Apple Cuts Engineer AI Token Budgets From $300 to $200 Per Day

Apple cut the daily AI token budget for some of its engineers from $300 to $200 last month, and Bloomberg reports that deeper reductions are already on the table. The next step could take the cap down to $100 a day for technical staff, and as low as $25 for some non-technical teams. It is a concrete number attached to a shift that is now visible across the industry: companies that spent freely on internal AI access in 2024 and 2025 are starting to treat that access as a managed cost.

The move matters beyond Cupertino. Apple sets procurement patterns that other large enterprises watch, and the cut lands at a moment when Microsoft, Tesla, Uber, and Meta have all recently capped internal AI spending as well, according to Bloomberg's reporting. When five of the world's most valuable technology companies restrain the same line item in the same window, that is a market signal, not an isolated budget adjustment.

What exactly changed at Apple?

The mechanics are straightforward. Apple gives employees access to external AI services, including Anthropic's Claude, OpenAI's ChatGPT, and Google's Gemini, Bloomberg reports. Employees consume those services through a "token budget" — tokens being the units of text that language models process and bill on. Last month, the daily allowance for some engineers dropped by a third, from $300 to $200.

The planned reductions go further. The ceiling may fall to $100 a day for engineers, and some non-technical teams could see budgets of just $25 daily. Employees whose work genuinely requires more compute can request exceptions from management — a process that, by design, adds friction and makes usage an explicit managerial decision rather than an ambient resource.

Some teams will also be shifted to cheaper models or to AI that runs locally on devices, on-device inference meaning the model executes on the hardware itself rather than through a paid cloud API. Local execution carries no per-token cost, which makes it attractive precisely when token budgets tighten. The takeaway: Apple is not pulling back from AI tools for staff, but it is repricing how staff use them.

How do the cuts square with Apple's rising R&D spend?

Here is the tension worth stressing. Bloomberg reports that Apple's research and development spending jumped 32 percent in fiscal year 2026, driven by higher infrastructure and AI costs. The company is simultaneously spending dramatically more on AI overall while rationing the AI credits its own engineers use day to day.

Those two facts are not contradictory, but they are revealing. A 32 percent R&D increase suggests money is flowing to infrastructure, silicon, and large strategic AI commitments rather than to broad internal consumption of third-party models. Every dollar an Apple engineer spends querying Claude or ChatGPT is revenue for Anthropic or OpenAI; every dollar spent on Apple's own infrastructure is not. Rationing third-party access while growing internal spend is consistent with a company trying to keep AI value inside its own platform.

The caveat: Bloomberg's reporting does not break down how the 32 percent is allocated, and Apple has not publicly detailed the budget changes. The signal is directional, not itemized.

What does new CEO John Ternus have to do with it?

The cuts fit the direction set by Apple's new CEO, John Ternus, who has been pushing to make the company leaner, Bloomberg reports. A CEO who campaigns on leanness will find few better demonstration targets than per-employee consumable budgets, which are visible, easily trimmed, and hard to defend as strategic when peers are doing the same.

That framing also carries risk. Budget discipline is easy to announce and hard to calibrate. If the $200 — or eventual $100 — cap starts binding on engineers doing genuine model development or evaluation work, the exception-request process becomes the real budget, and its efficiency depends entirely on how managers adjudicate requests.

In plain words: the policy signals a leaner Apple under Ternus, but its actual cost in lost productivity will not show up in any filing.

How are Apple employees reacting?

Internal reactions are mixed, according to Bloomberg. Some employees saw the old limits as overly generous to begin with — a reminder that $300 a day of premium-model access is, by the standards of most workplaces, an unusual perk. Others worry the cost-cutting could be a prelude to layoffs, reading tightened tool budgets as an early indicator of broader retrenchment.

Both readings are plausible, and neither is confirmed by anything in Bloomberg's reporting. What the two camps share is a recognition that the limits were generous enough, in absolute terms, that cutting them is defensible on paper while remaining ominous in context.

The one-sentence read: employees disagree about whether this is hygiene or a warning, and Apple has said nothing publicly to settle it.

Why is this a wider industry story?

Apple is the latest entrant in a pattern, not an outlier. Microsoft, Tesla, Uber, and Meta have all recently capped internal AI spending as well, Bloomberg reports. That breadth matters for at least three constituencies:

• AI vendors. Anthropic, OpenAI, and Google sell enterprise access on the assumption of expanding consumption. Simultaneous caps at five major buyers suggest the era of unmeasured internal AI spend is ending, and procurement is maturing into something that looks like conventional software budgeting.
• Enterprise buyers everywhere. Apple's specific numbers — $300 to $200, potentially to $100, with $25 tiers for non-technical staff — will function as reference points that other companies' IT departments can cite when setting their own limits.
• Apple watchers. The combination of rationed third-party access, a push toward on-device models, and a 32 percent R&D increase sketches a company consolidating AI investment into its own stack.

The skeptical note deserves emphasis. None of the reporting establishes that the caps reflect dissatisfaction with AI tools' value; cost discipline can be a sign of maturity rather than disillusionment. Equally, none of it establishes that the tools were being wasted. The data simply does not say, and any claim in either direction is speculation.

What is next: Bloomberg indicates more cuts are planned, not yet fully implemented, so the operative numbers — $100 for engineers, $25 for some non-technical teams — remain prospective. How many exception requests Apple grants, and whether the rumored layoff fears materialize, will determine whether this reads in a year as prudent normalization or as the first page of a larger retreat.