Travel report from HPE Partner Growth and HPE Discover, Las Vegas

HPE Partner Growth and HPE Discover in Las Vegas covered a lot of ground this year, but one common thread ran through most of the agenda: how AI is changing what we demand from the network. With more than 10,000 attendees and 225+ sessions, HPE’s message was clear: it is the network, not just compute, that determines which organisations can actually run AI agents at scale. From real time data to automation, from security to customer experience. Everything pointed towards the same concept: the Self-Driving Network.

Self-Driving Network: a cliché that has taken real steps forward

Rami Rahim, Executive Vice President, President, & General Manager for HPE Networking, was candid about this in his session: “Self-Driving Network” sounds like a cliché. It is a term the industry has used for years, but it is where HPE’s solutions are well on their way to getting. His point was that a great deal has changed in recent years. HPE has taken major steps, and that is where we are heading! It is worth being honest about what this means in practice today, and what is still a vision. For now, it is mostly about AI learning from real time data and being able to anticipate problems before they occur, and about the network in several ways being able to repair and optimise itself without any form of supervision.

The foundation: everything starts with real time data

Rahim was clear that the journey towards a self-driving network starts in one place: real time data. Without it, AI has nothing to work with. It sounds simple, but it places significant demands on how the infrastructure is built from the ground up. The network must see what is happening right now, not what happened yesterday.

Rahim summed this up in a simple model in three steps: learn, predict, heal. The network learns from the data it sees, anticipates problems before they occur, and can increasingly fix them itself. This is the engine behind the rest of the Self-Driving Network message.

From two divisions to one, and a promise delivered faster than expected

HPE has merged its Aruba and Mist divisions into a single AI Networking division. The background goes back to Barcelona in 2025, where HPE promised to take the best of the two platforms and integrate them. Six months later, they have delivered on that promise, faster than expected. Central and Mist are built on a microservices architecture, which means HPE can update both platforms in real time without affecting users, in the same way that apps on a smartphone are updated in the background without affecting the operating system.

From generative to agentic AI, and the proof on stage

A recurring theme in several sessions was the shift from generative to agentic AI. Where generative AI answers what you ask it, agentic AI is about systems that can act on their own, make decisions and carry out several steps of a task without a human having to follow every step. It is precisely this shift that lies behind much of what HPE showcased on the networking side, and it is the network, above all, that has to handle the new demands this creates.

The concrete delivery on this came when Rami Rahim showed how Marvis, the core technology in Juniper’s Mist platform, is now being extended into Aruba Central. Rahim himself described it as delivering on a promise made when the Juniper acquisition closed in July the previous year, saying that a year ago they had promised to bring together the best of Juniper and Aruba, and that they are now delivering exactly that.

The proof also came as a live demo on stage. Rahim showed how Marvis detected that more than 6 per cent of user minutes in a public venue were of reduced quality, found the root cause in radio frequency congestion, adjusted the channel bands automatically, and reduced peak load from 90 to 54 per cent, without a single ticket being raised with support. This is exactly what learn/predict/heal looks like in practice, shown live, not just described in a presentation.

Security must be built in, not added afterwards

Servers, storage, networking, software and security were highlighted as HPE’s five core pillars for AI. Of these, security deserves a little extra attention: the message was that security should not be something you add afterwards, but something that is designed in from the start.

This has a concrete consequence for the network: it becomes an active security layer in its own right, not just a passive transport route for data. Because the network already sees all traffic in real time, the same data that drives the Self-Driving Network, it can also detect anomalies and respond to them on the spot.

It can all be summed up in three phrases that came up again and again during the conference: the network must see more, react, and do more. Seeing more is about real time data. Reacting is about security and insight. Doing more is about automation, a network that takes action itself, rather than simply alerting you that something needs to be done.

A self-driving network

A concrete example of how this works in practice is Marvis Mini. It continuously scans the network using fictitious, synthetic users, so the network can be tested even when there are no real users on it. Typical examples are airports, educational institutions and workplaces outside opening hours, places where the network would otherwise sit idle and untested until people actually turn up.

The idea is simple but effective: the network should work when the real users come online. Troubleshooting should not start at that point. It should already be done.

What does this mean for the Nordic market?

It is worth being honest about where we actually stand. In 2026, “Self-Driving Network” in practice means a network that learns from real time data, anticipates problems and can increasingly handle them itself, not a network that is fully autonomous. It is still a journey, not a finished product. For the Nordic market, it is not accurate to talk about a single, unified picture. The countries are at different stages when it comes to AI and implementation, and that is a nuance worth keeping in mind.

We all want a network that runs on AI: a network that sees more, reacts faster and does more. Quietly, in the background, while it delivers.