Apple cuts engineers' AI spending cap to $200 a day

Felix Laurent 5 Nov 2026 4 min read

Big tech

Apple has cut how much some of its engineers can spend on AI each day. In October, the daily cap for some engineers fell from $300 to $200, Bloomberg reported on November 4, 2026. More cuts are planned, and some teams could drop to $100 a day.

Apple is the latest big company to put AI use on a budget. Tesla, Uber and Accenture set their own limits earlier this year. Together they show what happens when a company pays for every token its staff burn.

What Apple changed

The Bloomberg report, relayed by MacRumors and TechCrunch, describes three levels of cuts:

some engineers went from $300 to $200 a day in October;
some engineering allowances could fall further, to $100 a day;
many groups outside engineering already have a $100 daily limit, and some could drop to $25.

Employees who need more can ask executives for an increase. Apple is also asking some teams to use cheaper models or to switch to AI that runs on the device. Apple employees can use Claude, ChatGPT, Gemini and other models, according to MacRumors.

The savings may be small. Some employees told Bloomberg that few people were using their full $300 a day. Apple has not commented on the report.

The cuts arrive at a sensitive time inside the company. John Ternus became CEO on September 1 and reportedly wants to reduce layers and bloat. Apple has also brought in outside productivity consultants for some divisions, which has fed worries about a restructuring.

Tesla set a weekly AI limit in July

Rohan Paul, who shared the Apple news on X, compared it with Tesla's rule. Tesla capped employee spending on third-party AI tools at $200 a week starting July 6. TechTimes reported it, citing a memo first reported by The Information. Spending above that needs a manager's approval.

The cap covers models from Anthropic, OpenAI and Google. Beta products from xAI, including Grok, do not count against it. Tesla has invested $2 billion in xAI. Engineers largely prefer Anthropic's Claude over Grok, four people told Electrek.

Tesla's reason was cost. Some software engineers were using thousands of dollars of tokens every week. Internal dashboards had ranked employees by token use to encourage more of it.

That ranking idea came up at other companies too, and it ended the same way.

Uber and Accenture hit the same wall

Uber burned through its full-year AI budget in the first four months of 2026. In June it confirmed a new cap to Bloomberg. Each employee can spend up to $1,500 a month on tokens for each agentic coding tool, such as Cursor and Claude Code. Spending on each tool is tracked separately on an internal dashboard, Outlook Business reported.

Accenture went much lower. From October 1, it pays only $29 per person per month toward Claude, 24/7 Wall St. reported, citing an internal memo. Anything above that must be charged to a client project. Some consultants had been spending $500 to $1,000 a month on Claude Code.

Heavy use is still possible at Accenture, but a project now pays for it. "Our approach today is to provide the right tool for the right work," a spokesperson said.

The caps differ a lot in size. The table converts each one to an approximate month, assuming 22 working days and about 4.3 weeks.

Company	Reported cap	About per month	Since
Apple, some engineers	$200 a day, down from $300	~$4,400	October 2026
Apple, some non-engineering teams	$25 a day (planned)	~$550	not yet in force
Uber	$1,500 a month per coding tool	$1,500 per tool	June 2026
Tesla	$200 a week	~$860	July 6, 2026
Accenture	$29 a month, the rest billed to projects	$29	October 1, 2026

Even after the cut, Apple's engineers keep one of the largest allowances on the list. The question is why companies are cutting at all, when AI keeps getting cheaper.

Why AI bills grow while tokens get cheaper

Prices per token have fallen fast. As we reported earlier, a16z estimates that token prices fell 67% since May while token spend rose more than 120%. Cheaper tokens did not shrink the bill. People simply used far more of them.

Coding agents drive most of that growth. A tool like Claude Code reads files, runs commands and retries on its own, so one task can burn far more tokens than a chat. Dashboards that reward heavy use push the numbers higher still.

So the budgets companies set for AI in 2025 no longer match how their staff work. Each company is now picking its own fix: a hard cap, manager approval, cheaper models, or billing the cost to clients.

Will AI spending caps slow down AI adoption at work

For most staff, probably not. Apple's own employees said few of them used the old $300 limit. Accenture still lets heavy users keep working if a project pays. The caps mostly hit the small group of power users who ran up thousands of dollars a week.

The bigger shift is in how companies see AI. A year ago, many treated tokens as free and ranked staff by how much they used. Apple, Tesla, Uber and Accenture now treat AI like any other expense, with a budget and an owner. That is a sign the tools have become a normal part of work, and a cost that finance teams now track line by line.