HPE ProLiant Compute DL394 Gen12 Server with NVIDIA VERA CPU

The DL394 Gen12 departs from traditional high core count chiplet architectures by utilizing a monolithic CPU design. This approach eliminates the variable latencies and non-deterministic performance typically associated with non-uniform memory access (NUMA) in multiprocessor systems.

To support high-speed data entry, the server utilizes low-power double data rate 5X (LPDDR5X) memory. This configuration enables a total bandwidth of 1.2 TB/s with up to 14 GB/s per core, allowing the CPU to act as a high-speed orchestrator for demanding AI tasks while minimizing resource waste.

The platform is designed for environments where latency and reliability are critical. As part of a collaboration with NVIDIA and Redpanda, the infrastructure is being trialed by the New York Stock Exchange (NYSE) to handle the processing of over 1.1 trillion messages per day. The goal is to scale capacity while optimizing the throughput required for a resilient, AI-enabled market infrastructure.

Security is built into the hardware via the Silicon Root of Trust firmware. The server includes iLO 7, which provides a secure enclave to protect the system throughout its operational lifecycle. These servers are among the first to meet NIST's security requirements for resistance to quantum computing threats, making the platform suitable for regulated environments and sensitive data.

For system management, HPE has integrated Compute Ops Management. This tool provides a unified dashboard for managing distributed server environments. By using AI-driven workflows, the management layer aims to reduce manual maintenance efforts and minimize the risk of revenue loss due to system outages.

The HPE ProLiant Compute DL394 Gen12 is expected to join the NVIDIA AI Computing by HPE portfolio in fall 2026.