The Buzz: HPE Discover keynote day: self-driving networks take centre stage as HPE makes its AI-era argument

This epsisode of In The Channel is brought to you by HPE Discover 2026. Check out our full coverage of the event on ChannelBuzz.ca — you’ll find out HPE Discover 2026 News Hub in the menu bar at the top of the page.

This episode of The Buzz is brought to you by HPE Discover 2026. HPE Discover runs June 15 to 18 at The Venetian in Las Vegas. Discover what’s next at hpe.com/discover.

Welcome to The Buzz from ChannelBuzz.ca, I’m Robert Dutt, today is Wedneday, June 17th, and here’s what’s happening in the channel today.

We covered news elsewhere in an earlier episode of the Buzz, go check that out if you haven’t already. For this one, we’re drilling down on Tuesday’s news from HPE Discover 2026.

We’re right in the middle of the week here, and I want to bring you the highlights from Tuesday – keynote day, the day HPE makes its biggest arguments. And the argument on Tuesday was pretty clear: the network – not the GPU, not the server – is the foundation of the AI era. They had product announcements to back it up. Here’s what went down.

Let’s start with the afternoon, because honestly, the networking general session led by Rami Rahim – who heads up HPE’s networking business as EVP and GM following the Juniper acquisition – was the meatiest part of the day for the channel.

The headline is what HPE is calling self-driving networks. The idea is that AI-driven networking should be able to sense, learn, optimize, and heal itself in real time, without requiring a human to manually troubleshoot every issue.

Rami opened with an analogy I thought landed pretty well. He talked about the Millennium Tower in San Francisco – the luxury condo building that started sinking after construction because the foundation wasn’t built for the environmental load it was sitting on. His point: companies that are building AI on top of networking infrastructure that wasn’t designed for it are making the same mistake. “AI innovation can only move as fast as the network allows” was the line. It’s a good one.

So what did they actually announce? Four things worth flagging.

First: Marvis AI cross-pollination. Mist’s Marvis AI engine is coming to the Aruba Central platform, and Aruba capabilities are moving the other way too. Both platforms get stronger. And the important subtext for the channel: neither platform is being sunset. HPE has been clear about that, and it’s worth saying out loud, because there’s been plenty of speculation since the Juniper deal closed.

Second: a unified SASE orchestrator. HPE is combining its SD-WAN and Secure Service Edge capabilities into a single console with a consistent zero trust policy layer across the enterprise. But the most interesting piece is what they’re calling the AI Firewall – the ability to classify your users’ GenAI applications as sanctioned, unsanctioned and blocked, or tolerated with guardrails like prompt filtering and data upload controls. They demoed it blocking a data exfiltration attempt through a GenAI app in real time. If you’re an MSP and your customers are asking you how they let people use AI tools without losing control of sensitive data, this is a concrete answer to that question.

Third: the QFX 5140. This is a new inference switch – new this week, not a prior announcement – purpose-built for distributed AI workloads at the edge. AI-optimized load balancing and congestion control, designed to connect GPUs at distributed locations. The edge inference angle is where this gets interesting for partners who are thinking about AI at branch or remote sites.

And fourth – and I want to make sure this doesn’t get buried – the HPE Network Migration Program. Zero percent financing through HPE Financial Services, plus asset trade-in for legacy non-self-driving gear. If you’ve got a customer sitting on aging campus or branch infrastructure and the refresh conversation has stalled, this is the conversation starter to go back with.

On proof points: Rami said that over 80 percent of network incidents are now either fully self-remediating or instantly identified with a resolution ready – up from around 50 percent just a few years ago. He had big customers on stage: Ohio State University, the Royal Bank of Canada, Sentara Health. The RBC quote was notable – security is now “job number one” and it has to be managed at the network layer for what they called immutable evidence. That framing works particularly well in regulated industries, which is a big part of the Canadian market.

In the press Q&A afterward, Rami was direct about where the security and networking story goes: “When we say network and security are coming together, it’s not a tagline – it’s an investment strategy.” He also acknowledged that getting customers to trust full network autonomy is an adoption curve – most start with what they call trusted actions, where the system recommends and the human approves, before moving to full automation. I actually think that’s a reassuring thing to say rather than a weakness – it matches how enterprise IT actually works.

Now let’s go back to the morning. CEO Antonio Neri’s keynote set the strategic context for everything Rami built on in the afternoon.

Neri’s frame for the whole show is what he’s calling the agentic enterprise – the shift from applications that respond to user inputs, to autonomous agents that reason across your data and take action. And his point is that infrastructure has to be built to handle that, because agents deployed at scale without IT governance become the new shadow IT problem. He used the phrase “shadow cost” – the risk of an AI-heavy workforce operating outside IT’s visibility and control. That’s a real and near-term problem for your customers, and MSPs are typically the ones who get called when it goes sideways.

The most concrete illustration he gave was GreenLake Intelligence. The example: a major internal announcement gets added to the corporate calendar. The system sees it, anticipates that a large portion of the workforce is about to jump on a video call simultaneously, and proactively prioritizes video traffic before the strain hits – based on historical telemetry, no human in the loop. It’s a small example but it makes the concept real in a way that “agentic infrastructure” as a term doesn’t always do.

In the press Q&A after the keynote, Neri was notably direct on a couple of things. On the Juniper integration, he put a specific number on it: from close of the deal on July 2nd last year, to fully integrated data centre switching, routing, and campus portfolios – five months. That’s a credible timeline, and it matters for partners who’ve been watching to see whether the deal delivers or whether it turns into the kind of slow-moving integration that disrupts customer relationships for years.

And on competitive positioning, he was unusually blunt. Asked about HPE’s networking vision going forward, he said HPE is – direct quote – “better than Cisco in many ways, whether it’s campus and branch.” That’s not something you hear a CEO say casually at a press Q&A.

Now, for the Canadian channel specifically, there’s a layer here that tends to get underplayed in the broader coverage of a show like this.

The conversation in Canada right now isn’t just “upgrade your network because AI needs faster pipes.” It’s “bring AI workloads back on-prem or to Canadian colocation, because you can’t let that data live in a US-based cloud under current conditions.” Data sovereignty is a genuine buying driver right now in a way it hasn’t been before. And HPE’s self-driving networks story, and the broader private cloud AI play, maps onto that buying driver in a way that’s worth having a direct conversation with your customers about.

I’ll have more on the Canadian channel perspective in an upcoming In The Channel episode coming later this week from HPE Discover. But the framing I’d leave you with is this: self-driving networks don’t eliminate the managed services partner – they change what that partner does. The network takes on more of the routine work, but someone still needs to watch the dashboard, make strategic decisions, and bring the human layer. That’s still your business, and if anything it’s a higher-value version of it.

One more thing before we go – and this one’s a little off the beaten path. Someone asked Antonio Neri in the press Q&A who he’s picking for the World Cup. Being Argentine, he said he’d love to see Argentina win again – but acknowledged it’s tougher with an extra game in the format this time around. His final four: England, France, Argentina, and Spain. No bias there whatsoever.

That’s how we’re seeing the headlines from HPE Discover. I’m Robert Dutt for ChannelBuzz.ca, thanks for listening. Have a great day.