Apple has rolled out its custom "A" and "M" series chips across all of its devices, including the iPhone, iPad, and Mac, but a new report suggests the company may be expanding the scope of its custom silicon even further by developing its own AI accelerator.
According to information posted by the leaker "Phone Chip Expert" on China's Weibo, Apple is developing a custom chip for AI servers. While not going into much detail, they claim that this chip will be manufactured using TSMC's 3nm process.
According to the leaker, this custom chip will enter mass production in the second half of 2025. TSMC's 3nm process has several variations, and while Apple used "N3B" for mass production of the A17 Pro last year, if this AI server chip is indeed coming in late 2025, it's likely that "N3E," an improved version of N3B, will be used instead. The N3E node is said to be used in Apple's A18 Pro, the upcoming Snapdragon 8 Gen 4, and the Dimensity 9400. At this point, it remains unclear what Apple's ultimate goal is in developing a custom chip in this category.
Previous reports have suggested that with iOS 18, Apple may shift AI-related task processing to on-device rather than relying on the cloud. However, if this new information is accurate, it suggests that Apple isn't solely focused on AI that runs locally on-device. Devices capable of on-device processing would be equipped with cutting-edge chipsets like the A18 Pro, and to enable this, they may also feature larger Neural Engines. For older SoCs that don't have sufficient processing power to support on-device AI features, cloud-based processing would be necessary. Apple's AI server chip could improve the performance of fast cloud-based processing and other AI-centric features.
The claim that Apple would need AI servers and would design a processor with TSMC seems like a logical conclusion. However, since this is all speculation and not an official announcement, it's best to keep that in mind.
Sources