There's about 40 minutes to go until Antonio Neri, CEO of HPE, kicks off Discover 2026 with his opening keynote. I've already seen massive crowds heading in the direction of the auditorium, so I'm expecting a pretty packed audience.
Even busier than I thought!
The tannoy is promising that the program will begin "shortly". In the meantime we're being treated to some dramatic music...
And we're off. "What does it mean to be an architect?" asks the VO. Apparently, now that we're in a new AI era anyone can be an architect. I think the dictionary and architects in general might have something to say about this...
Apparently networks are the foundation of all this, which I predicted would be one of the big messages for the event. Expect to hear even more about this over the next 48 hours, not just in this first keynote.
Today, he says, we're in a new, shifting technology landscape, where bu=oth humans and AI are creating inputs.
You don't architect a building round one room, he says. For creating AI and agentic AI systems it's the same, you need a solid foundation. For AI, that's networking but also other HPE elements like Private Cloud.
For years, Neri says, Aruba (which has been part of HPE for over a decade) the company has provided networking in terms of wireless access etc. Witht he acquisition of Juniper, that has now come into the data center.
Neri has just mentioned "self driving" networks. We're being shown a recent ad for HPE featuring F1 drivers for Mercedes Petronas who are confused that this is about their vehicles not the network. And relieved that it's not – HPE, of course, is one of the sponsors of MErcedes F1.
The agentic AI era doesn't just need network, though, it needs compute as well. As AI moves from creating content to agents, there's greater pressure on compute.
HPE is ansering this with the release of Proliant DL394 Gen12, which features Nvidia chips.
The most advanced systems are only as good as the data that they can access, though, he says. Naturally, HPE has an answer for that, too, which is the new Alletra Storage MP X10000.
Next up: rising virtualization costs. HPE has had a virtualization play for about a year now, maybe two (excuse the fuzzy memory). HPE Morpheus VM Essentials was introduced at HPE Discover Barcelona either 2025 or 2024 (I will clarify this later) as a competitor to VMware, even if it hasn't been outright pitched as such.
I have missed a segment that I believe was about security – HPE acquired Zerto in 2021 – but Neri is now talking about how all of this is brought together in GreenLake, the company's private cloud "platform".
GreenLake intelligence, which is already part of Aruba Central acocridng to the new VO, and helps manage and operate agentic AI. Greenlake intelligence combines context and AI, turing "observation into understanding" and recommending or carryign out a remediation.
It's never a surprise to hear about GreenLake at HPE Discover, though it's shifted guises over the years. While it was originally pitched as a private cloud offering, it's now effectively both a management platform. I believe that at least some of the "consumption based IT" idea from the original iteration of GreenLake has been offloaded to HPE Financial Services.
Right: Neri is going to elaborate on two parts of the company's strategy, networking and "managing transformation into an agentic enterprise".
We're now watching a short Video about the company's partnership with Vultr, with a person on screen who I am 99% sure is generative AI. I suppose this is on topic for the theme of this morning.
Right, new announcement. HPE Juniper NEtworking now has the QFX5252 scale-up switch for AMD Helios. It's 100% liquid cooled, has open standards, and SONIC with AI-native operations.
In terms of specs, it has 1024 Tbps total capacity and comes with Juno operating system.
Neri is now talking about how AI data centers are now often not a single data center, they are several sites working together. This is where HPE edge offerings come in, he says.
The Juniper Networking MX301 can help with this. There's also a new switch, the QFX5140 switch for inference clusters. 16 Tbps total capacits un 1RU, optimized for inference and AI-native optimized with HPE Marvis, says Neri and the specs list next to him
New networking announcement: Aruba CX Switch is coming to Mist and Marvis actions is coming to Aruba Central. This is part of the company's "cross-pollination" efforts.
You can read more about that announcement here.
Another annoucnement, this is about agentic AI and updates to HPE AI Factory.
This includes: secure agentic operations, which mean agents only run in their own secure environments with guard rails. There's also Nvidia NemoClaw, which is an open framework. Also, Zerto is being included that includes data recovery if something does go wrong.
For a more coherent write up and more information, read more here.
Also coming to HPE AI Factory is HPE Alletra Storage MP X10000, which allows agents better access to data and metadata.
But wait, there's more: New inference capabilities, with multi-node inference, unified gateway, and enhanced conference, and shared KV cache.
HPE Unleash AI partner ecosystem, which has over 60 partners like CrowdStrike and Fortinix, and Protopia, also continues to grow says Neri, though there's no rate or previous partner figures for comparison.
apparently the Ryder Cup is using HPE kit for digital twins. No real information on that, but everyone loves a digital twin these days.
We're getting encouragement to come along to the Day 2 keynote, with Fidelma Russo tomorrow – it will be the third keynote of the event. Naturally, your intrepid reporter will be there.
Ok, another announcement: Confidential computing now comes as standard across HPE AI Factory with Nvidia
Ok., we're wrapping up. His three big messages to delegates, after thanking sponsors, are: Architect deliberately, start with the network and choose the right partner, which is to say HPE.
And it's over. Thank you for following along. Rami Rahim's keynote is in a couple of hours, so I will be back for that. But until then, goodbye!
We are back! It's 1pm in Las Vegas, the delegates are fed and the auditorium is once again full (though not heaving) with people ready to hear more about HPE's networking business. Rami Rahim, formerly CEO of Juniper Networks, should be taking to the stage any time now.
We are starting off with a video about self-driving networks, which can self optimize, self protect, and self heal.
Rahim is on stage and receives an extremely warm welcome. He's telling us a cautionary tale of the Millennium Tower in San Francisco, which ended up slowly collapsing under its own weight because its foundations weren't strong enough. Readers who were following this morning probably knows where this story is going...
Organizations are building AI applications and workloads, but if they don't have a strong foundation built on networking then their projects will go the way of San Francisco's Millennium Tower.
The network is front and center again, says Rahim. It's not just sitting in a corner somewhere, it's fundamental and as important as any other part of the data center and IT infrastructure.
An "old fashioned" network that's static isn't good enough though. It's not just about building a network for AI, it's about building AI for the network. He hasn't said the phrase yet, but this is the idea of the self-driving network.
This is all lovely, but the network has to be protected as well, he says. And it has to do it without impacting user experience or increasing the burden on IT teams.
" The networks of the future must be able to sense, to learn, to optimize, to protect, and heal themselves in real time," says Rahim.
HPE, he says, has made more progress than anyone in this technology.
Here comes the first guest: CIO of Ohio State University Rob Lowden.
Apparently OSU has 22,000 wireless access points across its campus, both indoors and outdoors.
They're updating their network at 'The Horseshoe', its stadium, where thousands of people can descend on a match day. There's also a nearby cancer hospital run by the university that also has connectivity needs. So it's now working with HPE to overhaul their whole network to support that.
Lowden's off, Rahim is back to talking about how if the self driving network can support a scenario as complex as OSU, it can support pretty much any need. A bold claim.
Sunalini Sankhavaram from HPE Campus and Branch is now on stage to talk about delivering a consistent self-driving experience on both Aruba Central and Juniper Mist. Thanks to microservices, HPE can build once and deploy twice – like developing an app and deploying it on both iOS and Android.
"We have an API first approach," she says, which makes HPE's MCP servers and tools "extremely powerful, delivering objective automation at scale".
Rahim says "most vendors talk about AI systems and agents and AI ops, but still rely on reactive human-driven troubleshooting, but if you can still have to fix the problem, where exactly is the self-driving in that?" He asks her for an example of how real self-driving networks work and why they matter.
"We all know that today's networks need to cater to high density requirements while also meeting the performance expectations of every user and application, but most networks are not designed to meet intermittent surges in the traffic," says Sankhavaram.
Showing an example scenario, she continues: "If we look at Marvis, we see that all the users in this public building automatically have (good connectivity), but was this the case all of last week? No, it wasn't. So, what happened? How did unhappy users become happy? Let's take a look.
"Last week, Marvis detected that over 6% of user minutes were bad, which may not sound like a lot, but it impacted hundreds of people. Marvis has a self-driving action for dynamically fixing the path solution. This action was enabled and was able to autonomously fix the problem by enabling fuel bands like river. This reduced the peak utilization from 90% to 54% enabling a better experience for users that are lacking."
All without a human or any tickets raised.
We're now going over HPE Marvis Actions in HPE Aruba Central. It's not about just more visibility, it's about better experience she says. There's also the opportunity to automate many actions, such as noticing there is a problem with a CCTV camera feed, analysing what the process is and remediating it automatically.
We're retreading some of the same ground as this morning about the CX switches.
According to Rahim, it only took a few moths to get Aruba and Juniper integrated and innovating, which gets a big round of applause from the audience.
Rahim welcomes a new guest, this is the head of engineering and networks at the Royal Bank of Canada.
Okey dokey, we're now onto Zero Trust netowrks. There are 5 elements of security, visibility, policy driven, enforcement, detection and response, explains Madani Adjlali, VP of product management (SASE and Security), at HPE. HPE's security portfolio includes firewall, NAC, SSE and SD-WAN.
It also has a unified SASE and everything's been combined into a single SASE orchestrator.
Oh dear, it's live demo time. Fingers crossed...
Ok, it's going well so far with no hiccups. SASE Orchestrator will be available to all customers later in the year, we've just been told
live demo of the AI Firewall now.
Rami says goodbye to Adjali, concluding the segment by saying: "At HPE, AI, networking, and security are no longer separate domains. They are converging into a single intelligent self-driving system".
Now on stage, Katrina Pickett to talk about the role and importance of routers.
We're being taken through an example demo of how Marvis might help detect and remediate a network problem involving an increase in latency.
Marvis fixed the issue, which was convenient. In future, the network manager – or whoever is in charge of it – can give Marvis permissions to carry out the same fix automatically if the issue occurs again in the network.
We've had a brief customer appearance and now we're onto another product demo – Data Center Operations, which is also part of Mist.
"Today you heard a consistent message from every customer, every demo ... the old way of operating networks has reached its limits" says Rahim, closing out the keynote. "the future belongs to networks that can think, that can adapt, that can optimize, and that can protect themselves in real time.
"Self-driving networks, not a futuristic idea, not a lab experiment, but as a practical necessity for operating modern infrastructure at scale."