The Apple internal AI budget is shrinking. In October 2026, Apple lowered the daily amount some of its engineering teams can spend on AI tools from $300 to $200, according to Bloomberg, as reported by TechCrunch — and more cuts are reportedly on the way. The move is a sharp reversal from earlier in the year, when the same company was handing teams generous daily token allowances and urging them to use AI as much as possible.
What changed is not Apple’s public AI offering. The caps apply to what Apple’s own employees can spend on AI tools in their daily work. But the reversal matters beyond Apple: Apple joins other companies that have introduced internal AI spending controls this year, and the bill for the tokenmaxxing frenzy has now come due.
What Apple changed: the Apple internal AI budget in numbers
The numbers, as reported by Bloomberg and summarised by TechCrunch, are straightforward. In October, Apple cut the daily AI spending cap for some engineering teams from $300 to $200 — a one-third reduction. Further cuts are under discussion: some engineering teams could see their allowance fall to $100 a day.
Outside engineering, the limits are already lower and may go lower still. Many non-engineering employees already work under a $100 daily cap, and Apple has considered lowering that to $25 a day for some teams. The reporting is attributed to people familiar with the matter, and Apple declined to comment on the reported cuts.
Two details in the reporting are worth keeping in mind. First, the $100 engineering figure and the $25 non-engineering figure are options under discussion, not confirmed policy. Second, employees who need higher limits for their work can request an exemption from senior leadership — the caps are a default, not an absolute ceiling.
How the daily AI allowances work
The caps are spending limits on tokens: the units of text and data that AI models process on every request. When an Apple engineer asks an AI coding assistant to draft, review or debug code, the model consumes tokens, and the provider bills for them. A daily dollar cap limits that metered spending.
The allowances cover internal use of outside AI models, including Anthropic’s Claude, OpenAI’s ChatGPT and Google’s Gemini. Apple has also asked some teams to use cheaper models, or to run AI models on-device, instead of the priciest cloud options — the same cost pressure, applied as a choice of model rather than a hard limit.
The caps therefore work as a spending control, not a ban: they limit how much can be spent in a day, while exemptions remain available for work that needs a higher allowance.
From tokenmaxxing to token rationing
The striking thing about the cuts is how recently Apple was doing the opposite. Just months ago, internal teams were handed daily budgets of up to $300 worth of Claude tokens and told to use them. Teams that significantly underused the allowance reportedly saw requests to refill vacant positions denied more often. Usage was not just encouraged; it was effectively enforced.
TechCrunch’s account of the wider shift captures the whiplash: “Earlier this year, a tsunami of tokenmaxxing took hold, with companies pushing staff to use AI as much as possible in the name of vaguely defined efficiency gains. The frenzy was short-lived, however, as bills for said tokens came due, and companies decided that token rationing might be a more fiscally sensible approach.”
There is an irony buried in the reporting: Bloomberg also reported that few Apple employees were actually hitting their full $300 daily limit in the first place. The research does not quantify how many employees a $200 cap will constrain.
The cost-versus-productivity question
Behind the caps sit persistent concerns about what all this AI spending actually returns, and whether the productivity gains match the cost. TechCrunch put it plainly: “Concerns persist about an apparent gap between what AI costs to build and run and the (often unclear) productivity gains the technology is supposed to generate.”
The numbers behind that concern are stark. Gartner forecasts total AI spending in 2026 will reach $2.59 trillion, up 47% on the previous year. Yet MIT’s Media Lab found that despite more than $30 billion in enterprise AI investment, 95% of pilots showed zero measurable financial return. Spending is up nearly half, while the measured return on those pilots is the open question.
The cost question is surfacing in public discussion as well: the word “tokens” appeared in 129 earnings calls in the second quarter of 2026, up from 57 the quarter before, according to AlphaSense analysis.
The contrast with the optimism of a few months ago is sharp. In February 2026, Meta’s chief technology officer Andrew Bosworth said a top engineer spending an amount equal to their salary on tokens saw a tenfold productivity increase, calling it “a no-brainer” with “no upper limit”. Meta has since killed its internal Claudeonomics usage leaderboard — which had tracked token use across more than 85,000 employees, burning through more than 60 trillion tokens in a single 30-day stretch — and moved to cap and centrally monitor internal AI spending instead.
Apple’s own accounts show costs rising. The company’s research and development expenses rose 32% in fiscal 2026, mainly because of higher infrastructure and AI-related costs, and total operating expenses climbed 22% to $75.5 billion. Those figures show Apple’s costs rising; they do not establish Apple’s precise reason for the caps.
What this means for Apple’s AI plans
On the available evidence, the cuts affect internal usage budgets, not Apple’s products or its public AI features. Nothing in the reporting suggests Apple is retreating from AI itself. Its annual report, filed on November 3, 2026, warns that AI agents can act on their own, including reaching systems, running or changing code, sending messages and completing transactions.
Whether tighter budgets slow Apple’s internal AI integration is an open question the reporting does not settle. What the reporting does document is a shift from encouraging usage to setting limits: employees can request exemptions from senior leadership if their work needs a higher allowance.
The cuts also follow a broader cost-discipline push from John Ternus, who became Apple’s chief executive in September 2026. The AI caps are one part of that push, not a standalone judgment on AI tools. And with Apple declining to comment, the company’s own reasoning remains unconfirmed — the reported cap changes rest on Bloomberg’s sourcing.
Is this a wider pattern?
Yes — Apple is joining a pattern, not starting one. Tesla told employees in July 2026 that their AI spending would be capped at $200 per week, after engineers were consuming thousands of dollars’ worth of tokens each week, according to The Information. Uber burned through its entire 2026 AI budget in four months, mostly on flagship coding models, and imposed a $1,500 monthly cap per employee per tool. Meta’s internal AI costs are approaching billions of dollars for 2026, and it has moved to cap and centrally monitor staff spending on outside AI tools.
A Financial Times investigation in June 2026 found that Amazon, Walmart, Cisco, Uber and Meta had all introduced spending caps, discouraged wasteful use, or pushed workers toward cheaper models. AT&T has started limiting some employees’ access to Microsoft’s GitHub Copilot. Amazon scrapped an internal leaderboard that ranked workers by AI usage after staff gamed it, and Shopify shut down its own token usage leaderboard in June.
Some companies are steering staff toward cheaper or internal models where suitable. AT&T’s chief AI officer, Andy Markus, has said firms can save as much as 90 percent by opting for less advanced models, noting that “for most use cases, the latest greatest frontier model isn’t needed.” Salesforce, meanwhile, says it now tracks “agentic work units” instead of tokens — a metric meant to measure output rather than raw usage.
That is the honest shape of the story, as far as the reporting goes. Apple cut a daily allowance from $300 to $200, with deeper cuts under discussion, at a time when token bills were rising and returns stayed hard to measure — though Apple has not confirmed its reasoning. It is not alone, and the reporting concerns internal usage budgets: it does not establish a retreat from Apple’s public AI products. But the question companies are now asking is not how much AI their people can use, but what each token actually buys.
