Apple is reportedly developing its own chips to run artificial intelligence software in data center servers.
According to a report in the Wall Street Journal, the chips, codenamed Project ACDC – which stands for Apple Chips in Data Centers – have been in development for several years. Apple has reportedly been working with TSMC to design and manufacture the chips.
Although it is unclear what the results of those initial trials have been, the WSJ report said the Apple chips are likely to focus on inferencing.
However, according to anonymous sources cited by the WSJ, even if the chips are successfully produced, it’s possible that Apple might never publicly announce them.
Last month, DCD exclusively reported that Apple had hired former Google, IBM, and Nvidia executive Sumit Gupta as director of products for Apple Cloud. At Google, Gupta said he was responsible for the company’s data center infrastructure products, including TPUs, Arm CPUs, storage, and networking.
He also served as the chief AI strategy officer and CTO of AI at IBM and prior to that was the general manager of AI and GPU-accelerated data centers at Nvidia.
The likes of Meta, Google, Microsoft, and Amazon have all developed their own custom silicon focused on AI, with the three cloud providers offering access to this hardware through their respective platforms.
Earlier this week, Apple launched its new iPad Pro containing the latest version of the company’s 3nm system on a chip (SoC) offering. Designed to improve performance for AI-related tasks in iPadOS, according to Apple, the CPU in the M4 is 50 percent faster than the M2, while the latest version of the chip also contains two more cores, four for performance and six for efficiency.
Analysis of the M4 provided by Geekbench said the M4 is rated for 38 Tops of Integer (8-bit) accuracy computation, giving it a five percent in Tops when compared to the M3.
