According to Bloomberg reporter Mark Gurman, Apple is advancing its own in-house enterprise AI server and server chip project. One of the self-developed chips may be based on the M7Ultra and is expected to be released in 2029, competing with NVIDIA after its launch. Previously, the tech media The Information reported that Apple is planning to use self-developed AI chips in enterprise servers.
According to reports, Apple plans to launch two versions of servers, with the base model equipped with two planned M8Ultra chips, and the high-end model equipped with four. Apple is also considering using NVIDIA's NVLink Fusion to connect these chips. This technology consists of switches, chiplets, and software, enabling high-speed collaboration among multiple chips in data centers. If this plan is ultimately adopted, it will also reflect a change in Apple's collaboration with NVIDIA in the AI infrastructure field.
Additionally, Gurman mentioned that Apple is also planning to launch a new AI server, expected to go online in early 2027. Overall, Apple is extending its efforts from terminal chip design to the data center AI infrastructure. However, specific product specifications, chip models, and release dates may still be adjusted.
