Microsoft EVP: 'tokenmaxxing is not what we are optimizing for' - division-level AI token budgets imposed as engineers spend thousands each per month

What happened

An internal email from Microsoft EVP Jay Parikh, first reported by 404 Media, told employees to curb AI token consumption and set formal AI token budget targets at the division level. Internal guidelines acknowledged many engineers spend from hundreds to a few thousand dollars a month in tokens. As of July 2026 every Microsoft division has an AI token spending cap, employees track usage on an internal dashboard, and the default internal model was switched to a cheaper alternative. Microsoft had already cancelled most Claude Code licences in its Experiences and Devices group in May 2026. TNW reports the same pattern at AT&T, Meta, Uber, Walmart, and Amazon: per-token prices fell roughly 98% since late 2022 while enterprise AI bills tripled, because agentic tools consume dramatically more tokens per task.

Causal vector

Uncapped per-engineer token consumption across a 200,000+ person organisation; agentic tools consume orders of magnitude more tokens per task than seat-based budgeting assumes

Source

Reported by The Next Web (citing 404 Media and an internal Microsoft email). Verified against the primary report.

Read the original report ↗