According to a Bloomberg report last week, Apple lowered the daily spending cap on AI tools, such as coding assistants and models from outside companies, for some of its engineers from $300 to $200. The first cuts took effect in October, and according to the report, for some teams the cap is expected to fall to $100 a day. Employees who need more can ask management for approval. Apple declined to comment.

To understand the numbers, it helps to translate them. $300 a day is about $6,000 in a 20-workday month, for one employee. Even $200 a day is about $4,000 a month, more than a junior developer's salary in many countries. According to employees who spoke to Bloomberg, few actually hit the old cap, so the first cut saves little. But the direction is clear: the move is part of a broader cost-cutting push by CEO John Ternus.

This story matters because Apple is not a company short of money. If even Apple is starting to ask how much each hour of AI costs and what it gets in return, it is a sign that the phase in which companies let employees use as much as they wanted, 'because it's the future,' is coming to an end. AI vendors, led by OpenAI and Anthropic, sell companies coding tools priced by usage, and agents that write code on their own can burn huge amounts of resources on a single task.

The real question is not whether AI helps engineers, but how much. No large company has yet published convincing data showing that an engineer with a $6,000 monthly AI budget produces much more than one with a $1,000 budget. As long as there is no such measurement, management will treat this spending like any other expense: cut it when the bottom line needs improving.

There is a practical lesson here for Israeli companies. Many startups in Israel adopted AI coding tools enthusiastically, and some even slowed hiring on the grounds that the tools replace developers. If the price of these tools rises, or investors start asking how much they cost, the savings on headcount may turn out to be smaller than they looked. Anyone not measuring today how much they spend on AI and what they get for it will find out in the financial statements.