Dell and AMD Expand On-Prem AI Servers for Enterprise Environments

Dell Technologies and AMD announced an expansion of their on-prem AI infrastructure portfolio by adding support for AMD Instinct MI350P PCIe GPU accelerators in Dell PowerEdge servers. The companies also introduced a new modular architecture for Dell AI Platform with AMD.

New AI capabilities for PowerEdge servers

Starting in the summer of 2026, Dell PowerEdge XE7745 and Dell PowerEdge R7725 will support AMD Instinct MI350P PCIe GPU accelerators.

The solution is aimed at organizations that want to deploy generative AI and agentic AI workloads in their existing data centers without requiring significant infrastructure changes.

Dell and AMD are focusing on PCIe accelerators designed for standard air-cooled server environments, enabling easier deployment without transitioning to liquid cooling or changing rack architecture.

AMD Instinct MI350P with up to 4600 TFLOPS performance

According to published specifications, AMD Instinct MI350P offers:

• up to 4600 peak teraflops with MXFP4 precision
• 144 GB HBM3e memory

The companies state that this is the largest memory capacity currently available in PCIe accelerator cards.

Targeted at AI inference and agentic AI

The new infrastructure is designed for various AI scenarios, including:

• small, medium, and large model inference
• retrieval-augmented generation (RAG)
• agentic AI applications

The software stack is compatible with popular AI frameworks and tools such as:

• PyTorch
• TensorFlow
• vLLM

Dell and AMD emphasize that integration requires minimal changes to existing code.

Modular AI platform for gradual scaling

Alongside the new servers, Dell is also updating the Dell AI Platform with AMD with a modular design that allows gradual expansion of compute and GPU resources without fully rebuilding the infrastructure.

The platform uses:

• AMD Enterprise AI Suite
• AMD ROCm
• AMD Inference Server

These cover training, fine-tuning, inference, and agentic AI workflow processes.

Dell defines the platform as validated for on-prem deployment environments, where organizations seek greater control over infrastructure and data location.

AI in on-prem data centers

The companies note continued interest in on-prem AI infrastructures, especially among organizations in regulated industries and environments with sensitive data.

PCIe GPU accelerators provide the ability to add AI processing to existing server environments without the need to deploy entirely new specialized systems.

By using air-cooled architectures and standard server designs, Dell and AMD aim to reduce the need for changes in:

• cooling
• power supply
• data center layouts

Broader AMD AI portfolio

For more demanding AI workloads, Dell already offers Dell PowerEdge XE9785 servers with AMD MI355X GPU accelerators and AMD EPYC processors.

These systems are positioned for:

• foundation model development
• large-scale inference
• large-scale AI deployment environments

According to Dell, the modular approach allows organizations to start with smaller AI projects and gradually expand their infrastructure based on real needs, without a complete change of architecture.