Apple is developing an enterprise AI server powered by its own M8 Ultra chips, targeting a 2029 release that would mark the company's first server hardware in nearly two decades, according to a report from The Information2,4,5.
The server, reportedly code-named Baltra, would come in two configurations using either two or four M8 Ultra chips3. Apple is also considering integrating Nvidia's NVLink Fusion interconnect technology into the product.
Ternus and the hardware bet
The project reportedly received backing from John Ternus, now Apple's CEO, when it began roughly a year ago, at a time when Ternus led Apple's hardware engineering division. Apple retired its Xserve rack-mount server line in 2011 and has largely ceded enterprise server hardware to other manufacturers since.
The return to servers follows strong demand for Apple's existing desktop hardware among AI practitioners. Apple's Mac Mini and Mac Studio, which use M-series Ultra chips, have seen booming sales as AI developers and companies adopt them for inference and other AI workloads. That demand has led to shortages of the devices.
ARM in the data center
Apple's ARM-based M-series processors offer a power-efficiency profile distinct from the x86 and GPU-centric architectures that dominate current AI server infrastructure. The growing demand for compute as the AI industry expands has apparently led Apple to see an opening for its chips in enterprise settings.
The potential Nvidia partnership is notable: NVLink Fusion is Nvidia's interconnect standard for linking its GPUs with non-Nvidia processors. Incorporating it could allow Apple's M8 Ultra chips to work alongside Nvidia accelerators or tap into Nvidia's networking ecosystem.
ANALYSIS A 2029 timeline places the product three years out, meaning Apple would need to align chip development, server design, and any Nvidia integration across multiple hardware generations. The gap between current Mac desktop popularity and a shipping enterprise server is substantial, and the competitive landscape for AI server hardware will have evolved considerably by then.
The two-or-four chip configuration suggests Apple is designing for scalable, multi-chip compute nodes rather than standalone appliances, positioning the product for data-center-class workloads rather than simply rack-mounting existing Mac hardware.