Hewlett Packard Enterprise(HPE) announced a major expansion of its NVIDIA-backed AI infrastructure portfolio, unveiling new systems designed to build secure, scalable AI factories and an advanced data center interconnect that boosts performance for AI workloads running across long distances or multi-cloud setups. HPE also introduced its first AI Factory Lab in the European Union, giving global customers a dedicated space to test and validate sovereign AI environments.
Antonio Neri, HPE’s president and CEO, said the company’s partnership with NVIDIA continues to strengthen the foundation for AI factories of all sizes, offering more performance options for a wider variety of workloads. He emphasized that the two companies are combining their strengths to deliver complete AI infrastructures that address the increasingly complex needs of enterprises.
NVIDIA founder and CEO Jensen Huang noted that every organization needs secure control over its AI production capabilities. He described the modern data center as evolving into an “AI factory,” designed to generate intelligence at scale. With HPE integrating NVIDIA’s full accelerated computing stack and Spectrum-X Ethernet networking, he said the two companies are establishing a blueprint for sovereign AI. The new AI Factory Lab, he added, gives customers a dedicated environment to safely convert data into value.
HPE and NVIDIA are opening the new AI Factory Lab in Grenoble, France. This facility allows customers to experiment with workload design on an air-cooled, sovereign AI setup using the latest versions of NVIDIA AI Enterprise (in government-ready configurations), HPE servers, HPE Juniper Networking PTX/MX routers, NVIDIA accelerated hardware, Spectrum-X networking, and HPE Alletra storage. Customers can evaluate performance on infrastructure physically located in the EU, supporting data sovereignty requirements and meeting strict regional regulations. HPE is also teaming up with Carbon3.ai to establish a Private AI Lab in London, based on HPE Private Cloud AI and NVIDIA AI Enterprise tools, to help UK companies accelerate AI adoption.
With European organizations placing greater emphasis on operational and data sovereignty, HPE has upgraded its Private Cloud AI offering with new configurations and capabilities. Customers now have access to NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs alongside NVIDIA Hopper-based options, as well as support for NVIDIA MIG-based GPU fractionalization to improve utilization and reduce costs. Secure, STIG-hardened and FIPS-enabled deployments of NVIDIA AI Enterprise are now available for air-gapped environments. Newly introduced Datacenter Ops Agents from HPE, NVIDIA, and World Wide Technology simplify AI data center operations across hybrid and agentic AI environments.
HPE has also expanded its sovereign AI factory roadmap with system designs that incorporate region-specific compliance requirements. These solutions offer adaptable security controls, validated reference architectures, and software protections suitable for regulated industries. HPE’s cybersecurity practice adds further support through advisory and management services, including sovereign-compliant solutions built with NVIDIA technologies.
Networking for HPE’s AI factory offerings now includes the NVIDIA Spectrum-X Ethernet platform and NVIDIA BlueField-3 DPUs, enabling accelerated connectivity across a range of production environments—from data center to data center, and data center to cloud. The portfolio also integrates HPE Juniper Networking, using MX and PTX routing systems to deliver secure, high-capacity, low-latency connections for distributed AI clusters and multi-cloud workloads.
To help businesses handle expanding AI demands, HPE also introduced the Alletra Storage MP X10000 Data Intelligence Nodes. This next-generation architecture transforms the X10000 into an active data layer equipped with NVIDIA accelerated computing. It analyzes and enriches data in real time as it moves through the pipeline, using HPE Data Fabric to unify access across edge, core, and cloud environments.
By applying NVIDIA’s AI Data Platform reference design and running NVIDIA AI Enterprise software directly in the data path, the X10000 processes incoming data instantly, detecting patterns and preparing it for AI workflows. It supports S3-compatible storage with RDMA and NVIDIA acceleration, optimizing data classification and throughput.
HPE also revealed the NVIDIA GB200 NVL4 by HPE, a compact and energy-efficient system built for high-performance AI inference and generative AI workloads. Combining two NVIDIA Grace CPUs with four NVIDIA Blackwell GPUs, it allows up to 136 GPUs per rack, offering a powerful yet cost-efficient option for organizations that need significant AI performance without large power demands.
In the security space, HPE has chosen CrowdStrike as a key protection partner for Private Cloud AI customers. The companies are extending their collaboration to secure end-to-end AI environments—including NVIDIA-accelerated LLM workloads—by unifying identity, endpoint, cloud, and data security across hybrid and multicloud deployments.
HPE is also partnering with Fortanix to support NVIDIA Confidential Computing. Using Fortanix Armet AI, enterprises can deploy secure, sovereign agentic AI systems in regulated environments. These capabilities can be used on HPE Private Cloud AI and HPE ProLiant DL380a Gen12 servers with NVIDIA RTX PRO 6000 Blackwell GPUs, offering secure and scalable AI infrastructure for customers across EMEA and globally.
HPE continues to position itself as a leader in essential enterprise technology, bringing together AI, cloud, and networking to help organizations operate more efficiently, manage data intelligently, and achieve greater business impact. HPE (NYSE: HPE) is a global technology leader delivering solutions across AI, cloud, and networking to help organizations operate more intelligently, convert data into actionable insights, and maximize impact. The company’s long-standing expertise empowers customers worldwide to pursue bold goals and redefine what’s possible.
