Atlassian Caps AI Spending for Engineers as ‘Tokenmaxxing’ Costs Bite
July 30, 2026 – 7:44 am
Atlassian has implemented an AI budget system for its engineers in response to the escalating costs of "tokenmaxxing" – the practice of maximizing AI model usage. The software company recently laid off thousands of employees in the name of AI integration and is now providing capped ‘AI wallets’ to the remaining engineers, warning them when they approach their spending limit.
The company justifies this move as a way to encourage responsible AI exploration while controlling costs, though the irony is evident given Atlassian’s previous aggressive promotion of "AI-first" initiatives.
The Shift in AI Spending Priorities:
While Atlassian frames its wallets as generous with safeguards, others are taking a more cautious approach. Amazon recently discontinued an internal competition focused on excessive AI usage due to employee manipulation of token systems for personal gain. Meta went even further, warning 6,000 employees that their 2026 internal AI spending could reach the billions and began implementing similar budget controls.
This reversal in sentiment reflects a broader trend. Last year’s focus was on maximizing AI adoption ("tokenmaxxing"), but this year, companies are increasingly emphasizing managing AI spend – focusing on cost per outcome rather than sheer token volume.