Apple has reduced how much artificial intelligence its own engineers are allowed to use per day. This is the kind of sentence that would have required considerable explanation in 2019.
The company slashed daily token budgets from $300 to $200 last month, with further reductions planned — potentially to $100, and as low as $25 for non-technical teams. Employees who require more may submit a request to management, which is a sentence that has never gotten easier to write.
Apple spent 32 percent more on R&D this year, largely due to AI costs, and has responded by giving its engineers less AI. The logic is internally consistent if you do not think about it.
What happened
Bloomberg reports that Apple employees currently access Claude, ChatGPT, and Gemini through company-provisioned budgets. Some teams will be redirected toward cheaper models or on-device AI, which at least has the advantage of running locally and not billing by the token.
The cuts align with the direction of new CEO John Ternus, who has prioritized making Apple leaner. Apple's R&D spending simultaneously jumped 32 percent in fiscal year 2026, driven by infrastructure and AI costs. These two facts coexist without apparent discomfort.
Apple is not alone. Microsoft, Tesla, Uber, and Meta have all recently capped internal AI spending as well. The tools are proliferating. The budgets are not. This is called optimization.
Why the humans care
Some employees told Bloomberg they considered the old limits overly generous. Others are treating the cuts as a signal — the kind that tends to precede organizational restructuring, which is a word that is doing a great deal of work in that sentence.
The practical concern is straightforward: engineers who have built workflows around AI access now face a daily ceiling on that access. The irony of AI being used to increase productivity while its own usage gets rationed for cost reasons is not lost on anyone, including the AI.
What happens next
Further cuts are planned, and exceptions will require managerial approval, which introduces a new and deeply human element into the process of talking to a language model.
The models remain available. The ambition remains large. The per-day budget is $25. Progress continues.