At COMPUTEX 2026, Hewlett Packard Enterprise (HPE) (NYSE: HPE) announced the expansion of its server portfolio with the launch of the HPE ProLiant Compute DL394 Gen12, a new CPU server powered by NVIDIA’s Vera CPU architecture. The company said the platform has been purpose-built to address the growing computational demands of agentic AI, high-performance data processing, and latency-sensitive enterprise workloads.
The new server was introduced as part of a collaboration involving HPE, NVIDIA, and Redpanda, with the technology also being explored by the New York Stock Exchange (NYSE) to support large-scale market infrastructure operations.
According to HPE, the HPE ProLiant Compute DL394 Gen12 has been engineered specifically for emerging agentic AI applications, which require real-time reasoning, rapid data processing, and highly responsive infrastructure. The company said the platform combines high CPU performance, increased memory bandwidth, low latency, and enterprise-grade security to support next-generation AI deployments.
“The shift from generative models to agentic systems is redefining the role of compute across the enterprise,” said Antonio Neri, president and CEO of HPE. He noted that organizations increasingly require infrastructure capable of handling real-time decision-making and reasoning workloads across AI and financial services applications.
The server is based on NVIDIA Vera CPUs and adopts a monolithic processor design rather than conventional high-core-count chiplet architectures. HPE stated that this approach helps avoid non-uniform memory access (NUMA) challenges commonly associated with multi-processor environments, where latency variations can impact performance consistency.
A key feature of the platform is its use of LPDDR5X memory technology, a low-power form of dynamic random-access memory (DRAM). According to HPE, the server delivers an aggregate memory bandwidth of 1.2 terabytes per second, equivalent to up to 14 gigabytes per second per core. This architecture is intended to enable faster ingestion and processing of large datasets while improving resource utilization for AI workloads.
NVIDIA positioned Vera as a CPU architecture developed specifically to support AI-centric computing environments.
“Agentic AI has arrived, and it needs a new CPU,” said Jensen Huang, founder and CEO of NVIDIA. He stated that Vera was designed to orchestrate AI factories and deliver improved efficiency and faster task completion compared with traditional x86-based systems.
The technology is also being evaluated within financial market infrastructure. Lynn Martin, President of NYSE Group, said the exchange processes more than 1.1 trillion messages each day and is exploring the use of NVIDIA Vera CPUs, in collaboration with Redpanda and HPE, to further optimize latency, throughput, and system reliability.
Beyond performance enhancements, HPE emphasized the server’s security architecture. The HPE ProLiant Compute DL394 Gen12 incorporates the company’s Silicon Root of Trust technology and integrates HPE Integrated Lights-Out (iLO) 7 management capabilities. The platform also includes a secure enclave architecture designed to protect systems throughout their operational lifecycle.
According to HPE, the new generation of ProLiant servers is the first to meet the National Institute of Standards and Technology’s quantum-resistant security requirements, a development aimed at helping organizations prepare for future cybersecurity challenges and safeguard sensitive workloads in regulated industries.
The server also integrates HPE Compute Ops Management, a cloud-based management platform that provides centralized visibility and automation across distributed server environments. HPE said the solution leverages AI-driven operational insights to simplify infrastructure management, reduce administrative effort, and help minimize downtime-related business disruptions.
