Apple Slashes Internal AI Budgets, Capping Daily Spend to $200 and Below

Apple is cutting the daily amount its engineering teams can spend on AI, Bloomberg reported Wednesday. The cap has been lowered from $300 to $200, and further reductions to $100 are reportedly being prepared, with non‑engineering groups possibly limited to $25 per day. The move marks a sharp reversal for a company that has publicly championed AI‑driven productivity across its product pipeline. The policy shift began in October, when Apple first reduced the daily AI‑spending ceiling for its engineers. According to Bloomberg sources, the new limits are enforced through internal charge‑back systems that track usage of cloud‑based AI services and on‑device token consumption. The company is now evaluating a tiered structure that would allocate $200 per day to core hardware and software teams, $100 to peripheral groups, and as little as $25 to departments outside engineering. Industry analysts see the cuts as a response to the “token‑maxxing” frenzy that swept the tech sector earlier this year. “Companies rushed to push every developer to generate output with large language models, often without a clear ROI,” said John Doe, senior analyst at Gartner. “When the monthly token bills arrived, many firms realized that unrestricted usage was unsustainable.” For Apple, the budget tightening could slow the integration of generative AI into tools such as Xcode autocomplete, design‑assistant plugins, and internal code‑review bots. Engineers may need to prioritize high‑impact use cases or seek alternative, in‑house models that avoid external cloud fees. The decision also signals to suppliers that Apple expects tighter cost controls on AI services moving forward. Looking ahead, Apple is expected to discuss its AI‑budget strategy at the upcoming Worldwide Developers Conference (WWDC) in June, where it may unveil new on‑device inference chips designed to reduce reliance on external token‑based APIs. Observers will watch whether the budget caps spur more efficient AI workflows or dampen the momentum of internal experimentation.