Apple cuts daily AI spending allowances as cost-conscious CEO tightens the belt

The numbers behind the cuts

Bloomberg reported on Wednesday that Apple lowered the daily AI spending cap for its engineering teams from $300 to $200 in October, citing sources familiar with the matter.TechCrunchconfirmed the details, noting that Apple is weighing further reductions that would bring some daily allowances as low as $100 for engineering teams. Outside of engineering, non-technical staff could see their caps drop to $25 per day.

The daily allowances cover tokens from third-party AI models — including OpenAI, Anthropic, and Google — that Apple engineers use for code generation, debugging, documentation, and internal tooling. They do not cover Apple's own Apple Intelligence services or first-party AI infrastructure, which remains on a separate budget.

The cuts come as part of a wider directive from CEO John Ternus to streamline operations and reduce costs across the company. Ternus, who took over as Apple CEO at the start of September, has emphasized operational discipline as a hallmark of his early tenure.

Why Apple is tightening the belt

Earlier this year, a wave of'tokenmaxxing'swept through the tech industry, with companies pushing employees to use AI tools as aggressively as possible. The philosophy — use AI everywhere, measure productivity later — was driven by a fear of falling behind in the AI race. But the bills for those tokens have come due, and companies are discovering that open-ended AI spending does not always translate into measurable productivity gains.

Apple is not alone in this realization. Across the industry, companies are starting to impose tighter controls on AI usage as the gap between the cost of running models and the revenue they generate becomes harder to ignore. TheUS tech giants' combined hidden AI debt has swelled to an estimated $3.2 trillionaccording to recent analyses, as companies continue to spend heavily on infrastructure and inference without clear paths to profitability.

For Apple specifically, the budget cuts represent a departure from the permissive approach that characterized the early phases of the AI boom. The company has historically been cautious about AI, with Ternus himself saying in late October that Apple doesn't 'want to ship AI for AI's sake.' The spending caps bring the company's internal tools in line with that public philosophy.

What the cuts signal for the AI industry

Apple tightening its AI budget is significant because it signals that even the most cash-rich companies are starting to ask hard questions about AI return on investment. If Apple — withover $150 billion in cash reserves— is rationing tokens, it suggests that the current model of unlimited AI consumption is not sustainable for anyone.

The cuts may also accelerate a shift toward more efficient AI models and deployment strategies. If companies cannot spend unlimited amounts on premium API tokens, they may turn to cheaper open-weight models, on-device inference, or specialized fine-tuned models that require fewer tokens per task.

For Apple's engineering teams, the impact is likely to be felt most acutely by groups that have integrated AI tools into their daily workflow. A senior engineer who previously had a$300 daily budgetfor code generation, documentation, and debugging now has $200 — and may soon have only $100. For non-engineering teams, the available budget could drop to a daily amount that barely covers a few API calls.

The move also underscores a paradox at the heart of the current AI moment: the same companies racing to build AI products and services are also struggling to manage their own internal AI costs. While Apple continues to invest in Apple Intelligence and AI-powered features for consumers, its own engineers are facing tighter constraints on the tools they can use to build those features.

A pattern across big tech

Apple is the latest but far from the first company to impose limits on internal AI spending. The pattern has been building for months as companies realize that the open checkbook approach to AI tokens is not sustainable in a world where AI model providers charge per token and enterprise usage can scale exponentially.

Anthropic's recent S-1 filing revealed the company spends$1.7 billion per quarter on payroll, or roughly $107,000 per employee annually. Across the industry, the cost of inference continues to be a major concern, with companies spending millions on API tokens that often go toward experimental or low-value use cases.

The AI cost question has become a boardroom issue. In October, a wave of companies began reviewing their AI spend more closely, with some implementing tiered access — giving generous budgets only to teams that could demonstrate measurable productivity gains. Others moved entirely to self-hosted open-weight models to escape per-token pricing entirely.

The question Apple's budget cuts raise is whether the current AI productivity boom was real or partly inflated by permissive spending policies. If engineers were generating more code because they had expensive AI tools at their disposal, will productivity drop when those tools get capped? Companies that have already moved to cheaper orself-hosted modelsmay find themselves at an advantage in the coming quarters.

The implications extend beyond just the cost of tokens. If Apple's engineering teams learn to be more efficient with AI, the knowledge and habits may carry over to how the company builds AI-powered products for consumers. A team that has internalized the true cost of AI inference is likely to build more efficient, more cost-effective products — a competitive advantage that may not show up on any balance sheet but could shape Apple's AI trajectory for years.

The Ternus factor

John Ternus's approach to AI has been characterized by a deliberate, product-first philosophy. In hisWired interview reported by TechCrunch on October 31, he said: 'We don't want to ship AI for AI's sake. When we're doing it right, people won't have to think about whether it was AI that made this happen or more traditional software.'

The budget cuts extend that philosophy from product strategy to internal operations. If Apple does not want to ship AI for AI's sake, it also does not want to burn cash on AI for AI's sake internally. Ternus appears to be applying the same cost-benefit lens to internal tooling that he applies to customer-facing products.

It is a pragmatic approach that other tech CEOs may be forced to emulate as AI spending continues to outpace the revenue it generates. The question is whether $200 per engineer per day — or even $100 — is enough to maintain the AI-assisted productivity gains that companies have come to depend on.

Looking ahead, the industry is likely to see more of these budget adjustments as the gap between AI infrastructure spending and revenue persists. For Apple, the caps may also influence how its engineering teams approach problem-solving — forcing them to be more deliberate about when to use AI tools versus traditional methods, which could itself be a productivity-boosting discipline.