HPE brings self-driving networks to AI factories, data centers, and edge

HPE has expanded its self-driving network strategy with new capabilities for data centers, campuses, edge environments, and AI factories. Announced at HPE Discover Las Vegas 2026, it reinforces a growing idea in enterprise infrastructure: AI does not scale solely with GPUs, servers, and storage. It also demands a network capable of low latency, high availability, continuous visibility, and automation enough to resolve issues before they impact workloads.

The company is integrating its HPE Juniper Networking portfolio more deeply into HPE AI Data Center Solutions and extending its AIOps agent capabilities over HPE Mist, HPE Aruba Central, and new SASE security platforms. The result is a proposition aiming to unify network, security, compute, and hybrid cloud within a single operational architecture, with increased automation and reduced manual tasks.

The underlying message is clear. As companies move from AI testing to real deployments, the network ceases to be a secondary layer. In inference clusters, AI factories, or distributed environments with agents, any bottleneck in traffic, latency, routing, or security can diminish infrastructure utilization. HPE wants the network to act as an intelligent layer, not just a set of devices checked only when failures occur.

Networks Designed for AI Workloads, Not Just Traditional Traffic

One of the most notable updates is the integration of HPE Networking into HPE AI Data Center Solution. The company incorporates HPE Juniper Networking QFX switches managed via HPE Networking Data Center Director to provide a preintegrated foundation of compute, networking, storage, software, and services.

The goal is to accelerate moving from experimentation to production in AI environments. Training and inference workloads do not behave like traditional enterprise applications. They require moving data between accelerators, servers, storage, and control systems with minimal delays, losses, or congestion.

HPE presents the QFX5140 as an option for inference and edge AI—areas expected to grow as AI moves beyond large centralized data centers and closer to industries, telecom networks, corporate campuses, or low-latency locations. The QFX5252 targets the AMD Helios platform, where the intra-rack network can determine how much GPU processing time versus data waiting time is allocated.

In AI, such distinctions matter. A costly GPU poorly fed by the network is capacity wasted. HPE’s promise is to reduce this downtime and improve infrastructure efficiency.

Agent-Based AIOps: Moving from Network Observation to Action

HPE is also advancing its integration of HPE Mist and HPE Aruba Central—two platforms from different origins that are now beginning to share capabilities. The company speaks of a “cross-pollination” strategy between Aruba and Juniper to unify hardware, consistent operations, and AI-driven network functions.

A key feature is the support for HPE Networking CX switches within HPE Mist. This enables wired access switching customers to benefit from native AI visibility, automated provisioning, layer 2 assurance, dynamic packet captures, service level indicators, and actions driven by HPE Marvis.

The next step is even more significant: HPE aims for the network not only to detect problems but also to reason about them and propose or execute remediation actions. The company is expanding its data center capabilities with proactive maintenance through predictive analytics and a reasoning agent capable of synthesizing telemetry data, millions of support cases, and a contextual graph from Data Center Director.

This reflects a shift in network operations. NetOps teams can no longer manually analyze all events, alarms, logs, and configuration changes in distributed infrastructures. AI can help identify patterns, anticipate optical or system failures, and accelerate root cause analysis. The challenge is ensuring these recommendations are reliable and do not generate excess noise.

Security and SASE: Integrated Strategies

HPE has also announced a unified SASE platform based on HPE Networking EdgeConnect and advanced firewall technology. It converges SD-WAN and Security Service Edge in a single management console, with a zero-trust approach and AI-assisted operations.

While the convergence of network and security is not new, it becomes more urgent with AI. Attackers also leverage automation to find vulnerabilities, run campaigns, and exploit weaknesses more rapidly. Meanwhile, organizations connect more users, devices, apps, agents, and services, expanding their attack surface.

HPE also emphasizes a “soberano” (sovereign) SASE foundation, where the combined SSE connector with Private Edge keeps traffic within the corporate boundary without always redirecting to external cloud PoPs. This can appeal to regulated sectors, government agencies, or companies with strict data residency and control requirements.

Security for a self-driving network cannot be an afterthought. If the network makes decisions, moves workloads, enforces policies, or connects agents, identity and access control must be embedded from the design stage.

Compute, Hybrid Cloud, and Network: Unified Experiences

Another significant announcement is the integration of HPE Mist Networking Data Center Assurance with HPE Compute Ops Management and HPE GreenLake. The goal is to reduce tool sprawl and provide a more unified experience across network, compute, and hybrid cloud environments.

While this may seem less glamorous than AI switches, it is practical. Many companies already juggle multiple consoles: one for networking, one for servers, virtualization, cloud, security, and observability. As AI demands faster operations, this fragmented approach does not scale well.

HPE envisions moving toward an autonomous data center where network, servers, and hybrid cloud share signals, recommendations, and actions. Not to replace human operators, but to reduce repetitive tasks and accelerate decisions in cases of degradation, saturation, or risk.

Juniper’s Impact Becomes Evident

The integration of the Juniper portfolio within HPE is becoming clearer. HPE is not just maintaining separate brands; it is incorporating Mist, Marvis, QFX, Data Center Director, and Juniper’s AIOps approach into a broader strategy that includes Aruba, GreenLake, compute, security, and AI solutions.

This was an expected move, but now it’s taking concrete shape. In campus and edge, HPE aims to combine its Aruba foundation with automation capabilities and user experience enhancements driven by Mist. In data centers, it seeks to strengthen with QFX switching and smarter operations. In security, it plans to unify SD-WAN, SSE, and zero trust under a single console.

Competition will be fierce. Cisco, Arista, Broadcom, NVIDIA, Juniper under HPE, and other players are vying to become the go-to network for AI. The differentiation will go beyond ports, speeds, or latency; it will be about operating networks with less complexity, integrating with compute and security, and demonstrating real-world performance in production.

Funding Becomes Part of the Strategy

HPE Financial Services launches a Network Migration Program to help organizations modernize their networks for AI readiness. The program combines hardware financing, 0% software financing, and a set of IT asset programs to unlock value from existing equipment.

Such initiatives are meaningful. Upgrading networks for AI, SASE, or edge environments requires investment, and many companies carry infrastructures that are operational but not optimized for new workloads. HPE aims to make this upgrade more financially manageable.

Redesigning networks often competes with budgets for servers, storage, cybersecurity, cloud, data, or AI initiatives. HPE seeks to directly connect this upgrade to the ROI of AI. The argument: reliable enterprise AI depends on a network that is ready.

Networks Are Moving Back to the Center of Infrastructure

For years, enterprise networks seemed like a stable, nearly invisible layer. AI is changing that perception. A distributed model, GPU clusters, an AI factory, or an agent-based environment requires a more predictable, observable, and secure network. If the network fails, AI fails.

HPE’s self-driving approach addresses this reality. It’s not just about automating configurations, but about creating a network that observes, reasons, anticipates, and integrates with compute, security, and hybrid cloud. It’s an ambitious vision, with success depending on the quality of operational data, model accuracy, and the confidence teams have in the recommended actions.

With the integration of Juniper, HPE has a clear opportunity to build a stronger network offering—at a time when AI infrastructure needs more than just bandwidth: it needs operational intelligence.

The question is no longer if companies will need advanced networks for AI, but whether they will be able to operate them without multiplying teams, tools, and complexity. HPE believes the answer lies in self-driving networks with agent-based AIOps, integrated security, and a common architecture from edge to AI factory.

Frequently Asked Questions

What did HPE announce at HPE Discover 2026?

HPE announced new self-driving networking capabilities for edge, campus, data centers, and AI factories, including QFX switches, integration with HPE Mist, new Marvis features, unified SASE, and deeper GreenLake connectivity.

What are self-driving networks?

Networks that leverage analytics, automation, and AI to detect issues, anticipate failures, recommend actions, and in some cases execute remediations with minimal human intervention.

What do the QFX5140 and QFX5252 switches bring?

The QFX5140 targets inference clusters and edge AI scenarios. The QFX5252 is a scale-up module for AMD Helios, offering low latency and high bandwidth for rack-scale AI infrastructure.

What role does HPE Marvis play?

HPE Marvis provides AIOps capabilities and self-driving actions for diagnosis, remediation, and operational improvement across wired, wireless, and data center networks.

Why does HPE combine SASE and self-driving networks?

Because network automation requires integrated security. The unified SASE platform combines SD-WAN, SSE, zero trust, and AI operations to secure users, devices, applications, and distributed resources.

via: hpe