Apple Reportedly Planning Custom AI Server Chips to Challenge NVIDIA by 2029
According to reports from Bloomberg's Mark Gurman and The Information, Apple is developing enterprise-grade AI server chips based on future M-series Ultra processors (such as the M7 or M8 Ultra) targeted for release between 2027 and 2029. The planned servers may feature multi-chip configurations with two to four Ultra processors and could potentially leverage NVIDIA's NVLink Fusion interconnect technology. If realized, this initiative signals Apple's ambition to reduce its dependence on third-party cloud infrastructure and directly enter the AI datacenter hardware space dominated by NVIDIA. Furthermore, adopting NVIDIA's NVLink Fusion technology highlights a potential technical collaboration between the two tech giants in data center infrastructure. Reported configurations include a standard version equipped with two Ultra chips and a high-end version scaling up to four interconnected Ultra dies. Gurman anticipates an initial AI server deployment in early 2027, followed by the M7 Ultra-based server chip around 2029.
## BACKGROUND
Apple currently relies on its proprietary UltraFusion packaging technology to fuse two Max dies into a single Ultra processor for workstation-class computing. Meanwhile, NVIDIA's NVLink Fusion is an ecosystem platform that provides high-bandwidth, low-latency interconnects between chips and compute nodes in large-scale AI data centers.