HPE Gen13 ProLiant servers: what to know about the AI and HPC launch

The launch report summarizes the specifications and workload uses HPE announced. Those descriptions explain how the company intends the systems to be used; they are not independent performance results.

How do the four Gen13 servers differ?

HPE describes four different configurations:

• ProLiant DL525: A 1U, single-socket server for data-intensive work including electronic design automation, simulation and AI inference. HPE says it supports a processor with up to 256 cores, 16-channel memory and PCIe Gen6 connectivity.
• ProLiant DL585a: A system aimed at GPU-accelerated workloads such as inference, retrieval-augmented generation, fine-tuning, analytics and agentic AI. HPE says it supports two sixth-generation AMD EPYC processors and as many as eight double-wide PCIe GPUs, with PCIe Gen6 connectivity.
• ProLiant XD245: An OCP-compliant, four-node design occupying one Open Rack Unit, with direct liquid cooling. HPE positions it for data-intensive AI and HPC workloads.
• ProLiant XD285: An OCP-compliant, two-node system occupying two Open Rack Units and using air cooling. HPE also targets it at data-intensive AI and HPC workloads.

The differences are concrete: processor and GPU capacity, node count, rack layout and cooling method. They give infrastructure teams specifications to compare with their workloads and facilities, but do not show that one system will deliver better results for a particular application.

What do the specifications establish?

The published details describe what HPE says each design supports. They do not measure how quickly a system will run a customer’s application, how much power it will use, or what it will cost to operate. The materials reviewed do not include independent Gen13 benchmarks or prices.

That distinction matters for buyers comparing AI or HPC capacity. A stated GPU count or processor configuration is not evidence of application performance, token throughput, cooling efficiency or total cost of ownership. HPE’s workload descriptions should be treated as intended uses, not verified results.

The designs also use different cooling approaches, but the announcement does not provide comparative cooling-efficiency results. Buyers can identify the systems’ stated cooling types; they cannot use the launch details alone to calculate operating costs or decide how a system will perform in their facility.

When will the Gen13 servers be available?

HPE’s announced schedule puts the DL525 later in 2026, the DL585a in early 2027, and the XD245 and XD285 in 2027. Those are future availability windows, not confirmation that the servers are generally available, orderable or shipping. The announcement does not specify prices.

The launch gives buyers a first look at four planned Gen13 configurations and their intended workloads. Order timing, pricing and independent results for real workloads remain unanswered in the materials reviewed.