Hewlett Packard Enterprise (HPE) is making a comprehensive bet on artificial intelligence, with CEO Antonio Neri using the opening keynote at HPE Discover 2026 in Las Vegas to outline a strategic vision that positions the company at the center of a fundamental shift in enterprise computing. According to Neri, AI agents are now running alongside end users in enterprise infrastructure, fundamentally changing how workloads move across networks and what compute and storage must deliver. The announcements spanning networking, compute, storage, agentic operations, and cloud management represent HPE's response to what Neri called one of the largest technology platform shifts in history.
The network as the AI foundation
For HPE, the network is the critical underpinning of any AI architecture. Neri emphasized that every byte, every token, and every decision across an AI workload must traverse the network, making it the first place where performance and scalability are determined. Much of his keynote focused on the integration and expansion of Juniper Networks technologies, which HPE acquired in 2024, as a cornerstone of its AI networking strategy.
HPE structured its AI network portfolio across four layers: scale-up within a single rack, scale-out across GPU clusters, data center interconnect, and edge inference routing. New QFX switches address the first two layers, providing high-bandwidth, low-latency connectivity for GPU-intensive environments. The PTX 12,000 routing platform handles data center interconnect with 800G routing capabilities, enabling fast data movement between geographically distributed AI factories. The SRX 4700 firewall delivers quantum-safe throughput at 1.44 Tbps in a single rack unit, while the MX 301 edge router brings the MX platform to the inference edge, leveraging Juniper's sixth-generation Trio silicon.
Neri highlighted the real-world implications of network latency at scale, noting that multiplying a small delay across hundreds of thousands of GPUs over weeks of training can mean the difference between training a new model in 90 days or 30 days. In the competitive landscape of AI development, that difference translates directly into the ability to achieve breakthroughs faster.
Additionally, HPE is integrating Marvis Actions into Aruba Central and Aruba CX switching into the HPE Mist portfolio, bringing AI-driven automation and observability to wired and wireless networks. This unification aims to simplify operations for enterprises managing hybrid work environments and increasingly complex agentic workflows.
Scaling compute for the agentic era
While networking connects systems, the actual compute horsepower must be organized and optimized for AI. HPE's compute portfolio is organized into three AI Factory tiers, catering to enterprise, service provider, and sovereign deployments. The new ProLiant DL 394 Gen 12 server is purpose-built for agentic AI and long-context workloads, addressing the growing need for handling inference across large models and complex knowledge graphs.
At the AI Factory at Scale tier, new configurations promise dramatic efficiency improvements: training with one-quarter the GPUs required by the previous Blackwell-generation platform, and inference at one-tenth the cost per million tokens. Private Cloud AI configurations now scale to 256 GPUs with multi-node inference support, enabled by a unified gateway that provides a single API for accessing both frontier and open-source models. A shared cache reduces the cost per first token, improving user experience and lowering operational expenses.
Neri emphasized that Private Cloud AI can now serve larger models across multiple systems with multi-node inference, so capacity grows with mathematical requirements rather than being constrained by single-server limits. This approach is designed to reduce execution risk and enable enterprises to move from ambition to outcome faster, accelerating time to token and ensuring environments are ready to perform from day one.
Storage: Making data ready for AI
Agents are only as capable as the data behind them, and HPE's storage strategy centers on the Alletra MPX 10,000, which now serves as the storage layer for Private Cloud AI. This system unifies file and object storage on a single architecture, adding real-time metadata enrichment and native MCP (Multimodal Content Processing) support. This enables agents to retrieve data across structured and unstructured sources, which is critical for tasks that require reasoning over documents, images, and databases simultaneously.
HPE claims that the Alletra MPX 10,000 delivers 7 to 12 times faster time to value compared to custom-built environments, largely because it eliminates the need for months of building custom AI data pipelines. Neri noted that traditionally, preparing data for AI required custom preparation for every use case, but the new unified architecture automates much of that work. The system also carries Nvidia Certified Storage validation, ensuring compatibility with leading GPU platforms and frameworks.
The storage improvements come at a critical time when enterprises are grappling with data sprawl. As AI agents proliferate, the ability to store, retrieve, and process data efficiently becomes a key differentiator between successful implementations and stalled projects.
Toward an agentic enterprise
The most transformative aspect of HPE's announcements is the governed agent layer built into Private Cloud AI. Neri described how agents now reason across data, applications, models, and workflows, helping users make decisions, automate processes, and increasingly take action on their behalf. However, the rapid proliferation of AI agents across enterprises often occurs outside formal IT oversight, creating governance and scale challenges that traditional IT management tools were not built to handle.
HPE's solution includes zero-code agent registration for agents built in any framework, with security controls on API calls, identity, and encryption applied without requiring code changes. A three-tier identity model verifies the user, governs the agent's permissions, and requires human approval for sensitive actions. This is backed by Nvidia Open Shell for isolated policy-enforced agent runtimes, NeMo Cloud for governed workflow blueprints, and Zerto for clean-state rollback when agents make errors, ensuring that agent failures do not corrupt enterprise data or processes.
Agentic governance addresses a growing concern among enterprise IT leaders: how to enable innovation with AI agents while maintaining compliance, security, and reliability. HPE's framework aims to balance flexibility with control, allowing developers to build agents rapidly while enterprise policies enforce boundaries.
Cloud and ecosystem expansion
HPE CloudOps consolidates virtualization, data protection, and cloud management into a single hybrid operating layer, simplifying the management of on-premises, edge, and public cloud resources. The Unleash AI program, which covers more than 60 validated partners, aims to accelerate the deployment of AI solutions by providing pre-tested integrations with leading software, hardware, and service providers.
Neri also addressed the power constraints facing the AI industry. He warned that the U.S. faces a 19-gigawatt power gap by 2028, with data centers projected to account for nearly half of U.S. electricity demand through 2031. HPE is investing in power-efficient hardware and cooling technologies, recognizing that as AI scales, the future will be defined not only by compute but by how efficiently enterprises can power, cool, and connect their AI infrastructure.
The announcements at HPE Discover 2026 make clear that HPE is positioning itself as a full-stack AI infrastructure provider, leveraging the Juniper acquisition and its own heritage in compute and storage to deliver integrated solutions. Neri's keynote provided a roadmap for enterprises navigating the shift from end-user-driven applications to agent-driven workloads, emphasizing the need for a network-first, scale-out architecture that can handle the unique demands of AI agents.
Source:Network World News