Large companies including AT&T and Meta have begun capping how much employees use AI tools. As internal spending on generative AI surges, a trend known as "tokenminimizing" — routing work to cheaper models and limiting spending — is spreading.
June 2026 · Enterprise AI
From "Tokenmaxxing" to "Tokenminimizing"
Months after racing to maximize AI usage, Meta, AT&T, Uber and others are slamming on the brakes — capping budgets and throttling access as internal AI bills soar into the billions.
$7,500
per heavy user, per month
~6,000
Meta staff told to limit external AI spend
4 mo
Uber burned its full-year AI budget
The cost shock, drawn to scale
Engineering-team annual token spend — and what agentic tools can do to a bill.
then agentic tools3× it →
Who is pulling back — and how
Meta
Memo to ~6,000 staff limiting external AI spend; building a real-time tracking-and-cap platform.
AT&T
Throttled some employees' GitHub Copilot access.
Uber
Spent its annual budget in four months; introduced monthly caps on some tools.
Amazon
Removed its internal token-usage leaderboard.
Not a kill switch — a gateway
The emerging fix isn't cutting access. "AI Gateway" platforms automatically govern spend in three steps:
ROUTE
Send each request to a cheaper model when possible
→
CAP
Set budgets and spending limits per team
→
MONITOR
Track usage and spend in real time
The likely real winners of this shift: the cost-control tools themselves.
The upside
Governed routing keeps "meaningful usage" alive while reining in runaway bills — smarter than a blanket on/off.
The risk
Abrupt restrictions could chill AI adoption culture, and uneven enforcement leaves the durability of caps in doubt.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…