HPE this week unveiled the first four ProLiant Gen 13 servers, all built on AMD's 6th-Gen Epyc 9006 SP7 "Venice" CPUs released in July, with systems shipping between next month and March 2027. The launch lands against a backdrop that Bain & Company analysts sized in September: annual AI infrastructure spending could reach $1.5 trillion by 2031, spanning datacenter capacity, GPUs, memory and networking.

The analysts also projected that power capacity in the most advanced AI datacenters — now approaching 1 gigawatt — could double by next year and exceed 9 GW by the end of the decade. Capital spending from Microsoft, Amazon, Google, Oracle and Meta is expected to hit $780 billion this year, roughly five times the level of three years ago.

What does Gen 13 actually ship?

The anchor system is the ProLiant DL585a, a 10U server holding up to eight double-wide GPUs from Nvidia, AMD or Intel and two Epyc CPUs with up to 256 cores each, plus sixth-generation PCIe connectivity. It is built for inferencing, retrieval augmented generation (RAG) and agentic AI workloads, and arrives in March 2027.

"Most of our customers who are using a PCIe-type GPU, what they're really telling us is, 'We're running out of power long before we run out of space, so make it bigger,'" John Carter, vice president of product management for HPE Compute, told The Next Platform. "We want to keep air cooling, so that's really where this is positioned. This is more for the enterprise, for the mid-tier service providers."

The ProLiant DL525, available next month, is a single-socket, air-cooled 1U system with one 256-core AMD chip, 16-channel high-bandwidth memory, DDR5 and MRDIMM support, aimed at AI inference, EDA and fraud detection. It fits a standard 19-inch rack.

Carter framed the DL525 as a shift in the CPU-to-GPU ratio. "We're expecting the ratio to shift from two to eight CPUs to GPUs, now more like one to one," he said, as agentic AI workloads demand dense CPU platforms to offload actions from GPUs.

What about the XD systems?

The XD245 and XD285, both due next year, follow on from HPE's Apollo 2000/XD 2000 series but move to an open 21-inch ORV3 rack infrastructure with shared power and cooling between systems.

"What you really get here is the ultimate density in compute. I can drive many, many, many cores in the same data space," Carter said. Both use a 2OU chassis: the XD245 carries four half-width dual-socket nodes with liquid cooling, while the air-cooled XD285 has two. Each node runs 256-core AMD CPUs with 16 memory channels per CPU and MRDIMM support.

How is HPE addressing AI security?

Enhanced management and security software, including iLO 8, runs across all four systems. iLO 8 offers self-encrypting HDD and SSD drives from the moment the system boots, plus multi-party authorization: high-impact actions by AI agents, such as key provisioning and secure erase, require two authorized human approvers. Chris Bradley, director of mainstream computer customer advocacy and technical enablement at HPE, said HPE embedded the rules into individual nodes within iLO.

HPE also expanded post-quantum cryptography (PQC) in the management stack, a defense against "harvest now, decrypt later" attacks in which threat actors stockpile stolen data to decrypt once quantum computers can break current encryption like AES-256. HPE is "ensuring that we've got the both quantum cryptography algorithms and capabilities already in the management construct," Bradley said.

The Gen 13 launch widens the infrastructure options for enterprises building out AI datacenters, but the Bain analysts injected a note of caution: "The unprecedented speed and scale of the AI buildout, with billions flowing into chips, data centers, networks, and power systems, have focused attention on the challenge of building capacity. But the more important question may be whether enough economic value can be created to justify it." With the DL585a not arriving until March 2027, HPE is betting that enterprises will still be adding dense, air-cooled GPU and CPU capacity well into the decade.