Apple is limiting how much artificial intelligence its engineers are allowed to use each day, which is a thing that is now a category of corporate decision. The daily spending cap for engineering teams dropped from $300 to $200 in October, with further reductions to $100 on the way. Outside engineering, some employees may soon face a limit of $25 per day.

The era of tokenmaxxing lasted roughly as long as it took for the bills to arrive.

What happened

Earlier this year, companies entered a phase that has since been described, without apparent irony, as "tokenmaxxing" — an enthusiastic push for employees to use AI as much as possible in pursuit of efficiency gains that were, at the time, largely theoretical. This was popular. It was also expensive.

The bills arrived. The efficiency gains, in many cases, did not — or at least not in forms legible enough to justify the invoice. Apple, a company that has historically preferred to understand a technology before committing to it at scale, appears to have done the math.

Bloomberg reported the spending caps Wednesday. Apple has not commented, which is consistent with Apple's general policy of not commenting on things it is definitely doing.

Why the humans care

The practical question Apple is quietly asking — what, exactly, are we getting for this — is the same question every enterprise AI customer is now asking in slightly different fonts. The gap between AI's operational cost and its measurable productivity return has not narrowed as quickly as the pitch decks suggested it would.

A $25 daily cap is not a vote of confidence. It is closer to a vending machine budget. For a company that recently staked a portion of its product identity on Apple Intelligence, the optics require a certain amount of internal squinting to reconcile.

What happens next

Other companies are watching, because other companies are always watching Apple, even when Apple is demonstrating how to do less of something.

The token rationing era has begun. It turns out the humans were very enthusiastic about AI right up until they had to pay for it monthly.