Apple is working on a server product built around its upcoming M8 chip series, according to a new report. The project would mark Apple’s first return to selling server hardware since it killed off the Xserve in 2011.
The Information reported the plans, citing people familiar with the effort. The new server would target AI inference and response generation, and Apple intends to sell it directly to business customers.
Apple’s New Server Chip Leans on Nvidia’s Networking Tech
Apple has talked with Nvidia about using its NVLink Fusion networking technology to connect chips inside the new server. Scaling Apple’s own chip-to-chip connections to server-grade levels would create cost and speed problems, according to The Information.
NVLink Fusion is Nvidia’s interconnect technology, built to let chips from outside vendors link into Nvidia’s broader server networking stack. Apple turning to a direct competitor’s networking hardware for its own silicon would be a notable shift for a company that designs nearly everything else in-house.
Apple already runs its own server hardware for Private Cloud Compute, the infrastructure behind several Apple Intelligence features. A commercial AI server would be a separate product aimed at outside businesses.
The Project Traces Back to John Ternus and Could Ship by 2029
Work on the AI server reportedly started roughly a year ago. Apple CEO John Ternus backed the initiative while he was still running hardware engineering, before taking over as CEO earlier this year. His support suggests the project has backing at the top of the company.
The Information’s sources say a release could land as early as 2029, though the project remains subject to cancellation before then. AI companies have already been buying Mac Mini and Mac Studio systems in bulk to run AI workloads, and that demand appears to be part of what pushed Apple to consider a dedicated server product.
The report points to gaps Apple would need to close before a server business could work.
- Better support for enterprise customers
- More investment in software resources for AI developers, including Apple’s MLX machine learning framework
Our Take
The Nvidia partnership is super important here. Apple has spent a decade proving it doesn’t need outside silicon partners for anything. Reaching for Nvidia’s networking stack tells me Apple’s own interconnect approach can’t scale to server density without a redesign it isn’t ready to build yet.
