HPE ProLiant Gen13: Open, Modular, and Built for Enterprise AI

By Jeni Barovian, Co-Founding Advisor, TechArena

The Open Compute Project (OCP) was founded on a straightforward premise: making hyperscale infrastructure innovation accessible to a broader market. Historically, hyperscalers drove much of that innovation, but now neoclouds, smaller data center operators, and enterprises are also adopting open infrastructure technologies to support their AI workloads. Organizations that spent the last few years on AI experiments are moving their workloads into production. They need compute infrastructure that can simultaneously support applications, fit existing facilities, and meet security and operational requirements.

HPE addresses these challenges with its new ProLiant Gen13 portfolio, combining workload-specific server architectures, open hardware standards, and enterprise security and management capabilities.

I spoke with Joseph "JBK" Knight, Chief Technologist for Service Providers at HPE, to take a closer look at the new systems and how HPE applies OCP standards to enterprise infrastructure.

Enterprise AI Requires More Compute Choices

One of the biggest challenges in enterprise AI is the diversity of applications. AI training has traditionally dominated infrastructure discussions. Now, inference and agentic AI are expanding the range of compute requirements. Contract analysis, manufacturing defect detection, and engineering simulation demand different levels of processing power, memory capacity, and acceleration.

Enterprises also face practical deployment constraints. Unlike hyperscalers that can build new facilities around specific infrastructure needs, many organizations need to work within existing power and cooling capacity. For some facilities, direct liquid cooling means substantial upgrades.

As a result, organizations need server designs that match their applications and deployment environments. While GPU acceleration dominated AI training deployments, racks with lower GPU:CPU ratios, or CPUs alone, are central to inference- and agentic-optimized infrastructure.

HPE's ProLiant Gen13 portfolio offers reliability through built-in security, simplified management, and lifecycle support. For organizations moving AI into production, these capabilities are essential to deploying and operating infrastructure at scale.

Four Platforms for Different Workloads

ProLiant Gen13 includes four platforms powered by 6th Gen AMD EPYC 9006 Series processors. Each addresses different performance needs, space constraints, and cooling capabilities, giving organizations more flexibility to choose infrastructure that fits their applications and facilities:

1. HPE ProLiant DL525 Gen13: This compact, single-socket server delivers high CPU performance and memory bandwidth for demanding applications, including AI inference, engineering simulations, and virtualization. With support for up to 256 cores, it provides substantial processing capacity without the added complexity of a multisocket system. Its modular design also allows for greater configuration flexibility.
2. HPE ProLiant DL585a Gen13: This accommodates up to eight GPUs for inference, agentic AI, fine-tuning, and other demanding workloads. Its air-cooled design is particularly valuable for customers that want to expand their AI capabilities without the cost and disruption of retrofitting facilities for liquid cooling.
3. HPE ProLiant XD245: Designed for organizations that need to maximize compute capacity within limited data center space, this system uses direct liquid cooling to enable higher-density deployments. Its OCP Open Rack V3, or ORV3, architecture also provides a modular foundation for customers adopting open infrastructure standards.
4. HPE ProLiant XD285: This air-cooled platform provides similar rack-scale flexibility, allowing organizations to increase compute density within existing facility constraints. It's particularly relevant for organizations that want the benefits of OCP-based infrastructure without introducing liquid cooling.

The DL585a's air-cooled design was particularly interesting to the customers JBK works with. "A lot of our customers are in spaces where air cooling is the only option," he said. "So we designed this to give the best air cooling for the GPU densities that they need for inferencing." This provides a practical path to GPU acceleration without major facility upgrades.

Together, these systems allow enterprises to select infrastructure based on their actual workload demands and operational constraints, rather than adapting their applications or facilities to a single server architecture.

Open Standards With Practical Benefits

I wanted to understand how open hardware approaches translate into practical advantages for enterprise customers.

JBK's answer focused on two priorities: greater power and cooling efficiency, and more flexibility in how organizations configure and deploy their infrastructure.

"Customers don't want to be locked in," he said. "They want freedom."

That's why HPE applies open specifications to develop modular systems with standardized components, enabling more configuration options without sacrificing reliability.

One example is OCP's Data Center Modular Hardware System (DC-MHS) specification, which establishes common designs and interfaces that manufacturers can use across different server platforms.

HPE chairs DC-MHS's plug-and-play workstream, making standardized modules compatible at the system level. Beyond physical and electrical connections, engineers must account for management interfaces and how components operate together.

JBK offered a practical example: HPE can reuse the DL525's host processor module in another single-socket configuration with different density requirements. Rather than develop an entirely new design, the company can build on validated components to create additional options for customers.

This flexibility extends to rack infrastructure. HPE supports both conventional EIA 19-inch racks and ORV3 configurations, giving operators a choice between familiar enterprise systems and open rack architectures designed for greater efficiency at scale.

HPE has been discussing these possibilities for years. "And now we're actually doing it," JBK said.

The significance is that HPE has put these specifications into practice across commercially available servers, so customers can benefit from modular designs and multiple deployment options without having to develop or integrate the underlying technology themselves, thereby accelerating their time to deployment.

Security and Management Complete the Platform

Modularity gives customers more flexibility, but it also introduces unique integration requirements. Manufacturers must ensure that security and management capabilities work consistently across standardized hardware components.

HPE draws on decades of experience in enterprise server management to make modular infrastructure easier to secure, monitor, and operate. Its Integrated Lights-Out (iLO) technology plays an important role in this approach.

As JBK explained, customers need the freedom to configure their infrastructure without compromising the ability to protect, monitor, and manage it.

ProLiant Gen13 addresses these operational priorities through several capabilities:

1. Hardware-based security: HPE incorporates two forms of silicon root of trust into ProLiant Gen13, providing a hardware foundation for verifying system integrity. It also controls and manufactures the silicon behind its root-of-trust implementation.
2. Post-quantum readiness: HPE introduced iLO 8 with ProLiant Gen13, including support for post-quantum cryptography (PQC). These protections extend to critical communications between system components, helping organizations prepare for evolving cryptographic threats.
3. Automated data protection: Self-encrypting drives and automatic key provisioning through iLO 8 help protect data from first boot, reducing manual configuration. HPE also provides multi-party authorization for sensitive operations, including key provisioning and secure erase, adding a safeguard against unauthorized changes.
4. Standardized management: HPE supports Redfish, the DMTF standard for hardware management, giving IT teams consistent interfaces for monitoring and managing different server configurations. HPE also participates in the standard's development.
5. Simplified operations: HPE Compute Ops Management helps IT teams identify affected systems, assess advisories, and guide remediation, simplifying infrastructure management at scale.

What I Took Away From Our Conversation

I left my conversation with JBK with three clear takeaways.

1. Enterprise AI requires fit-for-purpose infrastructure. There's no single ideal server configuration for every application, especially when organizations must account for existing power, cooling, and space constraints. ProLiant Gen13 offers distinct options for different processing demands and deployment environments.
2. Open standards provide greater hardware flexibility. As infrastructure requirements evolve, organizations need the freedom to adapt their systems without unnecessary complexity or dependence on proprietary designs. HPE's work with DC-MHS, ORV3, and Redfish shows how modular architectures and standardized interfaces can expand configuration choices and simplify development.
3. Security and management must keep pace with hardware flexibility. As organizations deploy AI, they need consistent protection and oversight across their infrastructure. Silicon root of trust, iLO 8, and automated data protection help enterprises maintain operational control across different server configurations.
4. ‍

What stood out to me most was HPE's ability to bring these capabilities together in commercially available systems. ProLiant Gen13 combines the modularity and efficiency associated with hyperscale infrastructure with the reliability and lifecycle management enterprises expect from HPE.

As AI moves further into production, organizations need infrastructure that fits their workloads, facilities, and operational priorities. HPE's approach gives them more ways to meet those needs without sacrificing security or manageability.

To hear more from JBK, watch our full conversation from OCP Global Summit and explore the HPE ProLiant Compute portfolio.

‍