A week ago, Apple released its own AI feature, Apple Intelligence, at WWDC, including direct integration of OpenAI’s ChatGPT.Since then, the stock price has risen by more than 7%, regaining the top spot in the world in terms of market value.
The series of AI features currently announced by Apple Intelligence are not very disruptive, such as AI emojis, AI summarization and transcription, priority notifications, smart replies, photo search and editing, smarter Siri, etc. But it can also be said that this shows Apple's practicality and caution.
But unfortunately, only the iPhone 15 Pro and 15 Pro Max, as well as iPads and Macs equipped with M1 or newer chips can use Apple’s new AI features.
This means that hundreds of millions of Apple devices around the world are currently unable to use Apple AI – even if you buy the latest iPhone 15.
Why?!
Apart from the fact that this may be a “crazy hint” from Cook urging everyone to buy a new iPhone, what other reasons are there why Apple phones from iPhone 15 to previous ones cannot use AI functions?
01 It’s really related to hardware
The first is the limitation of hardware, and the reason for the limitation of hardware is related to the concept that Apple likes to promote.
Jobs took privacy very seriously. “We think very differently about privacy than some of our peers in Silicon Valley. For example, we worry a lot about location on the phone. Before any app gets location data, we require them to ask,” he said 14 years ago. “A lot of people in Silicon Valley think we're very old-fashioned on this, and maybe we are. But we worry about it.”
“I believe people are smart and some people want to share more data, so ask them and let them know exactly how you’re going to use their data, that’s what I think.”
Now, fourteen years later, Apple executives have been consciously or unconsciously emphasizing this vision. However, Musk sarcastically said that if Apple cooperates with OpenAI now, it will have to feed data to OpenAI.
Musk satirizes Apple AI at X|Image source: X
In any case, Apple is trying to partially change this situation. At the developer conference, Apple executives said when announcing new AI features: “You shouldn't hand over all the details of your life to a cloud AI just to store and analyze them.”
Currently, various AI products on the market rely more on cloud computing. Regardless of whether users use ChatGPT, Midjourney, or Photoshop's AI function, the user's own device does not do any actual work, and the information must be sent to a remote cloud server for processing.
This is not what Apple wants. In order to fulfill the privacy concept that Apple has always advocated, it wants to process various AI tasks directly on the user's device.
“The cornerstone of the personal intelligence system is on-device processing. We integrate it deeply into your iPhone, iPad, and Mac and throughout your apps, so it can understand your personal data without collecting your personal data,” said an Apple executive.
This privacy policy of processing AI on the device can be directly called “endpoint AI”. Apple is not the first technology company to use this banner. Qualcomm's PC series chip products and the latest Snapdragon chip exposed last year are built for terminal-side generative AI, and the big selling point is also privacy. Microsoft's Copilot+ is also a similar concept. AI PC allows AI tasks to be processed on the device.
Since Apple wants users to run AI on their devices, back to the original question, the reason why many older iPhone devices cannot use the new AI features is due to hardware limitations.
Features of the A17 Pro chip used in Apple's latest iPhone 15 Pro and Max | Image source: Apple
According to Wired magazine, although Apple has introduced NPU (neural processing unit) in iPhone since 2017, including iPhone 8, iPhone 8 Plus and iPhone X, the computing power required for Apple Intelligence is far greater than before.
At the same time, RAM (memory) limitations are also a key factor.
When AI models run locally rather than in the cloud, they must be stored in RAM or vRAM. While the iPhone 14’s NPU chipset is said to be more powerful than the M1 MacBook, it only has 6GB of RAM, while the devices required for Apple Intelligence require 8GB of RAM.
This means that even if older devices have AI and machine learning capabilities, they will find it difficult to run the full functionality of Apple Intelligence.
But is there really no way?
There is a view that some new AI features are actually artificially limited to new hardware. A technology enthusiast named Albacore, who focuses on reverse engineering, has made some Microsoft Copilot+ features work on hardware that theoretically shouldn't run, such as a cheap Samsung Galaxy Book2 Go laptop, which costs one-fifth to one-quarter the price of a Copilot+ laptop.
Some technology enthusiasts use reverse engineering to run AI functions | Image source: Albacore
Of course, this reverse engineering has its flaws. According to the company, the Microsoft Recall function that was reverse engineered can be run, but there is a time lag in searching, and the image generation function Cocreator, although it starts and takes up a lot of computing resources, eventually crashes.
02 Can iPhone 16 fully utilize Apple AI?
Another question is, will the iPhone 15 Pro and 15 Pro Max, as well as iPads and Macs equipped with M1 or newer chips, or the iPhone 16 purchased in the future, be able to use all of Apple’s AI functions?
Wired magazine raised relevant questions and even directly stated, “Through logical reasoning, it can be seen that all those generative AI features designed to excite people cannot be completed locally on the iPhone 15 Pro Max, and may not be implemented on the iPhone 16 Pro Max.” “The rest begins to look very much like features that other mobile phones already have, or similar to the features that Albacore manages to run on hardware without NPU.”
Because realistically, the iPhone 15 Pro Max can provide about 2GB-4GB of RAM, and its NPU speed is 35 TOPS. As a reference, Nvidia's consumer-grade graphics card GeForce RTX GPU provides AI performance of 1300 TOPS. If Apple wants to run AI directly on consumer devices, it may have to be based on a small, relatively simple AI model.
When it comes to the computing resources needed to run AI on devices, manufacturers will consider how consumers can afford it and how to balance the price and cost of a consumer-grade device.
iPhone 16 mockup image circulated online|Image source: iPhoned
Consider that currently on the market, cards like Nvidia’s H200, designed specifically for AI, cost tens of thousands of dollars per card. These high-performance graphics cards power servers that execute AI computing requests initiated in the cloud. Each card has up to hundreds of GB of vRAM, and a large number of such cards are needed to support OpenAI’s latest AI models.
If we want to run AI completely on the device, it seems that it is still some distance away from ordinary consumer-level devices. This may also explain why Apple finally contacted OpenAI, and then was ridiculed and questioned by Musk.
However, this does not seem to contradict what Jobs said back then. After all, according to Apple executives, the choice of whether to use ChatGPT is still up to the user. If a user wants to ask Siri a question it doesn't know the answer to, it can suggest switching to ChatGPT, so that users can share data and explore various creations and answers. And when ChatGPT has a “hallucination” problem, it can also be said that it is its fault, not Apple's.
And then again, even if AI functions are also “squeezing toothpaste” in the future, for Apple itself, this may also be a benefit of “squeezing toothpaste”, that is, reducing various AI security controversies and risks. After all, in the generative AI competition among technology giants, there have been many “turnover” cases. For example, Google's transformation of AI search once suggested users to apply glue on pizza, and there were also controversies about various AI tools generating problematic images.
The launch of Apple Intelligence may prompt users to consider hardware upgrades. However, it remains to be verified whether the iPhone 16 will support all Apple AI functions or continue to be an exclusive selling point for the Pro series.
Apple's first AI features will arrive later this year with the public release of iOS 18, which is expected to be released with the iPhone 16 in September.
By then, consumers may be able to personally verify the actual effects and value of Apple AI.


