Apple Confirms iOS 27 AI Features Will Ship With Usage Limits
Apple says iOS 27 AI features including Siri AI and Image Playground will carry usage limits tied to server load, with paid access for users who hit the cap.

Updated
Why it matters
- Apple's support article published Wednesday states iOS 27 AI features including Siri AI, Image Playground, and AFM 3 Cloud models will have usage limits.
- Users who hit a limit enter a timeout period and can regain immediate access by paying a fee.
- Home app camera access via Apple Intelligence scales with iCloud+ plan: one camera on 2TB, two on 6TB, five on 12TB.
Apple will throttle heavy users of its new iOS 27 AI features and charge them for extended access, according to a support article published Wednesday.
The affected features include Siri AI — the new, smarter, more reliable version of Apple's voice assistant — plus Intelligent Photo Editing Tools such as Clean Up, Extend, and Spatial Reframing, Image Playground, the AFM 3 Cloud and AFM 3 Cloud Pro models in Shortcuts, and developer apps using Apple Foundation Models in Private Cloud Compute.
Apple's stated reason is infrastructure. Generative AI is still mostly cloud-based, meaning most of the processing runs on backend servers. When too many people tap into AI at once, those servers get bogged down juggling the workload, which can hamper performance for everyone. Apple says the latest Apple Intelligence features rely on powerful server-side models and therefore carry usage limits "to ensure a reliable experience for everyone."
The stakes are significant for the wider AI market. Every major consumer AI platform — from OpenAI to Google — is grappling with the cost of serving compute-hungry generative models at scale, and Apple's move signals that even a company with its financial resources is unwilling to offer unlimited free inference. How Apple prices and enforces these limits could set a template for how subsidized AI features become paid tiers across the industry.
Apple is fuzzy on how the limits will work. The company says only that limits may vary based on the feature, the complexity of the request, current load on the servers, system policies, and other factors.
The enforcement mechanism is a time out. Users who reach a limit on a particular feature enter a cooling-off period during which that feature is unavailable. When the timeout ends, access resumes. If a user genuinely needs the feature during the timeout, they can pay a fee to regain access immediately.
The two server-side models at the center of the policy serve different purposes. AFM 3 Cloud is the "workhorse," built for speed, efficiency, and performance. AFM 3 Cloud Pro is the more robust model, designed for demanding tasks such as agentic tools and complex reasoning. Developer apps built on Apple Foundation Models will also draw from the same pool of server-side resources.
One concrete example of tiered access already exists. Home app users with security cameras can apply Apple Intelligence to a number of cameras that depends on their iCloud+ plan: one camera on the 2TB plan, two cameras on the 6TB plan, and five cameras on the 12TB plan.
The vagueness of the policy appears deliberate. Apple is likely taking a wait-and-see approach to determine what loads its servers can withstand, when and how to limit users, and what to charge for increased access.
For now, iPhone users upgrading to iOS 27 get full access to Siri AI conversations, AI photo editing, and Image Playground's more photorealistic image generation — until they don't. The open question is where Apple draws the line between everyday use and the paid tier, and whether the timeouts arrive quickly enough to frustrate the early adopters most eager to test the new features.
Original: cc.zdnet.com
More from Rebecca Stone
Show full bio
Correspondent covering consumer brands and retail at AI In Context.
135 articles