Top Enterprises' AI Spending Drops 10% MoM in August, Model Price War Is the Main Driver

nashnova research
今天发布阅读约 6 分钟

Top-tier corporate AI spending posted its first notable decline — August per-employee spending for the top 1% dropped from $8,000 to $7,200, driven not by weakening demand but by model price cuts and a shift to cheaper tiers.

01

Why did the biggest AI buyers' bills suddenly shrink?

The Ramp AI Index shows that the top 1% of companies by AI spending saw per-employee monthly outlay fall to $7,200 in August, down roughly 10% from the July peak of $8,000.
This is the first notable pullback since the index began tracking.
Ramp chief economist Ara Kharazian attributed the drop to a "model price war." This means → the bills shrank not because firms used less AI, but because the same workload now costs less.
02

Where exactly did prices come down?

Kharazian pointed to "price cuts combined with a structural shift toward standard and lightweight model tiers" as the core driver.
In plain terms = companies stopped defaulting to the priciest frontier models and moved everyday tasks to cheaper "good enough" versions.
He specifically noted this round of decline was not driven by greater access to Chinese open-source models, but by direct pricing competition among model vendors.
03

Only the top is cooling — what about everyone else?

In sharp contrast, the top 10% of companies saw per-employee monthly spending rise to about $680 in August; the median company climbed to roughly $14.
Both groups have trended upward continuously since January 2024, with no interruption.
This reflects a widening split in AI spending — top buyers are optimizing costs, while most companies are still in the "just starting to spend" phase.
04

What does this mean for the AI market?

Kharazian said the data shows "more cracks in AI investment logic," with multiple spending drivers approaching a ceiling.
Ramp's tracking covers large-language-model subscriptions, coding agents — AI tools that help developers write code — API calls, and GPU cloud-computing spend.
This means → the top-buyer pullback looks more like a structural repricing than a demand contraction — but whether the biggest spenders have hit a spending plateau is the key variable to watch next.

市场有风险,内容仅供研究参考,不构成投资建议。