Dell Expands AI Platform With AMD Instinct MI350P Configurations

Dell is expanding its Dell AI Platform with AMD with two modular configurations based on PowerEdge XE7745 and R7725 servers, AMD Instinct MI350P PCIe GPUs and AMD EPYC CPUs. The company says the additions are designed to help enterprises move AI projects from pilots into production while retaining control of data, costs and governance.
- Dell adds two AMD Instinct MI350P platform configurations.
- The systems use PowerEdge XE7745 and R7725 servers.
- The platform is validated for AMD Enterprise AI and Red Hat AI Enterprise.
- Dell positions the modular design for scaling from pilots to production.
- The platform supports training, fine-tuning, inference and agentic workflows.
Two new configurations target production AI
Starting in November 2026, the expanded platform will offer configurations built on the Dell PowerEdge XE7745 and Dell PowerEdge R7725. Both use AMD Instinct MI350P scalable PCIe GPUs and AMD EPYC CPUs, bringing Dell’s modular architecture to AMD Instinct GPU systems.
Dell describes the platform as a full-stack, validated infrastructure solution. It combines AI-optimized PowerEdge servers with Dell PowerSwitch networking, Dell PowerScale storage, the Dell AI Data Platform, AI software and tailored services. The company says the system supports workloads ranging from training and pretraining to inference at scale, for enterprises and service providers.
The configurations follow an earlier option introduced in the summer, based on PowerEdge XE9785 servers with eight-way AMD Instinct MI355X GPUs and AMD EPYC CPUs. Dell says that configuration is intended for large global enterprises, NeoClouds and service providers requiring substantial capacity for training, fine-tuning and inference.
Software support expands across AMD and Red Hat
Dell says the platform is now validated for the AMD Enterprise AI reference architecture and Red Hat AI Enterprise, which includes Red Hat OpenShift. The additional software support is intended to give enterprises more flexibility in selecting tools for deploying and operating AI workloads.
AMD’s Enterprise AI Reference Stack includes AMD Inference Microservices, known as AIMs, and Solution Blueprints. According to Dell, these capabilities can simplify deployment of open-weight models and help organizations move more quickly from development to production.
The platform also draws on ROCm 10, which Dell describes as an AI-native software platform for developing and optimizing workloads on AMD GPUs. The company highlights AMD Skills for AI-assisted development and Hyperloom for advanced debugging as features intended to improve the developer experience and help teams make better use of AMD Instinct GPUs.
A modular approach to scaling inference
Dell’s announcement comes as more AI pilots and proofs of concept move into production. The company says inference demand is rising as AI reaches more functions, users and use cases, with agentic AI adding demand through systems that plan and complete multistep tasks.
Dell argues that token-based cloud services can become unpredictable and harder to govern at enterprise scale. Its on-premises platform is positioned as an alternative foundation that connects compute, storage and networking with enterprise data, software and services. The company says this approach can simplify data security and governance, lower total cost of ownership and reduce business risk.
The modular design is intended to let organizations begin with a single 2U compute node, reuse that infrastructure for production and add nodes, storage and networking as demand grows. Dell says customers can scale GPUs, CPUs, nodes, bandwidth or capacity in smaller increments, while built-in security and governance controls help protect sensitive data as AI use expands.
Dell Technologies: Dell AI Platform With AMD Adds New Modular Configurations


