AI Token Spending Growth Slows as Prices Fall 48% Since July

AI token spending grew just 0.2% month over month in September, the slowest monthly increase since November 2025.
That follows growth of only 1.0% in August, marking two consecutive months at or below 1%, the weakest such stretch in at least 18 months.
At first glance, the slowdown could look like weakening AI demand.
But pricing tells a different story.
Since July, the average price of processing one million tokens across leading AI models has reportedly fallen 48% to around $0.65, the lowest level since at least January 2026.
Models are simultaneously becoming cheaper, more efficient and more capable.
That means dollar spending can grow slowly even while the underlying volume of AI usage continues expanding.
This is an important distinction for the AI economy.
Lower inference costs reduce the barrier to deploying AI across software, customer service, coding, agents and other high-volume applications.
So falling token prices could pressure revenue per unit of compute while also unlocking significantly greater consumption.
The next phase of AI adoption may depend less on expensive frontier access and more on what happens when intelligence becomes cheap enough to use everywhere.