but the facts here don't specify the methods.
What's clear is that the old approach of treating AI as a discretionary tool won't hold. When it becomes a utility, you can't just ignore the bill. You need to budget for it, monitor it, and find ways to use less without losing the benefit.
The next step is watching how OpenAI and others respond to that pressure. Pricing models, consumption caps, or new efficiency tools could emerge. But for now, Altman's forecast sets the terms: AI usage is going to explode, and the only question is who figures out how to handle the cost.
