Hewlett Packard Enterprise's new ProLiant server, built around Nvidia's Vera CPU, marks the first major enterprise infrastructure play for agentic AI — a market Nvidia says is worth $200 billion.
Back
HPE launches Vera-powered server targeting $200B agentic AI market
Hewlett Packard Enterprise's new ProLiant server, built around Nvidia's Vera CPU, marks the first major enterprise infrastructure play for agentic AI — a market Nvidia says is worth $200 billion.
Hewlett Packard Enterprise on Monday introduced the HPE ProLiant Compute DL394 Gen12, a 2U server powered by Nvidia's Vera CPU that the companies say is purpose-built for agentic AI workloads. The server, unveiled at Computex in Taipei, delivers 1.2 TB/s of aggregate memory bandwidth using LPDDR5X — up to 14 GB/s per core — addressing the memory bottleneck that has constrained AI inference on CPU-based systems.
"The shift from generative models to agentic systems is redefining the role of compute across the enterprise," Antonio Neri, president and chief executive officer of HPE, said in a statement. "These workloads require high-performance servers with exceptional CPU performance to enable real-time reasoning across agentic AI and financial services applications."
The Vera CPU uses a monolithic design rather than the high-core-count chiplet architectures common in x86 processors, which suffer from non-uniform memory access issues that create variable latencies. By integrating LPDDR5X memory directly, the system achieves deterministic performance — a requirement for financial services where microsecond delays carry real costs. The New York Stock Exchange, which processes more than 1.1 trillion messages daily, is exploring the platform in collaboration with HPE, Nvidia and data streaming company Redpanda.
"Agentic AI has arrived, and it needs a new CPU," Jensen Huang, founder and chief executive officer of Nvidia, said. "Vera was built to orchestrate AI factories — delivering 2x the efficiency and faster task completion than x86."
Why Vera's architecture matters for AI workloads
Nvidia's Vera CPU represents a departure from the company's traditional GPU-centric strategy. Unlike x86 server chips that rely on chiplet designs with multiple dies connected via interconnects, Vera uses a single monolithic die with integrated LPDDR5X memory controllers. This eliminates NUMA-related latency variability, which becomes critical when AI agents must reason and respond in real time.
The 1.2 TB/s memory bandwidth compares with roughly 500-700 GB/s on current-generation x86 server processors from Intel and Advanced Micro Devices, according to published specifications. For agentic AI workloads — where models must retrieve context, reason across multiple data sources and generate responses sequentially — memory bandwidth often determines throughput more than raw compute clock speed.
HPE said the server will be available in fall 2026 as part of its Nvidia AI Computing by HPE portfolio. The company also offers financing through its 90/9 Advantage program, which defers payments for 90 days followed by nine months at 1%.
Security and management as competitive differentiators
The DL394 Gen12 is the first HPE server to meet NIST's quantum-computing resistant security requirements, the company said. It embeds security at the firmware level through HPE's Silicon Root of Trust and uses the secure enclave in its Integrated Lights-Out 7 management controller to protect servers across their lifecycle.
HPE Compute Ops Management provides a unified dashboard for managing distributed server environments, using AI-driven operations to reduce management time and minimize revenue loss from downtime, the company said.
The competitive stakes are high. Nvidia's Vera CPU opens a new front in the company's rivalry with Intel and AMD, which dominate the $200 billion server CPU market. If Vera delivers the 2x efficiency gain Huang claimed, it could erode x86's hold on enterprise data centers — particularly for AI inference, where latency and memory bandwidth matter more than raw core count.
HPE shares rose 1.8% in pre-market trading Monday. Nvidia shares, which have gained more than 80% over the past year, were little changed. The server's fall 2026 availability means any revenue contribution is at least two quarters away, but the NYSE pilot program provides a marquee reference customer that could accelerate enterprise adoption.
This article is for informational purposes only and does not constitute investment advice.
[1] HPE Introduces CPU Server with NVIDIA Vera CPU, Purpose-built for Agentic AI[2] MediaTek Collaborates with NVIDIA on the RTX Spark to Power the Next Wave of Windows PC Experiences[3] Nvidia launches RTX Spark, an ARM-based super chip for Windows PC[4] Nvidia Shares Dropped After Stellar Earnings. Is This a Sign of What's Coming for Artificial Intelligence (AI) Stocks?
