Nvidia Opens Its Infrastructure to Rival Chipmaker d-Matrix
Nvidia has opened its NVLink Fusion interconnect technology to d-Matrix, a Microsoft-backed inference chip startup valued at $2 billion. This move allows d-Matrix's Raptor processors to integrate into Nvidia-powered systems using the high-speed fabric that connects GPUs within data center racks. The partnership is part of Nvidia’s broader strategy to expand its ecosystem and maintain control over key infrastructure components like networking and switches, even when competitors supply other hardware elements. Nvidia CEO Jensen Huang has emphasized a shift in AI economics where compute becomes revenue through toll collection on interconnects rather than just chip sales. By licensing NVLink Fusion, Nvidia is positioning itself as the standard for data center connectivity while still profiting from networking and software components regardless of who supplies the accelerators. This strategy aligns with hyperscalers like Microsoft investing in specialized inference chips alongside traditional GPUs. If d-Matrix’s Raptor proves more efficient at processing tokens, it could reduce Nvidia's GPU usage in certain workloads without displacing its broader infrastructure role.
Topics: AI accelerators d-Matrix Marvell Microsoft Nvidia NVLink Fusion