How networking is shaping AI

At HPE Discover 2026 in Las Vegas, there were two key messages:

• If you’re serious about AI, you need to start with your network. We’ve heard variations of this before, but this time it felt different - less like a headline, and more like a reality check. As organisations push forward with AI, it’s becoming obvious that the network isn’t just supporting it, it’s shaping what’s actually possible.
• Secondly, and kudos to the teams involved, the seriously speedy development that’s taken place to integrate and cross pollinate Aruba and Juniper, not just between Mist and Central but dual platform hardware. They’ve brought Marvis Actions with SLEs into Central, CX Switching into Mist, adding further telemetry capabilities and dual platform APs, giving more choice to customers. It’s clear that regardless of the platform selected, the customer will not be missing out.

The network is no longer “just connectivity”

As a Network Architect, (and previously an Engineer), the network itself has always meant more to me than “just connectivity.” It’s great to see that mindset being backed up by how HPE has positioned the network. Not as a plumbing layer, and not just as critical infrastructure but as both the enabler of AI and most importantly, the constraint if the network isn’t ready.

We’re moving away from thinking about networks in isolation, and towards something much more integrated; where networking, compute, storage and cloud are all working together as a single system, underpinned by AI-driven operations.

Self-driving networks are here

One of the more interesting moments for me was how far HPE has taken the idea of self-driving networks beyond a simple vision. It’s become something tangible, particularly with the continued evolution of Mist and Marvis across campus, edge and into the data centre and with Marvis Actions coming to Central.

What’s changing isn’t just the tech, it’s the operating model. We’re moving from watching dashboards and reacting to alerts and manually fixing issues to something that can genuinely detect problems early, understand what’s going on, and in some cases, fix it automatically!

From insights to action (finally), this was probably the most important shift for me. We’ve had AIOps platforms for a while now, but they’ve mostly been about giving us better visibility. HPE is now demonstrating something more complete with its agentic AIOps model, where, in simple terms, the workflow becomes - sense what’s happening across the network, then reason using AI, consider taking action automatically and learn from the outcome.

The key difference is that last step - action!

Instead of raising a ticket or sending an alert, the network can actually step in and resolve issues before users even notice, that’s a big deal operationally.

HPE presented several self-driving demos for issues such as VLAN mismatch or stuck port scenarios, but one demo that stuck with me was dynamic capacity optimisation. Rather than an engineer spotting high utilisation and tweaking RF settings, the system detects load pressure on APs, works out the best response within policy, adjusts things (like moving to dual 5 GHz radios) then checks if user experience actually improved. It sounds simple, but it highlights the bigger point, this kind of tuning stops being a manual task and becomes part of a closed-loop system.

SASE orchestration

As expected, there was also a big focus on SASE and security. HPE demonstrated its Unified SASE Orchestrator, bringing SD-WAN, SSE and zero-trust policy into a single operating model and control plane. It also introduced the AI Aware Firewall, demonstrating the capabilities that can help so many organisations with discovering AI usage, classifying it and inspecting prompts and outputs, subsequently allowing customers to control and audit AI use to ensure it protects the users and business.

Data centre networking has evolved

Another standout was how HPE is evolving data centre networking. Bringing Mist and Marvis into the fabric, they're shifting the conversation from device visibility to actual service and application experience. With capabilities like intent modelling, AI-driven root cause analysis and prioritised remediation actions. It feels like the data centre is finally getting the same level of operational intelligence we've seen developed on the campus side.

What does this all mean?

• The network is no longer just supporting infrastructure; it is a critical control layer for AI workloads and digital services.
• Operating models must evolve. Ticket-led, alert-driven and manually intensive approaches will struggle to keep pace with AI-driven environments.
• Networking, security and operations are converging within unified, AI-native platforms.
• Governance is as important as automation. As autonomy increases, HPE continues to emphasise policy control “guardrails”, auditability and human oversight.

If you would like to understand what this could mean for your environment, or explore how these technologies could deliver value for your organisation, your Softcat Account Team would be happy to start that conversation. If you’re a new customer, you can also reach out to our Sales team.

Gary Hawkins, Chief Technologist - Cloud and Datacentre, has also written a blog that gives a general overview of HPE Discover. You can read that here.