Apple has cut the amount its engineers can spend on third-party AI coding tools each day, from $300 down to $200 — and it might not stop there.

According to a Bloomberg report carried by Gadgets 360, the company has told staff the cap could drop again to $100 a day if costs keep climbing. For a company that has spent years building its reputation on premium hardware, the move is a reminder that even Apple is watching its AI bills closely.

Why Apple is tightening the cap

The main cost driver is Anthropic's Claude Code, the AI coding assistant many Apple engineers rely on. Apple is said to pay roughly $200,000 a year per engineer for the licence, and its most active users were already hitting the old $300 daily limit. That kind of spending adds up fast when thousands of engineers use the tool every day.

The timing matters too. Apple has been training its own models and leaning more on outside tools to speed up development, especially as it pushes AI features across the iPhone, Mac and its services. The company spent about $4.5 billion on AI training this year, and the costs of running large models keep rising.

The numbers behind the cut

  • Daily limit: $300 down to $200, with $100 possible next
  • Anthropic licence: about $200,000 per engineer per year
  • AI training spend this year: about $4.5 billion
  • Q4 revenue: $113.7 billion, a record for the quarter

The cut is small next to Apple's overall numbers, but it sends a clear signal inside the company: AI tools are now a real budget line, not a free perk.

What it means for engineers

For most engineers, the lower cap will not change much day to day. The $200 limit still covers a lot of coding help, and the tools remain available. But heavy users — the ones building and testing the most complex features — will need to be more careful about how much they lean on the most expensive models.

Apple is not alone here. Other big tech firms have been quietly reworking their AI budgets as model costs rise, and some have started shifting work to cheaper in-house models. Apple, too, has been working on its own AI tools, and it will be interesting to see whether this push nudges more engineers toward them.

For now, the message is simple: AI coding tools are powerful, but they are not free — even at Apple.

Read the full story on Rohan Tech Official.