Apple’s New CEO Moves to Rein In Internal AI Spending and Token ‘Maxxing’

Apple is reportedly tightening the reins on how its employees use internal generative AI tools, as new CEO John Ternus looks to curb spiraling compute costs. Inside the company, some developers had been “token‑maxxing” — deliberately running more AI workloads than necessary because internal metrics rewarded higher token usage. In large language models, tokens are the basic unit of text processing, and heavy use translates directly into higher cloud and data‑center bills.

Under the new leadership, Apple is said to be revising internal AI credits and usage quotas so that experimentation remains possible, but idle or vanity use is discouraged. The goal is to steer expensive compute resources toward projects that have a clear product or efficiency payoff, rather than treating AI usage as an end in itself. At the same time, Apple’s overall research and development budget continues to swell, with R&D spending up 32% year‑on‑year in fiscal 2026 and total operating expenses up 22%, signaling that the company is far from pulling back on AI.

The shift reflects a broader challenge facing big tech: how to manage the mounting infrastructure costs of the AI race. Apple’s approach suggests a next phase in corporate AI strategy, where internal incentives and governance are re‑engineered to favor efficient, targeted deployment of AI rather than raw consumption of tokens.