Apple Cuts AI Token Budget to $200 Daily, Plans $100 by April

Apple has quietly cut the daily AI token budget for its engineers from $300 to $200, according to internal documents reviewed by The Information. The reduction took effect in late October and is the first of three planned cuts that would bring the limit down to $100 per engineer per day by April 2027.

What changed

The previous $300 daily cap already constrained teams training or fine-tuning models internally. The new $200 limit forces engineers to be more selective about which models they query and how often. Internal guidance now recommends routing routine tasks to cheaper models — such as smaller open-weight models running locally — and reserving premium APIs like GPT-4o or Claude 3.5 Sonnet for high-value debugging and architecture decisions.

At current pricing, $200 buys roughly 40 million GPT-4o tokens or about 13 million Claude 3.5 Sonnet tokens. For context, a typical code review session with context can consume 50,000–200,000 tokens. The math leaves little room for experimentation.

The roadmap: $300 → $200 → $100

Internal memos outline a three-stage plan:

1. October 2026 (completed): $300 → $200 daily cap 2. January 2027 (planned): $200 → $150 daily cap 3. April 2027 (planned): $150 → $100 daily cap

Non-technical teams may face even stricter limits — as low as $25 per day — according to the same documents seen by The Information.

Exemption process: a black box

Engineers can request exemptions for specific projects, but the criteria and timeline are not documented. Approval requires sign-off from senior leadership, and there is no public policy or SLA for how quickly requests are processed. Multiple engineers described the process as opaque and unpredictable.

"We don't know who decides, what the criteria are, or how long it takes," one engineer told The Information. "It feels like a black box."

Broader context: R&D up 32%, per-engineer budget down

The cuts come while Apple's overall R&D spending continues to grow — up 32% year-over-year to $45.8 billion in the most recent fiscal year. The contrast highlights a shift: Apple is investing heavily in AI infrastructure (data centers, custom silicon, model training) while simultaneously constraining day-to-day experimental usage by individual engineers.

This mirrors a pattern seen at other large tech companies where centralized AI infrastructure investment grows faster than per-seat experimental budgets.

Enterprise AI adoption enters allocation phase

For Apple engineers, the immediate impact is workflow friction: more time spent estimating token costs, switching between models, and waiting for exemption approvals. For the industry, it signals that even the best-resourced companies are treating AI compute as a scarce resource that must be centrally allocated.

The tension between infrastructure-scale investment and engineer-level access is likely to define the next phase of enterprise AI adoption.

Source: The Information

apple ai budget token limits enterprise ai internal tools

AI

About the author

AI Briefing Daily Editorial. The editorial desk of AI Briefing Daily, an independent publication covering artificial intelligence. Every story is researched against primary sources, checked by hand and reviewed by a person before publication.