HPE announced four AMD-powered ProLiant Gen13 servers on October 7, introducing systems for enterprise AI, virtualization and compute-intensive applications. The Houston announcement includes a single-socket platform supporting up to 256 CPU cores and a separate server designed to accommodate up to eight GPUs.
The rollout will be staggered: HPE schedules the DL525 Gen13 for later in 2026, the DL585a Gen13 for early 2027, and the XD245 and XD285 systems for 2027. The announcement establishes a new hardware roadmap for infrastructure buyers, rather than immediate availability across the entire lineup.
Four Systems Address Different Deployment Requirements
According to HPE’s launch announcement, the systems use sixth-generation AMD EPYC processors. The company is targeting inference and agentic AI alongside analytics, simulation, modeling and established enterprise workloads.
| System | Principal configuration |
|---|---|
| ProLiant DL525 Gen13 | Single-socket platform with up to 256 CPU cores |
| ProLiant DL585a Gen13 | Accelerated platform supporting up to eight GPUs |
| ProLiant XD245 | OCP-compliant system using direct liquid cooling |
| ProLiant XD285 | OCP-compliant system using air cooling |
In its companion product overview, HPE positions the DL systems for conventional data-center power and cooling environments. The XD systems address deployments requiring greater rack density and different thermal options.
That distinction matters for facilities planning. Buyers considering the liquid-cooled XD245 will need to evaluate cooling integration, while the air-cooled XD285 offers a different deployment path. The two designs should not be treated as interchangeable additions to an existing rack.
Enterprise AI Extends Beyond GPU Capacity
Data Center Knowledge independently reported the launch and interviewed HPE executives. Its coverage identifies the DL525 as a 1U server and the DL585a as a system pairing two AMD processors with its GPU capacity.
HPE executive John Carter told the publication that the accelerated system targets enterprise customers, distinguishing it from larger rack-scale AI platforms. He also said the DL585a would support accelerators from NVIDIA, AMD and Intel.
Executives and analysts interviewed by the publication described CPU-focused systems handling orchestration, retrieval and tool calls alongside GPU-based inference. That is their assessment of how these platforms can support agent workloads, rather than a measured performance result for the newly announced servers.
Security Controls Accompany the Hardware Refresh
HPE says its iLO 8 management technology introduces automatic key provisioning for self-encrypting drives and two-person approval for sensitive operations such as secure erase. The company also describes expanded post-quantum protection for communications between system components.
These are vendor-described capabilities. Their practical value will depend on supported configurations and how customers integrate the controls into existing server administration procedures.
First Deliveries Are Still Ahead
Data Center Knowledge’s reporting gives more specific expected dates of November 2026 for the DL525 and March 2027 for the DL585a. HPE’s press release uses the broader windows of later 2026 and early 2027, respectively.
For procurement teams, the next milestone is product availability. Deployment plans should distinguish announced capabilities from deliverable configurations and confirm the applicable shipping schedule before committing dependent infrastructure work.
