Apple is reportedly developing enterprise-grade servers powered by its own silicon to target the AI hardware market. The company aims to sell these systems to AI developers, corporate clients, and government agencies. The hardware is expected to feature the upcoming M8 Ultra chips, with configurations offering either two or four processors per unit.
To facilitate high-speed communication between the M8 chips, Apple is considering the use of Nvidia's NVLink Fusion technology, which includes switches, chiplets, and software. Project participants view this as a leading interconnect solution. The potential collaboration signals a warming relationship between the two companies, which previously experienced long-term friction.
The project, which has been supported by Apple CEO John Ternus since his time leading hardware engineering, is currently slated for a potential release no earlier than 2029. The initiative remains subject to change, including the possibility of cancellation or a shift away from Nvidia's technology.
This move marks a potential return to the server business for Apple, which previously sold Xserve units between 2002 and 2011. The company is focusing on AI inference, aiming to provide a solution for organizations that prefer deploying models on private infrastructure rather than public clouds. Demand for Mac mini and Mac Studio systems among AI labs has grown significantly over the past year.
