Editorial illustration for Apple slashes internal AI budgets, shifting teams to cheaper models
Apple slashes internal AI budgets, shifting teams to...
Apple cut the daily token budget for some engineers from $300 to $200 last month, Bloomberg reports, and the company isn't done trimming. Internal plans call for that number to drop as low as $100 a day, with some non-technical teams facing caps as tight as $25. Workers who burn through their allowance faster than that can ask managers for an exception, but the default is now smaller than it was a month ago.
The squeeze comes even as Apple's engineers keep access to Anthropic's Claude, OpenAI's ChatGPT, and Google's Gemini. Bloomberg says some teams will get pushed toward cheaper models instead, or onto AI that runs locally on the device rather than through paid API calls. The timing lines up with John Ternus, Apple's new CEO, who has been pressing the company to run leaner.
The budget cuts sit oddly next to Apple's books: R&D spending rose 32 percent in fiscal year 2026, pushed up by infrastructure and AI costs. Reactions inside the company, per Bloomberg, split between employees who thought the old limits were already too generous and those reading the cuts as a warning sign of layoffs to come.
Tech companies are cutting internal AI spending, and Apple is the latest to follow suit. According to Bloomberg, Apple slashed the daily token budget for some engineers last month from $300 to $200. More cuts are planned, potentially dropping to $100 a day, and as low as $25 for some non-technical teams.
Why this matters
For developers and founders building on Apple's stack, this is a signal worth watching closely. A company with Apple's balance sheet rationing token budgets down to $25 a day for some teams tells us internal AI costs are being treated as a real line item, not an afterthought, even at the top of the industry. That has practical implications: if Apple is pushing engineers toward cheaper models and on-device inference instead of defaulting to Claude, ChatGPT, or Gemini, expect more products built with hybrid architectures that mix frontier models for hard problems and smaller local models for everything else.
For researchers, it's a data point on how far "good enough" models have come, since Apple wouldn't make this bet if cheaper options couldn't handle most workloads. And for anyone pricing API access into a business model, Ternus's cost discipline is a reminder that the token-spend era of AI development may be getting shorter, not longer, even inside companies that can clearly afford it.
Related Reading
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google to End Free Access to Gemini AI for Non-Subscribers
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
From Archives GPT Image 2.6 Model Upgrade Targets Real Animated GIF, Not Sprite Sheets
From Archives Tiny GPT Memorizes Facts But Fails to Connect Them
Further Reading
- Papers with Code - Latest NLP Research - Papers with Code
- Hugging Face Daily Papers - Hugging Face
- ArXiv CS.CL (Computation and Language) - ArXiv
