Sources: Apple is developing AI servers that could use M8 Ultra chips connected via Nvidia's NVLink Fusion, with plans to sell them to developers and others

Apple may offer AI servers for sale, giving businesses an alternative to cloud-based AI infrastructure.

Apple is reportedly developing AI servers that could utilize its M8 Ultra chips and Nvidia's NVLink Fusion technology, with a potential release in 2029. This initiative aims to bring Apple back into the server market, a segment it exited in 2011 after discontinuing Xserve. The proposed servers would be sold to external customers, primarily companies seeking to run AI models on-premises for tasks like inference. Apple's consideration of Nvidia's NVLink Fusion stems from potential cost and speed limitations in scaling its internal chip-to-chip connections. The project, which gained traction about a year ago, follows increased demand for Apple hardware among AI businesses buying Mac mini and Mac Studio systems.

This potential move signals Apple's intent to re-enter the enterprise hardware market with a specialized AI offering, leveraging its in-house silicon alongside third-party networking solutions. The focus on inference suggests Apple is targeting practical AI deployment scenarios for businesses, aiming to capture a segment that prioritizes local data processing and control over sensitive AI models. The collaboration with Nvidia, even for networking components, represents a significant step in bridging Apple's historically closed ecosystem with key industry partners for advanced computing.

The development could affect the competitive landscape for AI hardware, offering enterprises a choice beyond hyperscalers and specialized AI hardware vendors. Success hinges on Apple's ability to deliver robust software support, a competitive price point, and a compelling total cost of ownership compared to existing solutions. Investors and competitors will watch whether this foray into AI servers marks a sustained commitment to the enterprise or a niche offering.