The HPE AI Factory with NVIDIA is a full-stack AI platform designed to help enterprises move from pilots to production-grade, agentic AI. It combines HPE’s infrastructure, storage, and data software with NVIDIA’s accelerated computing and AI software stack.

In practical terms, it gives you:

• Turnkey AI factory capabilities via HPE Private Cloud AI, co-engineered with NVIDIA, so you can deploy AI agents with built-in governance, observability, and efficiency.
• Agent-focused software through the NVIDIA Agent Toolkit (including NVIDIA Nemotron open models, NVIDIA NemoClaw, and NVIDIA OpenShell secure runtime) that acts as an “operating system” for AI agents—supporting reasoning, monitoring, policy enforcement, and risk reduction.
• Optimized compute with systems like HPE ProLiant Compute DL394 Gen12 using NVIDIA Vera CPUs for high-performance data processing and agentic AI workloads.
• Integrated security and governance across networking, servers, storage, and software, so you can run AI agents with more control and compliance.

Overall, the HPE AI Factory with NVIDIA is built to help organizations reimagine how they operationalize AI agents—moving beyond experimentation into secure, scalable, production environments.

HPE’s approach is to embed security and governance into every layer of the AI stack so you can move agents from development to production with more confidence.

Key capabilities include:

• Secure and governed agentic AI in HPE Private Cloud AI, giving you controls to approve AI models, skills, and tools through secure local agent registration while enforcing centralized governance and security policies.
• NVIDIA Agent Toolkit with NVIDIA OpenShell secure runtime, which lets you monitor agent behavior, enforce policies, and reduce deployment risk.
• HPE Zerto Software with new capabilities to detect rogue agent actions and use continuous data protection to “rewind” to a clean state if something goes wrong.
• NVIDIA Confidential Computing integrated into the HPE AI Factory, protecting models and private data during execution for on-premises or sovereign deployments. It establishes a chain of trust with cryptographic attestation and encryption at every stage—hardware, software, and datasets—supporting compliance with standards such as CMMC, NIST 800, STIG, and FIPS.
• NVIDIA BlueField and NVIDIA DOCA across the HPE AI Factory to enable zero-trust policy enforcement, runtime threat detection, and networking encryption while maintaining performance.

Together, these capabilities help you secure AI workloads and data, govern how agents behave, and align with regional or industry regulations without sacrificing scalability.

HPE focuses on both the data layer and the inference layer to help you get more value from your AI investment while managing cost and performance.

Data and pipeline optimization

• HPE Private Cloud AI helps convert unstructured data into AI-ready pipelines in minutes, improving inference efficiency.
• HPE Alletra Storage MP X10000 automatically applies metadata and governance policies to prepare data for AI applications. In benchmark testing, this configuration delivered up to a 20.4× improvement in time to first token (TTFT) for the NVIDIA Nemotron 70B model.
• HPE Data Fabric Software broadens data availability for agentic workflows by extending Model Context Protocol (MCP) support to Apache Airflow and introducing an enterprise AI inventory that enriches distributed data with metadata. A standalone HPE Data Fabric appliance on HPE ProLiant servers simplifies deployment.

Tokenomics and compute efficiency

• HPE Private Cloud AI is designed to improve prompt processing efficiency and can boost token throughput by up to 20× based on internal HPE benchmark data.
• Capabilities such as a unified model gateway for governed access to frontier models, active workload prioritization, and multi-node inferencing for up to 256 GPUs help you control token costs and maximize GPU utilization.
• Fine-tuning of pre-trained models, including NVIDIA Nemotron open models via NVIDIA NeMo, lets you securely adapt models to your enterprise data for more relevant, efficient agentic AI.

By combining optimized storage, data fabric, and GPU-accelerated infrastructure, HPE helps you rethink how you manage token usage, performance, and data readiness across your AI estate.