Apple Reportedly Cuts Some Engineers' Daily AI Budgets to $200

What happened

Bloomberg reported on Wednesday 4 November 2026 that Apple cut the daily AI spending cap for some of its engineers from $300 to $200 in October. The report, which cites people with knowledge of the change, says the company is planning further cuts for some engineering teams that could bring allowances down to $100 a day. The cuts were made quietly, according to the same report, and Apple declined to comment.

The cap applies to the AI models employees use, and the budgets cover tokens, the units of text that AI models process and charge for. TheNextWeb reports that employees who need more can ask senior leaders for an exemption. The reports do not say how many engineers or teams are affected, and they do not state how much money the cuts would save.

The reports tie the moves to the leadership of John Ternus, who became chief executive on 1 September 2026 and is pushing for a faster, leaner company. Some staff reportedly say few people used their full $300 allowance, while others read the cuts as a sign of wider cost cutting and worry about layoffs. TechCrunch's coverage of the Bloomberg report describes the change as a big shift for a company that has worked to bring AI into its engineering processes.

Why it matters for businesses that automate work

The story sits inside a broader correction in how companies pay for AI. TechCrunch notes that earlier in 2026 many companies pushed staff to use AI as much as possible in the name of efficiency gains. Once the token bills arrived, several moved toward rationing usage. The same report says concerns persist about the gap between what AI costs to build and run and the productivity gains it is meant to produce.

Apple's own figures show the scale of the spending pressure. TheNextWeb reports that Apple's research and development costs rose 32% in its 2026 fiscal year, driven by higher infrastructure and AI costs, according to the company's annual report published that week. That is a company-level figure and does not describe the cost of any single workflow.

For a business that automates routine work, the practical lesson is that AI cost grows with volume. A workflow that reads incoming invoices, drafts follow-up emails or checks reconciliations all day consumes usage every time it runs. Usage that is cheap in a test can become a large monthly line once it covers every customer, supplier or branch.

What changes in practice for a business that automates work

Most major AI providers now let an account holder set a spending limit, and the design of those limits matters. OpenAI's documentation describes monthly spend limits at the organisation and project levels. A spend alert sends a notification while traffic continues. A hard limit makes affected requests fail once the tracked spend reaches the amount set. Because the hard limit stops requests, a finance workflow could halt in the middle of a month-end run if the limit is set too low.

Anthropic's Claude Console works in a similar way with workspaces. An administrator can set a monthly spend limit on each workspace, but not on the default workspace, and organisation-wide limits still apply even if workspace limits add up to more. Giving each automated process its own workspace or project means one faulty process cannot use up the budget for the whole company.

A practical sequence for an owner or operations head is short:

1. List every automated workflow that calls an AI model, and name the person who owns each one.
2. Give each workflow its own project or workspace, with its own monthly limit and an alert set well below that limit.
3. Decide in advance which workflows may pause when a limit is reached, and which must keep running, such as payment reminders due on a fixed date.
4. Review the usage figures each month and compare them with the work the workflow completed.

Automation does not remove the need for checks. A cap controls spend. It does not check whether an invoice match or a drafted message is correct. A person still needs to review exceptions, approve payments and sign off on anything that affects customers, staff pay or tax records. Automation is also not worth its running cost where a task happens only a few times a month and takes a few minutes each time. In those cases, a simple template or a shared checklist may cost less and be easier to control.

What to watch next

• Whether Apple confirms the $200 cap and any move to $100 or $25. As of the reports, Apple has not commented.
• Whether the reported $25 limit for teams outside engineering is put in place.
• Whether the layoff concerns inside Apple lead to a formal announcement. The reports describe worry among staff, not confirmed job cuts tied to these budgets.
• Whether other large employers publish similar internal AI budgets, and how those budgets are measured against output.

For business owners, the useful question is not whether Apple's figures apply to them. It is whether they know what their automated work costs each month, which team owns each workflow, and what happens when a limit is reached.