Particle.news

Apple Reportedly Testing AI Server With M8 Ultra and Nvidia Networking

The plan would mark a cautious move back into enterprise hardware to serve growing AI compute demand.

Overview

  • The Information reported Wednesday that Apple is developing an AI-focused rack server that would use two or four of its next-generation M8 Ultra chips, though the company has not confirmed the work.
  • Apple has discussed using Nvidia’s NVLink Fusion connectors to link multiple M8 Ultra dies so chips can share high-bandwidth memory and I/O for large AI models.
  • The effort is described as early-stage and aimed at AI developers, businesses and governments, with no commercial availability expected before 2029 and the project subject to change or cancellation.
  • The push responds to rising AI compute needs and strong Mac demand, with Apple’s Mac sales up 29% year over year in fiscal Q3 2026, and reports that AI teams have been buying Mac minis and Mac Studios for model work.
  • If true, the move would be Apple’s first dedicated server product since Xserve ended in 2011 and would be notable because Apple and Nvidia have had a long-strained relationship; Apple and Nvidia declined to comment and Reuters could not verify the report.