AI inference chipmaker d-Matrix will use Nvidia’s NVLink Fusion technology to connect its next-generation Raptor XPUs to Nvidia’s AI infrastructure platform, the companies said.The move will allow d-Matrix to integrate its custom inference chips with Nvidia’s NVLink scale-up and Spectrum-X scale-out networking, as well as its MGX rack architecture and other components of its AI platform.The partnership is aimed at giving d-Matrix a faster route to deploying its inference chips at scale, while allowing it to use existing rack, networking, power and cooling infrastructure rather than building a separate architecture around its chips.“Demand for inference is soaring, but capital, time and energy remain finite,” said Sid Sheth, cofounder and CEO of d-Matrix, during a press briefing. “With NVLink Fusion and MGX, we can integrate our Raptor XPUs into a broadly deployed, liquid-cooled architecture, giving customers a faster, lower-risk path to deploy and scale ultralow-latency inference.”Also Read: Palantir, Nvidia bring sovereign AI to Nvidia’s complex supply chainNVLink Fusion is Nvidia’s technology for connecting custom XPUs and CPUs to its infrastructure stack. It is designed to allow chipmakers to focus on their processor architectures while using Nvidia’s networking, rack systems and other infrastructure for large-scale AI deployments.For d-Matrix, the integration will provide access to the Nvidia MGX ecosystem, including validated rack designs, supply-chain infrastructure, power and cooling systems. The common rack architecture is designed to support GPUs, CPUs and XPUs, potentially allowing data centres to use different types of processors without deploying a separate rack design for each.d-Matrix plans to connect its Raptor XPUs using Nvidia NVLink within a high-bandwidth, low-latency scale-up domain. Its systems will also be able to operate alongside Nvidia GPU-based systems such as the Vera Rubin NVL72 for disaggregated inference, the company said.The chipmaker also plans to integrate Nvidia Vera CPUs, ConnectX-9 SuperNICs, BlueField-4 DPUs and Spectrum-X Ethernet networking into its systems.Also Read: OpenAI launches ChatGPT for financial services industryThe partnership comes as AI infrastructure is increasingly being designed around specialised chips for specific workloads, particularly inference, rather than relying solely on general-purpose GPUs.NVLink Fusion is intended to allow customers to combine different compute architectures within a common AI factory infrastructure, giving them more flexibility to match processors to individual AI workloads.Nvidia has also announced other partners building systems around its AI infrastructure platform, including specialised XPU architectures. The company said the broader platform is designed to support different AI workloads and model architectures while optimising performance per watt and cost per token.