Microsoft Limits AI Token Use for Engineers to Stop 'Tokenmaxxing'
When tech giants sell infinite AI power to the world, they somehow forget to check their own cloud bill. Now top software engineers are getting put on strict token allowances because hype gets expensive real fast.
Internal memos leaked from Microsoft reveal that executive vice president Jay Parikh ordered engineers to stop "tokenmaxxing" after giving them internal access to OpenAI's GPT-5.6 model. Management is rolling out individual AI token budgets for developers, forcing staff to track their personal consumption instead of spamming giant prompts for every minor bug fix.
The sudden frugality stems from corporate panic over skyrocketing infrastructure costs. Earlier this year, GitHub Copilot shifted to a pay-per-token pricing model after users burned through basic quotas too fast, while Microsoft began swapping out expensive third-party models from Anthropic and OpenAI in favor of its cheaper, homegrown MAI models inside Microsoft 365.
This belt-tightening isn't unique to Microsoft either. Amazon, Adobe, Atlassian, and Citi have all cracked down on internal AI spending after Amazon accidentally blew $1.8 million on a single internal system powered by Claude Sonnet before demoting the AI to simpler, cheaper tasks.
Selling the revolutionary future of unlimited artificial intelligence is much easier when someone else pays the cloud bill, but watching multi-billion-dollar tech titans put their own developers on digital food stamps proves the AI boom is running directly into financial reality. The grand promise of automated work seems to hit a hard wall the second corporate finance departments actually look at the invoice.
Source: 404 Media
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.