HPE has unleashed a sweeping array of new products and enhancements at its annual Discover event in Las Vegas, designed to help enterprises build, manage, and secure AI-infused networks from the core data center to the far edge. The announcements span next-generation switches, deeper artificial intelligence operations (AIOps) integration, a unified Secure Access Service Edge (SASE) platform, and expanded partnerships with Nvidia. Collectively, these moves aim to position HPE as the go-to vendor for the networking backbone required to support the exploding demand for AI inference, agentic AI, and high-performance computing.
New QFX Switches for AI Workloads
At the heart of the hardware rollout is the HPE Juniper Networking QFX5140 switch, a 1RU fixed-configuration data center switch purpose-built for AI inferencing and edge AI deployments. The QFX5140 offers a 16 terabit-per-second switching capacity and can be configured with 24 ports of 400G QSFP112, 8 ports of 800G OSFP800, and 2 SFP28 ports. This port flexibility allows enterprises to connect anything from older 10G gear to the latest 800G accelerators without forklift upgrades.
Critically, the QFX5140 supports RDMA over Converged Ethernet (RoCEv2), which enables direct GPU-to-GPU communication without involving the host CPU. Combined with advanced congestion management features such as Priority Flow Control (PFC) and Explicit Congestion Notification (ECN), the switch ensures low latency and zero packet loss – essential for training and inference workloads that rely on massive parallel processing. HPE CTO Fidelma Russo emphasized that these capabilities allow enterprises to run AI fabric, spine-leaf, and border leaf topologies efficiently, scaling from a few racks to hundreds.
Alongside the QFX5140, HPE introduced the QFX5252 module specifically designed for its AMD Helios turnkey package, which packs 72 GPUs per rack. The module integrates directly into HPE’s high-end AI factory blueprint, combining CPUs, GPUs, and open Ethernet networking into a unified platform that can be deployed quickly for large-scale training and high-volume inference.
The new switches fill out the mid-tier of the QFX family, which already includes the top-end QFX5240/QFX5250 (102T capacity) and the entry-level QFX5100 (100G). By offering a clear tiered portfolio, HPE gives customers the ability to choose the right balance of performance and cost for their specific AI workloads. The addition of these switches also signals a deeper integration of the networking portfolio acquired from Juniper Networks into HPE’s broader AI infrastructure offerings. HPE’s Data Center Director management platform now includes full support for the QFX family, providing a centralized view of all network components and simplifying troubleshooting across hybrid environments.
Mist AI and Marvis: Extending Autonomous Operations to the Data Center
Perhaps the most significant software news is the deepening integration between HPE’s Aruba portfolio and the Mist AI platform acquired from Juniper. HPE announced that it will unify the Mist AI natural-language engine with HPE Aruba Central, and vice versa, all powered by the core Marvis AI engine. Marvis collects telemetry and user state data from Juniper routers, switches, access points, firewalls, and applications to detect and resolve a broad range of enterprise networking problems.
A key component is Marvis Actions, an AI-based system that identifies and prioritizes network problem remediation. HPE president of networking Rami Rahim stated that Marvis Actions will be extended to Aruba Central by the end of the year, bringing proactive AIOps across wired, wireless, and SD-WAN domains. This means that autonomous remediation – such as port resets, policy adjustments, or traffic steering – will happen without human intervention, reducing mean time to resolution from hours to minutes.
In addition, HPE said it will integrate the HPE Aruba CX switching portfolio with HPE Mist, giving CX customers capabilities like AI-native visibility, zero-touch provisioning, wired assurance for Layer 2 access, and Marvis AI-driven support. This cross-portfolio integration is a direct response to the growing demand for a single pane of glass that spans campus, branch, and data center networks.
Within the data center itself, HPE expanded Mist’s capabilities to include predictive analytics for proactive maintenance. For example, Mist can now use AI/ML to predict potential optics failures before they cause outages. The system also introduces an advanced reasoning AI agent for high-confidence remediation. This agentic AI ingests millions of Technical Assistance Center (TAC) cases and a contextual graph database from HPE Networking Data Center Director to deliver precise root cause analysis inside the data center. Rami described this as “Marvis AI engine for data center operations,” combining telemetry, application flows, operational context, and historical knowledge to diagnose issues rapidly.
The result is a “self-driving network” that extends from campus to core. Gartner senior director analyst Mike Leibovitz noted that this momentum is critical for HPE as it leverages the core Mist platform by expanding Marvis across the entire portfolio. “Agentic NetOps is the most exciting area of innovation in enterprise networking in more than 20 years,” Leibovitz said. He added that while most enterprises start this journey with their existing vendor, leadership in the space is still up for grabs.
Unified SASE Orchestrator
HPE also addressed the growing complexity of hybrid work and cloud connectivity with a new SASE Orchestrator. The package ties together HPE’s SD-WAN and SSE (Secure Service Edge) capabilities with a unified policy engine that uses AI to manage branch, remote user, and cloud connectivity from a single console. With the policy engine, customers can define security policies once and deploy them consistently across all sites, reducing configuration errors and speeding zero-trust adoption.
The Orchestrator promises intelligent traffic steering based on application awareness, ensuring that critical AI workloads get priority bandwidth and low latency. As Rami put it, the platform delivers “simpler operations with AI operations, faster zero-trust adoption, and a better user experience through intelligent traffic steering and application awareness.” For organizations that run AI inference at the edge or in multi-cloud environments, the unified SASE Orchestrator offers a way to enforce consistent security and performance policies without running separate management tools.
Nvidia Integration Deepens
HPE also announced tighter integration with Nvidia, a key partner for its AI factory offerings. The HPE Private Cloud AI platform – a turnkey AI factory co-engineered with Nvidia – is being enhanced with support for Nvidia’s Agent Toolkit software, including Nemotron open models, NemoClaw, and OpenShell secure runtime. This gives enterprises an agent operating system that can reason, monitor agent behavior, enforce policies, and reduce deployment risk. Effectively, it turns the AI factory into a managed environment where AI agents can be deployed, governed, and scaled safely.
Additionally, HPE is integrating Nvidia Confidential Computing into the AI factory through HPE Services. Confidential Computing protects models and private data during execution, which is critical for on-premises or sovereign deployments where data cannot leave the premises. By encrypting data in use, HPE and Nvidia aim to address concerns about data leakage during AI training and inference, especially in regulated industries like healthcare and finance.
Zerto and Morpheus Updates
Finally, HPE updated its data protection and cloud management software. A new release of HPE Zerto Software can now identify rogue agent actions and rewind to a clean state using data protection. This is particularly important in agentic AI environments where autonomous agents might inadvertently introduce errors or security risks. The Private Cloud AI package also supports secure local agent registration, allowing customers to approve AI models, skills, and tools while adhering to centralized governance.
On the cloud management side, HPE Morpheus now offers a compelling incentive for customers looking to move away from VMware. HPE announced that any customer who purchases or owns HPE VM Essentials for one year will receive the first year of licenses for free, along with free Zerto migration licenses during that period. Zero-interest financing for HPE cloud ops software over three years further supports migration. Russo described this as a way to “mitigate the double-bubble cost problem” that customers face when switching platforms. By removing financial friction, HPE hopes to accelerate adoption of its unified cloud management suite, which manages VMs, containers, and cloud resources across multiple environments from a single control plane.
These announcements represent a significant step forward in HPE’s strategy to own the AI networking stack. From the high-performance QFX5140 switch to the agentic NetOps of Marvis and the unified SASE Orchestrator, the company is delivering a vision where networks become intelligent, self-healing, and automated – essential for the next wave of AI-driven business.
Source:Network World News