Apple cut engineers’ daily AI budget from $300 to $200, Bloomberg reports

Apple has started rationing the AI its own staff use. In October the company lowered the daily AI budget for its engineering teams from $300 to $200 a person, Bloomberg reported on 4 November, citing people familiar with the decision. TechCrunch picked up the story the same evening. Apple did not answer TechCrunch’s request for comment.

The new limits

Bloomberg describes three tiers:

• engineers: $200 a day, down from $300;
• some engineering allowances: $100 a day, a cut Apple is preparing;
• staff outside engineering: a cap as low as $25 a day, under discussion.

The caps cover outside AI models that Apple staff use for coding and other work. Several outlets that relayed the report name Anthropic’s Claude, OpenAI’s ChatGPT and Google’s Gemini among them.

What the old cap was worth

We converted the caps into yearly figures, taking about 250 working days:

• $300 a day comes to about $75,000 a year per engineer;
• $200 a day comes to about $50,000;
• $25 a day comes to about $6,250.

At $100 a day, an engineer has to choose which tasks deserve a long agent run. A tighter AI budget also pushes teams toward cheaper models for routine work and keeps the expensive ones for hard problems.

Anthropic, OpenAI and Google charge for their models by the token, the unit of text a model reads and writes. A coding agent re-reads files and its own earlier steps on each turn, so one long task can cost far more than a chat.

From tokenmaxxing to rationing

Earlier this year many tech companies told staff to use AI as much as they could, a habit TechCrunch called “tokenmaxxing”. In April TechCrunch reported that developers who chased token counts were less productive than they believed. HR Dive has reported that companies are not yet saving money with AI. TechCrunch wrote that the frenzy faded once the token bills arrived and companies turned to rationing.

At Apple, John Ternus became chief executive on 1 September and told Wired that Apple does not want to ship “AI for AI’s sake”. In its September quarter Apple also paid higher prices for memory chips, which the AI data centre boom made scarce, as we wrote before the results.

What it means for you

AI vendors price their services on how much companies consume. Apple, one of the world’s richest companies, now counts its engineers’ tokens. If you run an AI agent for trading or research on a paid API, set a hard daily limit in the provider’s console before you start, and log what each task costs.

Details come from Bloomberg’s report as relayed by TechCrunch. Apple has not confirmed the figures. Yearly totals are our estimate.