Apple Weighs First Server Since Xserve, Leans on Nvidia for Networking
Apple is reportedly considering its first server since the Xserve, which was discontinued in 2011. The new server would be designed for enterprise AI workloads and could potentially launch in 2029.
The server would likely use Apple's M8 Ultra chips, with two or four possible configurations. Nvidia's NVLink Fusion technology is being considered to link the chips together, as Thunderbolt-based Mac clustering falls short of data center bandwidth requirements.
Apple has been working on plans for an enterprise AI server built on its own silicon, but it needs help from Nvidia on the networking front to make it work. The company's current flagship, the M5 Ultra, is a quad-die design with up to 512GB of unified memory and 1.2TB/s of bandwidth.
However, a data center-esque server would require considerably more power under the hood to meet its AI needs, which would also depend on how fast the interconnect between chips is. Nvidia's sixth-generation NVLink connects up to 72 accelerators at 3.6TB/s each, roughly 240 times faster than Thunderbolt 5.