According to reports, Apple is developing its artificial intelligence (AI) chipsets so that data centres can run software. If accurate, this would represent a significant shift in strategy for the massive Cupertino, California-based tech business, since prior reports indicated the company was concentrating primarily on on-device AI features for its products. But the advancement of AI chips suggests Apple is also thinking about AI functions that run on servers. It’s possible that some of these features will be revealed at the June 10 Worldwide Developers Conference (WWDC) in 2024.
Apple has reportedly been working on its server project, internally referred to as Project ACDC, for a number of years, according to a Wall Street Journal story. The report stated that the tech giant is currently aiming to produce unique chipsets for these data centres so they can run AI computing, citing anonymous persons acquainted with the topic.
The tech giant, as per reports, is designing and manufacturing these AI chips in collaboration with Taiwan Semiconductor Manufacturing Co. (TSMC), although it’s unclear if the produced processors produced a conclusive outcome. It is reported that these chips are not employed for training large language models (LLMs), but rather are solely utilised for inference (running the AI model). Notably, TSMC manufactures the majority of the silicon used in Apple products and is a significant chip-making partner of the business.
If the reports are accurate, Apple’s AI approach may be about to change significantly. The corporation was reportedly primarily interested in providing on-device AI technologies that could be processed fully locally, according to numerous earlier reports. The new iPad Pro 2024 with the M4 chip, which has a new Neural Processing Unit (NPU) to perform AI tasks, was unveiled as proof of this ambition.
However, because of the high hardware requirements, no smartphone manufacturer has been able to provide a suite of AI features that split the processing between the device and the server and work just locally. The AI capabilities of Oppo and Samsung’s Galaxy AI are noteworthy examples.
This implies that non-privacy-invading features may be hosted on servers, while locally processed features that access user data may still be handled. Additionally, as these devices lack the hardware necessary to handle on-device AI functions, Apple may be able to bring similar features to its earlier models. Even if these are merely conjectures, if Apple discloses its AI features at WWDC 2024, we will know for certain.
