AI token prices are cooling — but why?

A measure of daily spending on AI usage has fallen since its peak in May — although interpreting what the decline means is challenging.

The Silicon Data LLM Token Expenditure Index (SDLLMTK) is a daily snapshot of the state of the market. It draws data from a multitude of providers and produces a blended rate, expressed in US dollars per one million tokens. The Index currently stands at 1.62, an increase from the index’s inception in December last year, but down 20% lower than its peak in May.

Because the index weights frontier model and open-weight model usage differently, it’s difficult to identify the causes of the decline. Are enterprises pushing vendors to lower prices? That won’t be good news for those AI businesses going for an IPO,

Is there a backlash against AI as some organizations are finding the downsides? There has been some public reaction to job losses and the attack on human creativity, causing AI supporters to be booed at university campuses. And there is also resistance to building new data centers to run AI models.

Or are users simply switching to less token-heavy models?

AI vendors and their customers alike are certainly facing a dilemma. There’s the perception that AI is the future, that it can enhance productivity and ultimately save costs, but there are a host of other issues to look at as companies attempt to justify AI spend by pointing to ROI and finding it hard to calculate.

It’s only a snapshot, but the Silicon Data Index may be the first sign that the rush for AI may just be slowing down.

This article first appeared on InfoWorld.