Apple is preparing a 2029 AI server launch and evaluating Nvidia’s interconnect technology to support its internal compute ambitions.

Apple is reportedly developing an enterprise AI server designed to leverage Nvidia’s high-speed interconnect technology. Targeted for a launch as early as 2029, the system is specifically aimed at the “AI inference” market—handling real-time user requests—rather than model training.

According to a report by U.S. tech publication The Information on the 16th (local time), Apple is evaluating the adoption of Nvidia’s “NVLink Fusion” to enhance connectivity between its proprietary server processors. NVLink Fusion integrates switches, chiplets, and software to enable high-bandwidth, low-latency connections across chips from different manufacturers. This architecture allows enterprises to unify their own CPUs, AI accelerators, and Nvidia GPUs within a single infrastructure.

The server will be built around Apple’s in-development M8 Ultra chip and is expected to ship in two configurations: one featuring dual M8 Ultra processors and another with quad setups. Primary target customers include AI developers, corporate enterprises, and government agencies. While some Apple engineers involved in the project consider NVLink Fusion the best connectivity solution currently available on the market, The Information cautioned that the initiative remains subject to cancellation. Neither Apple nor Nvidia has officially confirmed ongoing negotiations.

The project reportedly gained traction approximately a year ago under John Ternus, who was then leading Apple’s hardware engineering division before becoming Apple’s CEO earlier this month. Sources familiar with the development note that his sustained executive backing has been instrumental in keeping the initiative alive.

Apple’s push into the AI server space follows an unexpected surge in Mac demand. In the most recent quarter, the Mac emerged as Apple’s fastest-growing product category, generating $10.4 billion (approximately 14.3 trillion won) in revenue, a year-over-year increase of roughly 29%. A wave of AI developers purchasing Mac minis and Mac Studios in large volumes triggered widespread supply shortages, simultaneously highlighting the limitations of consumer-grade hardware when tasked with enterprise-level AI workloads.

The potential partnership also holds strategic significance for Nvidia. As the dominant force in the AI server market, Nvidia currently derives roughly one-fifth of its data center revenue from networking equipment. An Apple adoption of NVLink Fusion would establish a massive new enterprise client while validating Nvidia’s strategy of monetizing interconnect technology, even as competing silicon continues to erode its GPU market share.