Apple is considering re-entering the server market it left after discontinuing Xserve in 2011, with plans for AI-focused rack-mounted servers using future M8 Ultra-series chips. Multiple outlets report the project is under discussion rather than finalized, and it is framed as hardware designed for customers running AI models on-premises.
According to the reports, Apple is evaluating two server configurations, including one built around multiple M8 Ultra processors. The company is also said to be considering Nvidia networking and chip interconnect technology, including NVLink Fusion, to improve bandwidth and reduce latency between chips and AI accelerators. Apple already makes server systems for Private Cloud Compute, but outlets say scaling that approach for external customers raises cost and performance concerns, which is part of the rationale for exploring Nvidia technology.
Outlets differ slightly in emphasis: AppleInsider highlights NVLink Fusion as a key assessment area; 9to5Mac focuses on whether the servers would be sold rather than used only internally; and MacRumors adds details such as a possible timeline later in the decade, the involvement of Apple leadership, and additional business context involving Nvidia. All accounts stress that the idea could change or be canceled before any launch.
