Apple cuts employee AI spending limits to manage rising costs

The shift toward cost control

Apple has begun a process of tightening the purse strings on how its employees use artificial intelligence. According to reports from Bloomberg, the company is lowering the daily spending limits for AI tokens, which are the basic units of text or data processed by these models. This move is part of a broader effort by CEO John Ternus to manage the company's escalating expenses.

The adjustments come at a time when the cost of maintaining AI infrastructure has become a significant line item for tech giants. While Apple previously encouraged staff to integrate AI into their workflows, the sheer cost of doing so at scale has prompted a pivot toward fiscal discipline. The company is now looking to ensure that the productivity gains from these tools justify the high price tag of the underlying technology.

This trend is not unique to Apple, as other major firms are also reining in their AI spending. For example, Accenture recently reduced its centrally funded allowance for the Claude model to just $29 per employee each month. These moves suggest that the initial phase of unrestricted AI adoption is evolving into a more managed and cost-conscious corporate strategy.

How the limits are changing

The specific changes to spending caps vary depending on the department and the seniority of the staff. Some engineers who previously had a daily allowance of $300 have already seen that limit reduced to $200. Furthermore, the company is planning additional cuts for several engineering teams, which could see daily allowances drop to as little as $100.

For employees outside of the engineering departments, the restrictions may be even more significant. Many groups currently operate under a $100 daily cap, but Apple is considering lowering that limit to $25 for certain teams. These limits apply to various popular AI services that employees can access, including Google’s Gemini, OpenAI’s ChatGPT, and Anthropic’s Claude.

To manage these new restrictions, Apple is implementing a system of exemptions for specific roles. Employees who can demonstrate that they require higher spending limits to perform their essential job functions will be able to request permission from senior leadership. This allows the company to maintain strict oversight while still providing the necessary resources for high-demand tasks.

The rise of tokenmaxxing

The need for these limits stems from a cultural trend in Silicon Valley known as 'tokenmaxxing.' This practice involves employees consuming as many AI tokens as possible, often viewing high usage as a primary metric for their own productivity. Some companies even went as far as creating internal leaderboards to rank employees based on how many tokens they were able to process.

This behavior accelerated as AI agents became significantly better at complex tasks like software coding. Tools such as Cursor, OpenAI Codex, and Claude Code became standard for developers, leading to a surge in demand for high-volume processing. As the costs of these tools rose, companies began to question whether the massive consumption of tokens was actually resulting in meaningful work or just high overhead.

Apple's move to curb this usage reflects a broader corporate realization about the economics of large language models. While the technology offers immense potential, the infrastructure required to power it is expensive. By setting caps, Apple aims to prevent 'tokenmaxxing' from becoming an uncontrolled expense that drains the company's resources without a clear return on investment.

Impact on Apple's operations

The move to cut costs is particularly notable because Apple's overall spending is actually increasing. The company reported that its R&D expenses rose 32% to $45.8 billion in its most recent fiscal year. This spike was driven largely by the high costs associated with AI infrastructure and the development of new technologies.

Despite the spending cuts for employees, Apple remains heavily integrated with third-party AI providers. Bloomberg reports that the company is essentially 'running on Anthropic' for many of its internal operations. This reliance on external models like Claude makes it even more critical for Apple to manage the costs of every interaction its staff has with those systems.

Employee reactions to these new limits have been divided across the company. Some workers noted that few people were actually using their full $300 daily allowances, suggesting the cuts might not affect everyone. However, others view the move as a warning sign of broader cost-cutting measures, leading to concerns that the company might eventually resort to layoffs to balance the books.

Moving toward local AI

To further reduce costs, Apple is looking for ways to move away from expensive cloud-based models whenever possible. Some teams are now being encouraged to utilize cheaper AI models or to run models locally on their own devices. By processing data on-device, the company can bypass the per-token costs associated with large-scale cloud providers.

This strategy aligns with the broader industry trend of optimizing for efficiency. If Apple can successfully transition more of its internal workflows to local models, it could significantly lower its operational overhead. This would allow the company to continue its AI initiatives without the same level of financial pressure from infrastructure costs.

The long-term goal for Apple appears to be a balanced ecosystem where AI is a standard tool for every employee. By implementing these spending caps now, the company is attempting to create a sustainable model for AI usage. The success of this plan will depend on whether they can maintain high productivity levels while significantly reducing the cost of every prompt and response.