Apple is reportedly developing AI servers for outside customers, with proposed systems built around two or four of its planned M8 Ultra chips. The Information’s Sept. 16 report, covered by 9to5Mac, describes a possible 2029 launch aimed at AI developers, businesses and governments.
The plans remain unsettled. Apple is considering Nvidia networking technology to connect the processors, and the server project could still be canceled.
Selling hardware for customers’ own AI
The intended customers want to run AI models on equipment they own. MacRumors’ account of the report describes a focus on inference, the process of using a trained model to generate responses. It also links the project to AI businesses buying large numbers of Mac mini and Mac Studio computers.
Apple has built server hardware before. Its 2024 introduction of Private Cloud Compute describes custom Apple silicon servers that handle demanding Apple Intelligence requests beyond what a user’s device can process. Those machines support Apple’s services. The reported product would put Apple servers in outside customers’ hands.
That would revive a business Apple left when it discontinued Xserve in 2011, according to MacRumors.
Nvidia could supply the connections
The networking option is NVLink Fusion. Nvidia introduced the technology in May 2025 to let partners integrate custom chips with its AI infrastructure. It gives Nvidia a way to participate in systems built around other companies’ processors.
Apple and Nvidia already have a separate, publicly disclosed AI collaboration. In June, Apple said it was working with Google and Nvidia to run new Apple Intelligence workloads on Google Cloud as part of an expansion of Private Cloud Compute. That announcement concerns Apple’s cloud services, not an agreement to supply networking for the proposed commercial servers.
For the server project, the reported 2029 timing remains a target. MacRumors says both the product and the Nvidia arrangement are still unfinalized.
