The Information reports Apple considering AI server with M8 Ultra chips and Nvidia NVLink Fusion interconnect. Sold to outside customers — distinct from Private Cloud Compute internal infrastructure.
Scaling existing chip connections presents cost/speed problems; Nvidia networking under discussion. OpenAI and Anthropic demand for Mac minis/Studios for agent training adds industry context.
Project backed by John Ternus (hardware chief, now CEO). Potential 2029 release; cancellation possible. Bloomberg’s Gurman: M7 Ultra possible instead of M8 by 2029.