· AI
Apple Cut Engineers' Daily AI Budget by a Third. Its Own AI Spending Still Went Up.
Apple capped what an engineer can spend on AI in a day. It did not cap what Apple spends on AI overall, and that gap is the actual lesson here.
Apple cut the daily AI token budget for some of its engineers from $300 to $200, a cut of a third, and is weighing further cuts to $100 for engineers and as low as $25 for some non-engineering teams, according to Macworld's and The Next Web's reporting on an original Bloomberg scoop. The same fiscal year, Apple's R&D spending rose 32 percent, driven largely by AI infrastructure, and operating expenses climbed 22 percent to $75.5 billion. A company spending tens of billions more on AI is simultaneously telling individual employees to spend less of it. That is not a contradiction. It is the actual discipline most businesses skip.
My position: the dollar figure on the cap is close to irrelevant. What matters is that Apple put a hard number in front of every employee's AI use at all, because a number is the only thing that forces someone to ask whether a cheaper tool would do the job just as well.
Here is the fair case against this move. Per Bloomberg's reporting, few engineers were hitting the old $300 ceiling in the first place, which means cutting it to $200 may save Apple close to nothing in real dollars. The timing makes it worse optically: Apple cut more than 200 jobs in August, some divisions have brought in outside efficiency consultants, and staff told reporters they read the AI cap as one more sign of belt-tightening under new CEO John Ternus. If the old limit was never binding, a tighter one that is still not binding is a morale cost with no financial upside. That is a real criticism, not a strawman, and I do not think Apple's own reporting rebuts it directly. Ternus has said publicly that he does not want to "ship AI for AI's sake," which is a mission statement, not a cost defense.
But the criticism misses what the cap is actually for. The Next Web's reporting notes that some Apple teams are being told to switch to cheaper models or on-device models once they approach their limit. That is the mechanism, and it has nothing to do with whether $200 a day is generous. A budget line with a ceiling is the only thing that makes an engineer stop and ask which task in front of them actually needs a frontier model's reasoning and which one a smaller, cheaper, or local model handles just as well. Remove the ceiling and that question never gets asked, because there is no moment that forces it.
Apple is not alone in reaching for this tool, and the variation across companies is the interesting part. Microsoft set AI budgets at the division level in July. Tesla caps staff AI spending at $200 a week, a tenth of Apple's old daily rate. Uber limits each AI coding tool to $1,500 a month per use case. Accenture caps Claude access at $29 a month per seat. None of these numbers are comparable to each other, because the companies, the use cases, and the scale are not comparable. What is comparable is that every one of them decided a flat, unmonitored AI subscription was the wrong way to buy this, and replaced it with a metered budget tied to a person, a team, or a tool.
That is the part worth taking from this story even if you run a five-person shop with no engineer anywhere near $75.5 billion in operating expenses. Apple can absorb basically any AI bill it wants and caps spending anyway, because the cap's job is to generate information, not savings. Most small businesses have the opposite problem: real AI spend that nobody has ever metered by person or by task, which means nobody has ever been forced to ask which of those calls could run on a four-dollar-a-month local model instead of the most expensive frontier one available. Set a number, even a generous one, and watch where your own team bumps into it. That is where your real routing decision is sitting, and it is the specific question we get hired to answer at MojoAI: not whether to use AI, but which model earns its price on which task, and where a cheaper one quietly does the same job. If you have never run that exercise on your own AI bill, that is worth an hour before your next renewal, not after it.
Sources
References used in this article. Links also appear alongside the relevant claims.
Let's make it happen.
You don't need a polished brief. A couple of lines about where you want your company to go is plenty, and we'll come back with what we'd tackle first.
We scope the work around your goals and timeline before quoting anything, so you know exactly what you're getting.