Nvidia integrates Microsoft-backed d-Matrix into its NVLink Fusion fabric, pivoting toward a 'toll-collector' model that secures networking revenue regardless of which accelerator dominates the rack.
Nvidia opens NVLink Fusion to d-Matrix and Marvell, potentially making its interconnect standard and securing recurring networking revenue even if inference shifts to rivals.
Marvell joins NVLink Fusion partnership, reinforcing its position in AI networking and interconnect.
Microsoft backs d-Matrix, signaling hyperscaler interest in specialized inference chips alongside Nvidia GPUs.
Nvidia is expanding its NVLink Fusion ecosystem by integrating d-Matrix's Raptor processors, allowing the startup's chips to operate within Nvidia-powered racks. By positioning its proprietary interconnect as an industry standard, Nvidia aims to capture recurring networking and infrastructure revenue even if hyperscalers shift inference workloads to specialized third-party silicon.
By opening its rack-scale fabric, Nvidia transforms its proprietary interconnect into an industry standard, ensuring it collects revenue from networking and infrastructure components even when customers choose rival accelerator chips.