Apple is developing an enterprise server built around its future M8 Ultra processors, according to reporting cited by Ars Technica. The proposed system would target AI training and inference and could arrive in 2029, although the project may still be changed or canceled.
Two configurations are reportedly under consideration, using either two or four M8 Ultra chips. Apple is also discussing whether to connect the processors with Nvidia networking technology, including NVLink Fusion. Such an arrangement would combine Apple-designed compute with an interconnect already used to scale accelerators inside data centers.
Demand for Mac mini and Mac Studio systems among AI developers appears to be part of the rationale. The report says OpenAI has purchased tens of thousands of those machines for reinforcement-learning work, while Anthropic has rented Mac minis through Amazon Web Services. A rack-oriented product could offer the same architecture in a form better suited to large deployments.
The plan remains several years from a possible release and relies on unannounced chips, making performance and availability uncertain. Memory shortages affecting the broader data-center industry add another constraint. If launched, the machine would mark Apple’s return to enterprise servers after it discontinued Xserve in 2011.