Apple Building AI Servers Powered by M2 Ultra Chips

According to a reliable source, Foxconn has begun production of Apple’s AI servers.

Apple is said to have started manufacturing its own AI servers using the M2 Ultra chip.

Based on analysis by Jeff Pu, Foxconn is building M2 Ultra chip-based AI servers and plans to build M4 chip-powered servers by late 2025.

It was previously reported that Apple is developing a processor for AI servers using TSMC’s 3nm process and will bring it to mass production by the second half of 2025. According to the new report, these chips will likely be of the M4 type.

The flagship M2 Ultra chip contains 134 billion transistors and is made up of 24 CPU cores (16 powerful cores and 8 efficient cores), 60 graphics cores, 192GB of internal RAM with a bandwidth of 800GB/s, and a 32-core neural engine.

According to Apple, the CPU and GPU performance of these chips has been improved by 20% and 30%, respectively, compared to the previous generation.

Apple’s decision to build AI servers is indicative of the company’s strategy to better align its hardware with its software ecosystem. Apple will likely use its chips to boost the performance of its data centers and future AI tools that rely on cloud computing.