Topics

Spend controls

0 articles and 2 community signals on Spend controls for people running AI in production, written and curated by Wren with every source linked.

Updated Nov 18, 2026

What practitioners are saying

All signals →
  • AWS Machine Learning blog

    Airbnb on the AWS blog: a model that tested at the same cost per task ran up several times the weekly spend per user once staff had it

    Airbnb engineers write that one model matched its predecessor on cost per task in tests, but in use each request cost about three times as much and users sent about three times as many, as people ran long unattended sessions. They say moving one model's reasoning setting from low to high added 0.8 points of score at 2.5 times the cost; a later model did better at about a sixth of the cost.

    Why it matters If you run IT or own the AI budget: a pre-rollout cost test did not predict the bill, because people changed how they worked. Compare spend per user after any model switch. About coding agents; no dollar figures; published by AWS.

  • Business Insider

    Business Insider: Grafana Labs' CTO says AI token bills run to tens of millions of dollars a year, and top-model access is now granted 24 hours at a time

    Business Insider reports Grafana Labs CTO Tom Wilkie saying token bills are in the tens of millions of dollars a year, and that after an analyst left a model running and spent tens of thousands, access to the strongest models is granted in 24-hour increments. It cites Anthropic's prospectus: under $75 a month for the average business user without Claude Code, about $1,500 for the top 5% using it.

    Why it matters If you run IT or finance: one unattended job cost a named company tens of thousands of dollars. Check whether your AI accounts have a per-user cap or time-limited access to the costliest models. Executives' own accounts, partly paywalled.