This article may contain affiliate links. Learn more on our Partner Program page.
Apple is tightening restrictions on its internal use of AI. Some engineers will be allowed to spend a third less per day on AI models, and further cuts are under consideration. This comes as research spending has just reached a record high.
According to a Bloomberg report, Apple reduced the daily AI budget for some of its engineers in October. This move comes during the first few months of John Ternus's tenure as CEO, who aims to make Apple faster and leaner.
Bloomberg has good connections regarding internal Apple matters and relies on sources familiar with the process. No document relating to this is publicly available.
Key Facts at a Glance
- Since October, some Apple engineers have been allowed to spend $200 per day on AI instead of $300.
- Some engineering teams are being asked to pay $100, while those outside the engineering field may be paid as little as $25.
- Anyone who needs more can request an exception from management.
- Some teams are expected to switch to cheaper or locally powered models.
- In fiscal year 2026, Apple's research spending had increased by 32 percent.
What is valid and what is up for discussion
The budgets refer to tokens, the data units that AI models process. Programming and debugging consume a particularly large amount of these. According to Bloomberg, Apple employees have access to Claude from Anthropic, ChatGPT from OpenAI, and Gemini from Google.
| Group | So far per day | New each day | Status | Reduction |
|---|---|---|---|---|
| Some engineers | $300 | $200 | since October | a third |
| Some engineering teams | – | down to $100 | in conversation | – |
| Some groups outside of technology | $100 | down to $25 | considered | three quarters |
For heavy users, there is still a way out: Those who need more for their work can request an exception from management. Some teams will also be required to switch to cheaper models or those that run locally on the device.
What the annual limits mean
Over the course of a working year, the sums are considerable. With 220 working days, the old limit of $300 corresponds to a limit of $66,000 per person, while the new limit of $200 corresponds to $44,000. At $100, it would be $22,000, and at $25, $5,500.
This allowance was apparently rarely fully utilized. According to Bloomberg, some employees say that only a few took full advantage of the $300. Therefore, it's unclear how much Apple is actually saving, and the savings are unlikely to become noticeable until further cuts are made.
Apple is not alone, but more generous
Other companies are imposing stricter limits. Tesla has capped its employees' AI spending at $200 per week since July; anything more requires approval. With a five-day workweek, that's $40 a day, or one-fifth of Apple's new limit for engineers.
The situation at Accenture is disputed. According to 404 Media, since October the company has only been contributing $29 per month per employee for Claude; the rest must come from project or team budgets. Accenture, however, denies to Forbes that there is a fixed monthly limit, stating that the figure only represents a portion of the funding.
Saving despite record spending
Apple's costs are rising in the opposite direction. Research spending increased by 32 percent to $45.8 billion in fiscal year 2026, primarily due to higher infrastructure costs, including investments in AI, according to Apple. Total operating expenses rose by 22 percent to $75.5 billion.
According to Bloomberg, the cuts have been met with mixed reactions among employees. Some see them as another sign of cost-cutting measures and fear layoffs, especially since some departments have brought in external consultants to review their processes. In August, Apple had already eliminated over 200 positions at Vision Pro, Siri, and in software teams.
A budget is not yet a withdrawal from AI
I don't see the cut as a cause for alarm. Those who are theoretically allowed to spend $44,000 a year on AI models aren't exactly being squeezed, and it seems only a few have reached the old limit. More interesting are the exceptions made at the executive level and the switch to less expensive models, because these turn the choice of model into a question of cost.
This shouldn't change anything noticeable for customers. Some teams are already reportedly switching to cheaper or locally sourced models, and I expect such requirements to increase if further cuts are implemented.
Is $200 a day a reasonable limit for you, or does Apple risk hindering the very engineers it needs to catch up in AI? Let us know in the comments where you would draw the line.
