Apple designs AI servers with future M8 Ultra chips

Apple is reportedly working on a high-performance enterprise server designed specifically for artificial intelligence workloads. According to reports, the server is slated for a potential 2029 release and would represent Apple's first official foray into the server market since it retired the Xserve in January 2011. The project began a year ago with backing from John Ternus, who recently succeeded Tim Cook as Apple's chief executive officer on September 1.

The proposed servers are expected to come in two distinct configurations, housing either two or four of Apple's future M8 Ultra processors. To link these powerful chips, Apple is reportedly in discussions with Nvidia to potentially utilize its NVLink Fusion networking technology. While the project remains subject to change or cancellation, integrating Nvidia's data center networking could significantly boost the system's scaling capabilities.

This hardware push is driven by the massive popularity of Apple's consumer desktops among AI practitioners. Companies like OpenAI have reportedly purchased tens of thousands of Mac Studio and Mac mini computers to train AI agents using reinforcement learning. Similarly, Anthropic has rented Mac mini hardware through Amazon Web Services. An official enterprise server would allow these developers to run heavy workloads on native Apple silicon without relying on clustered consumer desktops.

However, Apple's ambitious server roadmap faces significant industry-wide headwinds. A persistent global shortage of memory chips, driven by the rapid expansion of AI data centers, has already driven up manufacturing costs and retail prices for various consumer electronics. Securing enough high-bandwidth memory for these multi-chip M8 Ultra configurations will be a critical hurdle for Apple to overcome before its projected 2029 launch.