HPE expands ProLiant Gen13 server lineup for AI workloads

Key Takeaways

• HPE’s first four ProLiant Gen13 servers are built around 6th‑Gen AMD EPYC CPUs, delivering up to 96 cores per socket and PCIe Gen 6 bandwidth.
• The lineup targets AI inference, agentic‑AI workloads, high‑speed analytics, and mission‑critical virtualization while embedding post‑quantum‑ready security via iLO 8 and Compute Ops Management.
• Four models address distinct density and performance needs: DL585a, DL525, XD245, and XD285.
• According to McKinsey’s 2026 AI Trust Maturity Survey, ≈ 66 % of enterprises cite security and risk as the primary obstacle to scaling autonomous AI.

HPE ProLiant Gen13: A Trust‑Centric Platform for the AI Era

HPE has launched the inaugural quartet of ProLiant Gen13 servers, each powered by AMD’s Zen 4‑based EPYC 7004 family. The processors deliver up to 96 cores, 384 GB/s of memory bandwidth per socket, and PCIe Gen 6 lanes, enabling unprecedented data‑movement rates for modern AI pipelines.

Beyond raw performance, the Gen13 series embeds HPE iLO 8 enhancements—automatic firmware hardening, hardware‑rooted trust, and a roadmap toward post‑quantum cryptography—plus Compute Ops Management for unified visibility across distributed environments. These capabilities directly address the security‑risk concerns highlighted by the McKinsey survey, positioning the servers as a “trusted infrastructure foundation” for autonomous AI workloads.

Model‑by‑Model Highlights

DL585a – The GPU‑Heavy Workhorse

The DL585a is engineered for workloads that demand massive parallelism. Its eight double‑wide PCIe Gen 6 GPU bays can accommodate the latest NVIDIA H100 or AMD Instinct MI250X accelerators, delivering up to 2 TB/s of aggregate GPU‑PCIe bandwidth. Coupled with dual EPYC sockets, the server can sustain > 4 PFLOPS of FP16 performance for inference‑first AI pipelines.

DL525 – Compact, Single‑Socket Powerhouse

At just 1U, the DL525 packs a single EPYC processor with up to 256 GB/s of memory bandwidth and 16‑channel DDR5 support. Its high‑speed I/O makes it ideal for latency‑sensitive analytics—think real‑time fraud detection or high‑frequency trading where every microsecond counts.

XD245 & XD285 – OCP‑Compliant Density Champions

Both XD245 and XD285 follow the Open Compute Project (OCP) 3.0 spec, offering shared power and cooling to reduce rack footprint. The XD285 pushes density further with eight single‑width GPU slots, enabling up to 1.8 PFLOPS of mixed‑precision training in a 2U chassis.

Security & Operational Simplicity

• iLO 8 now includes automatic firmware attestation and hardware‑rooted trust that align with emerging post‑quantum standards.
• Compute Ops Management provides a single pane of glass for health monitoring, firmware lifecycle, and policy enforcement across all Gen13 nodes.
• Integrated Secure Boot, TPM 2.0, and encrypted memory protect data at rest and in motion, directly mitigating the security concerns flagged by the McKinsey survey.

Bottom Line

HPE’s ProLiant Gen13 servers marry AMD’s 6th‑Gen EPYC performance with hardened, post‑quantum‑ready security and OCP‑grade efficiency. Whether an enterprise needs a GPU‑dense inference platform (DL585a), a space‑constrained single‑socket server (DL525), or ultra‑dense OCP‑compliant compute (XD245/XD285), the Gen13 family offers a purpose‑built solution that tackles both the performance and trust challenges of today’s agentic AI deployments.

Reference

1. McKinsey & Company, AI Trust Maturity Survey 2026, accessed Oct 2026.