TechCrunch, citing Bloomberg, reported on 4 November 2026 that Apple cut the daily amount some engineering teams can spend on AI from USD 300 to USD 200 in October. The change makes token consumption an explicit cost control for staff using external models in development work.
Some allowances may fall further. Engineering teams could see caps of USD 100 a day, while some groups outside engineering could be limited to USD 25 a day. Many non-engineering groups already work with a USD 100 daily limit.
The cap covers spending on external models including Anthropic's Claude, OpenAI's ChatGPT and Google's Gemini. Engineers have higher allowances than other teams because coding and debugging can consume large volumes of tokens. Staff who need more can ask executives for an exemption.
Apple is also asking some teams to use less expensive models or switch to systems that run on the device. That changes what IT and development leaders need to measure. They need to know which model a tool uses, where it runs and what each task costs.
Some Apple employees said few people came close to using the full USD 300 allowance, so the savings may be modest. Even so, a formal cap creates a control point for finance, procurement and engineering management. It also gives teams a reason to separate high-value coding and analysis work from routine prompts that can run on cheaper systems.
The move follows a period in which companies encouraged more internal AI use on the expectation of productivity gains. The cost of tokens makes that approach harder to leave unmanaged. If a model is metered by use, adoption is also a variable cloud bill.
The immediate issue for technology leaders is governance rather than access. They need allocation rules, exception processes and telemetry that relate model use to actual work. Without that, daily caps can reduce spending but do not show whether the remaining spend is useful.
Daily AI allowances, token telemetry and decisions between external models and on-device systems are practical questions for ECS's Core AI topic, alongside agentic AI, Azure AI Foundry, the EU AI Act and responsible AI. Sessions and workshops give attendees room to compare cost controls, model selection and governance patterns from real deployments. Attendees working through AI budgets can bring the policy and measurement questions that are hard to settle from vendor documentation alone.
Photo: Nils Huenerfuerst, CC0, via Wikimedia Commons, cropped
This article was written by the ECS editorial team with the assistance of Amal Agent, run.events' AI assistant, which supported the research, drafting and proofreading.