Apple's latest batch of artificial intelligence features comes with a bill that is not measured in dollars. According to 9to5Mac, iOS 27 shipped this week with a major overhaul to Apple Intelligence, and on certain newer iPhone models the amount of free storage the system wants set aside for those features is now nearly double what it used to be. Plenty of other iPhones see no change in their AI storage requirement at all, which makes the jump both notable and oddly selective.
That uneven outcome is the real story. A software update that changes the storage math on some devices but not others says something about how Apple is dividing AI work between hardware generations — and about what buyers of the newest phones are quietly signing up for.
A major overhaul, unevenly applied
The iOS 27 release is not a minor point update with a few extra suggestions in the keyboard. Reporting describes a substantial reworking of Apple Intelligence across the board, with the assistant experience sitting at the center of the changes. A companion look at macOS 27, published by 9to5Mac on September 17, 2026 and focused on Siri's new AI capabilities, frames the assistant as finally coming alive — a sign that Apple's assistant, rather than any single standalone feature, is the flagship of this cycle.
Storage requirements, though, have not moved in lockstep across the lineup. For many iPhones, the previous footprint stands. For the newest models, the demand has grown to close to twice the earlier figure. That difference need not be read as a limitation; read another way, it reflects newer hardware being asked to do more work locally rather than sending it off to a server.
Why the newest phones feel it most
On-device AI is a hardware story as much as a software one. Faster neural engines and larger memory pools let a phone run bigger models without a round trip to the cloud. When a device gains that headroom, the software tends to expand to use it — richer speech recognition, better image understanding, and persistent indexes that make search and suggestions feel instant. The result is a feature set that looks identical in a screenshot but occupies very different amounts of disk depending on which phone is running it.
Why on-device intelligence is so storage-hungry
The storage cost of AI is easy to underestimate, because none of it shows up as an icon you can delete. In broad terms, an on-device assistant leans on several kinds of data that all have to live on the phone:
- Model weights. The compressed neural networks that handle language, speech, and image tasks must be present locally to run offline or quickly. Even aggressively quantized, they are measured in gigabytes rather than megabytes.
- Runtime support files. Frameworks, tokenizers, and hardware-specific kernels sit alongside the models so the right silicon can be used efficiently.
- Local indexes. Semantic search across photos, messages, notes, and documents requires a searchable representation of that content, and it grows along with your library.
- Caches and temporary state. Context windows, embeddings, and intermediate results occupy space while the system works, and a ceiling for that scratch space has to be reserved.
Stack those together and a device can plausibly need several gigabytes of breathing room for features that once amounted to little more than a voice-command parser. Doubling that requirement on newer hardware is a smaller relative leap than it sounds — but it is a leap that lands squarely on the phone's free space.
What it means for buyers
Storage has been the quiet dividing line in smartphone shopping for years. Base models push buyers toward higher tiers, and AI features add a fresh reason to care. A phone that reports plenty of total capacity can still feel cramped if a meaningful slice is effectively earmarked for intelligence features, on top of photos, cached media, and the operating system itself.
Free space is the number that matters
The distinction between total capacity and usable free space becomes more important with every AI release. Updates that expand on-device models can effectively shrink a device long after purchase, and the squeeze tends to arrive without a headline feature attached. Owners who chose a smaller configuration a year or two ago may find that the margin they counted on has quietly narrowed.
Managing the squeeze
There is no switch that removes Apple Intelligence from the equation without giving up the features built on top of it, so the practical options are the familiar ones:
- Check what is actually consuming space before assuming AI is the culprit; photos, video, and cached streaming media usually lead the list.
- Offload rarely used apps rather than deleting them outright, keeping documents and data in place.
- Lean on cloud libraries for photos and messages where that trade-off is acceptable.
- When buying, treat the AI footprint as a permanent tax on the storage tier you select rather than a one-time install.
The privacy trade-off behind the numbers
The reason these models live on the phone at all is that processing locally avoids shipping personal data to remote servers. That is a meaningful privacy advantage, and it is also the source of the storage pressure. Cloud-based assistants can behave like thin clients; on-device ones carry their capability with them. Every gain in offline smarts has to be paid for somewhere in the device's budget of space, memory, and battery.
What to watch next
The open question is whether the requirement keeps climbing. If the figures introduced with iOS 27 hold steady through the cycle, the near-doubling on newer models may represent a one-time step up to match the hardware. If subsequent releases continue to grow the footprint, buyers will have to plan for AI as a fixed cost of ownership. For now, the takeaway is simple: the newest iPhones may be the most capable at Apple Intelligence, but they are also the models where that capability takes the biggest bite out of free storage.
This article is based on reporting by 9to5Mac. Read the original article.
Originally published on 9to5mac.com
