Apple is big on AI for everyone — but inside the company, it's rationing. According to a report from Bloomberg's Mark Gurman relayed by MacRumors, Apple quietly cut the daily AI spending allowance for some engineers from $300 to $200 last month, and more cuts are on the table: some engineering teams could drop to $100 a day, while groups outside engineering — already capped at $100 — could fall to as little as $25.
The caps cover AI "tokens," the units of data models process per request, across the tools Apple's engineers actually use: Anthropic's Claude, OpenAI's ChatGPT, Google's Gemini, and others. Coding is the expensive habit — that's why engineers had the biggest budgets in the first place. Employees who genuinely need more can appeal to senior leadership for an exemption, and Apple is reportedly nudging some teams toward cheaper models or on-device AI instead.
Here's the awkward part: some Apple employees told Bloomberg that few people were even using their full $300 daily limit, so the actual savings may be small. This looks less like a real budget fix and more like a cultural signal from a new CEO two months into the job. John Ternus took over on September 1 and has been pushing Apple to run faster and leaner — cutting management layers, dismissing engineering program managers in September, and, per MacRumors, bringing in outside productivity consultants for some divisions. That last detail has employees reading the tea leaves: the AI rationing and the consultants together have fueled worries about a restructuring.
The context makes the rationing look almost theatrical. Apple's annual report, filed Tuesday, shows R&D at $45.8 billion — up 32% on AI and infrastructure — and operating expenses up 22% to $75.5 billion. Against that, clawing back $25,000 a year per engineer who actually maxed the old cap is a rounding error. But it's the same logic Apple just aimed at its customers: yesterday's news that server-side Siri AI features will carry daily usage limits, with expanded access sold for a fee. Inside and outside, Apple has decided AI compute is a metered utility, and the era of unlimited tokens — for users and engineers alike — is ending.
Apple declined to comment to Bloomberg. The question now is whether the cuts are really about the token bill, or about establishing a new operating culture under Ternus. Watch the exemptions: if engineering teams start regularly escalating to executives for more tokens, the policy becomes theater. If they quietly adapt — smaller models, more on-device work, fewer prompts — then Ternus has done exactly what he set out to do.
— via Bloomberg