Apple Might Make Servers Again to Cash In on the AI Rush
The Information reported on 16 September 2026 that Apple is developing an enterprise AI server built around future M8 Ultra silicon, in two- and four-chip configurations, and has discussed using Nvidia’s NVLink Fusion interconnect to wire them together. Nothing is finalised, nothing ships before 2029, and the project could be cancelled. This article separates the single-sourced claims from the confirmed facts: Apple’s Houston plant has been shipping Private Cloud Compute servers since October 2025, the shipping M5 Ultra runs at 1.2TB/s of unified memory bandwidth, and Apple’s roughly $13 billion of annual capex sits against $725 billion of combined 2026 guidance from four rivals.