Apple is reportedly exploring the integration of Nvidia's NVLink technology into its upcoming custom AI servers. The project, which began approximately a year ago, aims to enhance connectivity for servers powered by Apple's future M8 Ultra processors.
The company is considering two primary server configurations: a smaller model featuring two M8 Ultra chips and a high-end version equipped with four. Apple currently utilizes its own connectivity solutions for Private Cloud Compute, but these are reportedly insufficient for large-scale commercial deployments.
The potential adoption of NVLink infrastructure could include switches, specialized chiplets, and software stacks. This move suggests that Apple may look to leverage Nvidia's broader data center ecosystem, including Spectrum-X Ethernet or Quantum-X InfiniBand, for its infrastructure needs by 2029.
While Apple is a member of the UALink Consortium, which develops industry-standard accelerator interconnections, the company appears to be evaluating Nvidia's proprietary technology. This potential collaboration marks a significant shift in the historical relationship between the two firms.
