A 2U server purpose-built for the era of agentic AI. The NVIDIA® Vera CPU puts high core counts, low latency, and fast memory on the critical path of every reasoning loop — wrapped in the security, manageability, and lifecycle tooling enterprises already trust from HPE ProLiant.
The first wave of AI was about training and inference on GPUs. Agentic AI changes the math: models now plan tasks, call tools, run code, query data, and grade their own output — and most of that work runs on the CPU. When the CPU stalls, the whole agent loop waits. The DL394 Gen12 is engineered to keep those loops moving.
Running code, processing files, querying databases, and driving applications are CPU workloads. As agents act instead of just answering, that load climbs fast.
Agent-generated code runs in isolated sandboxes to be tested and validated before it ships. Spinning up and tearing down thousands of them is pure CPU and memory pressure.
Reinforcement learning loops continuously evaluate outputs, compute rewards, and feed results back — sequential work where slow CPU latency throttles total throughput.
A focused set of capabilities that matter when you are sizing infrastructure for agentic AI and reinforcement learning.
88 NVIDIA-designed Olympus cores for orchestration, runtimes, analytics, and agent workloads.
Each core runs two tasks at once with predictable performance for multitenant, high-concurrency AI.
LPDDR5X memory feeds data-intensive AI, analytics, and reinforcement-learning pipelines.
Arm®-compatible Olympus cores with NVIDIA Scalable Coherency Fabric (SCF) for efficient data movement.
HPE iLO 7 with silicon root of trust establishes a verified chain of trust from the moment you power on.
HPE OneView and HPE Compute Ops Management deliver AI-driven insight from data center to edge.
Unlike CPUs tuned purely for traditional virtualization, the Vera CPU is built for fast, responsive environments that launch, run, and evaluate thousands of concurrent tasks under sustained load.
Volatile-market analysis and faster trading decisions that depend on low-latency compute.
Agent loops that reason, act, and re-plan in line with live data instead of in batches.
Reward computation and environment evaluation that scale across many concurrent runs.
More accurate risk assessment and fraud detection on continuous streams of transactions.
Data-heavy research pipelines that pair CPU-centric and accelerated work in one loop.
Streamlining operations with agents that monitor, decide, and trigger actions in real time.
Vera CPU options, memory, storage, networking, GreenLake consumption, and lead times — there's a lot to get right before a PO. Tell us what your AI workloads need to do, and we'll land you on the right DL394 build, get you pricing that makes sense, and stay on the order until it's racked.
Enterprise AI needs more than raw throughput. The DL394 extends the security, manageability, and lifecycle tooling that IT teams already rely on across the ProLiant portfolio.
HPE iLO 7 with silicon root of trust verifies every layer of firmware and software, protecting against firmware attacks, unauthorized access, and tampering.
A centralized interface to manage and monitor servers, storage, and networking — automating and streamlining day-to-day IT operations.
AI-driven insight and automation from data center to edge: forecast energy costs, pinpoint issues, and manage a global server footprint from one place.
Agentic systems plan tasks, call tools, run code, query data, and evaluate their own results — activities that run primarily on the CPU. The DL394 Gen12 is tuned for fast, responsive environments that launch, run, and evaluate thousands of concurrent tasks with predictable performance under sustained load.
88 NVIDIA-designed Olympus CPU cores provide high-performance processing for orchestration, runtime environments, analytics, and agent workloads. Spatial multithreading lets each core run two tasks simultaneously while maintaining consistent performance. Up to 1.2 TB/s of LPDDR5X memory bandwidth feeds data-intensive AI and RL pipelines, while Arm®-compatible Olympus cores and the NVIDIA Scalable Coherency Fabric (SCF) provide strong single-thread performance and efficient data movement.
Security is built into the foundation with HPE iLO 7 and silicon root of trust, verifying every layer of firmware and software from the moment the server powers on. HPE OneView provides a centralized interface to manage and monitor servers, storage, and networking, while HPE Compute Ops Management adds AI-driven insights and automation from data center to edge — helping IT teams forecast energy costs, pinpoint problems, and manage a global server footprint.
Tell us about your agentic AI, RL, or analytics workloads and a ServerComputeWorks specialist will help you configure, price, and order the right system.
Talk to an Expert*NVIDIA and Vera are trademarks and/or registered trademarks of NVIDIA Corporation. Arm is a registered trademark of Arm Limited. All other third-party trademark(s) is/are property of their respective owner(s). Specifications and availability are subject to change.