Two separate reports outline Apple's plans to build a custom AI server powered by its M-series Ultra chips, with a target launch in 2029. Both sources agree on the core design: a server packed with Apple's own silicon, aimed at data center AI workloads. The reports are based on unnamed sources, so the details remain unconfirmed.
The reports diverge on one significant point. Tom's Hardware says Apple is reportedly interested in using Nvidia's NVLink Fusion interconnect for these servers, a notable development given the historically bitter relationship between the two companies. Ars Technica, by contrast, makes no mention of Nvidia and instead emphasizes that this would be Apple's first enterprise server in decades.
If accurate, the project would mark a major step in Apple's data center AI push. But the Nvidia angle is only present in one source, so readers should treat that specific claim with extra caution until Apple or Nvidia comment.