Apple could pair its own AI chips with NVIDIA’s NVLink to re-enter the server market

Apple appears to be preparing an unusual move for its server business: The company is considering using NVIDIA NVLink technology to connect its own AI chips. This will help integrate future Apple servers with modern artificial intelligence infrastructure.

It was previously reported that Apple and Broadcom are collaborating to develop the first dedicated artificial intelligence chip for servers, code-named Baltra. It is expected that its production will be organized using TSMC’s N3E 3nm process technology, and its design will adopt a chiplet architecture. At the same time, Apple wants to independently control final assembly and even keep key development features secret from partners.

Now The Information reports that Apple is also exploring servers based on its own M8 chip. The company is said to be considering integrating NVLink Fusion, an NVIDIA technology designed to connect third-party accelerators to scalable artificial intelligence systems. This approach will allow Apple to ensure compatibility of its solutions with NVIDIA’s server infrastructure and potentially provide devices to enterprise customers.

If the plan comes to fruition, Apple will actually be returning to the server market years after Xserve was discontinued in 2011. The bet this time will not be ordinary servers, but the infrastructure used to handle artificial intelligence tasks.

Other publications on this topic