Apple quietly put some engineers on a $200 daily AI budget

What Apple actually changed

Apple has quietly lowered the daily AI spending limit for some of its engineers from $300 to $200, according to Bloomberg, and the company is preparing further cuts for several engineering teams. The change took effect last month, and it applies to the internal budgets employees use when they run prompts through third-party models.

The allowances are measured in tokens, the units of text and data that a model processes when it handles a request. Coding and debugging sessions can consume large volumes, especially with stronger models, so a daily cap is a practical way for a company to slow spending without banning the tools outright.

Why token bills became a problem

Over the past year, tokenmaxxing, the practice of consuming as many AI tokens as possible as a proxy for productivity, spread through Silicon Valley. Some companies introduced internal leaderboards, occasionally unofficial ones, that ranked employees by how many tokens they burned, turning a cost line into a scoreboard.

That enthusiasm created a budgeting problem. Inference is metered, and heavy agentic workloads multiply the number of calls behind a single task. Companies that once celebrated usage began asking whether the productivity gain justified the bill, and finance teams started treating per-seat AI costs the way they treat software licences. The LLMjacking cases this site covered showed how quickly a stolen or careless credential can turn into a six-figure cloud bill.

The same arithmetic applies to every company running agents at scale. A single looped workflow can issue thousands of model calls in an hour, and each call is billed by the token. When usage is measured per employee rather than per product, the fastest way to cut the bill is to cap the person rather than the workload.

The cuts beyond engineering

Engineering is not the only place where limits are tightening. Many groups outside engineering already work under a $100 daily cap, and Apple has considered reducing that ceiling to as little as $25 for some teams, according to people familiar with the planning.

Employees who need a larger budget can request an exemption from senior leadership, Bloomberg reported. That structure keeps the tools available for the work that needs them while making the default cheaper, which is the same pattern large employers use for travel, hardware and contractor seats.

For Apple, the awkward part is timing. The company spent a year telling staff that AI fluency was part of the job, and now the same staff are being asked to justify a daily allowance. That is a management problem as much as a finance one, because it changes what counts as normal use.

Which tools employees can use

The reports do not identify which vendors are affected by the new ceilings, but Bloomberg said Apple staff have access to several assistants, including Anthropic's Claude, OpenAI's ChatGPT and Google's Gemini. Some teams are also being encouraged to try cheaper models or lighter configurations for routine work.

Apple has leaned on Anthropic's models internally, according to earlier reporting by Mark Gurman, even while it partners with Google on Siri. That mix of vendors makes the internal budget a small test of how a large company allocates work across competing model providers without committing to one. Enterprise buyers have been weighing the same question since Anthropic's revenue surge made clear how much work is already running through a single vendor.

Vendor choice now carries a budget consequence. If a team shifts routine coding to a cheaper model and reserves a frontier model for hard problems, the daily allowance stretches further, which is exactly the behaviour the new caps are meant to encourage inside Apple.

How staff reacted to the caps

Reactions inside the company have been mixed. Few workers were using their full $300 allowance, according to Bloomberg's sources, which makes the old ceiling look generous in hindsight. Other employees see the reductions as part of broader cost cutting and worry that layoffs could follow.

Some divisions have brought in outside consultants to review efficiency and operations, adding to that anxiety. Apple declined to comment on the report. The silence is normal for the company, and it leaves the internal reaction documented mainly through unnamed people.

The cost-cutting context at Apple

The limits arrive under chief executive John Ternus, who has pushed for faster product launches and fewer management layers. They also land days after Apple closed its fiscal year with research and development spending up 32% to $45.8 billion, driven partly by infrastructure and AI costs. Power and cooling are part of that line item, an issue this site traced in the move to on-site jet engines.

Read against that number, a $200 daily cap on engineers is small. Its significance is symbolic: the company that told staff to use AI aggressively is now measuring the bill, and it is treating model access as a managed expense rather than an open-ended benefit handed to everyone.

For employees, the practical effect is a return of procurement logic. Anyone who wants a larger allowance has to explain why the work needs it, and that explanation is now part of the job in a way it was not during the year of unlimited experimentation.

Apple is not the only company doing this arithmetic. Microsoft set division-level AI budgets in July, Tesla caps staff at $200 a week, and Uber limits each AI coding tool to $1,500 a month, according to The Next Web. The caps differ in size, but the logic is identical: usage is now a line item someone has to approve.

What is established and what is not

Established: Apple lowered some engineers' daily AI spending limits from $300 to $200 in October, further reductions to roughly $100 are planned for some teams, and non-engineering groups may fall from $100 to $25. Staff can seek exemptions, and Apple declined to comment.

Not established: how much money the caps will save, which vendors lose the most usage, or whether the reductions will spread to every team. The open question is whether this is a temporary correction after a year of heavy experimentation or the start of permanent rationing inside large companies.