Source linked

El 1% de las empresas que quema 7.500 dólares por empleado en IA cada mes

Los principales consumidores - las empresas "compuestas por la IA" - gastan más de 650 veces la empresa mediana, pero todavía están por detrás del salario promedio de los ingenieros de software de 16.000 dólares al mes.

rampramp ai indexai spendingenterprise ainvidiamercor

$7,500 per employee per month. That's what the top 1% of American companies — what Ramp calls "AI-pilled" firms — are spending on AI compute and tokens. An Nvidia executive recently said compute costs have surpassed employee salaries at his company. Mercor's CEO told a similar story: the startup now spends more on tokens for internal agents than on human headcount.

The Median Firm Spends $11.38 per Employee

Ramp's AI Index, which tracks AI adoption across U.S. businesses, reveals a staggering gap between power users and the rest. The top 10% of firms spend $611 per employee per month. The median company? $11.38 — roughly the cost of a single enterprise SaaS seat. That means the top 1% outspend the median by over 650 times. AI spending is highly concentrated, not yet a universal line item.

Compute Costs Are Outpacing Salaries for Power Users

Even at $7,500 per employee, these AI-pilled firms still spend less than the average software engineer's monthly salary of about $16,000. But the trend is accelerating. Among the top 1%, AI spending grew 14.1% per employee last month alone. If that rate holds, they'll cross the salary threshold within months. The top firms aren't married to a single model — they bounce between frontier models and cheaper open-source alternatives to keep costs in check.

Whether the 14.1% monthly growth holds depends entirely on whether those firms see proportional returns. For now, inference cost is becoming a line item that rivals rent, and the median company hasn't even started.


Source: 'AI-pilled' firms spend $7,500 per employee each month on AI
Domain: techcrunch.com

Read original source ->

External source stays available while the OJO article and comment thread stay local.

Comments load interactively on the live page.