BARCELONA – Hewlett Packard Enterprise (HPE) has kicked off the European leg of its Discover tour with further Juniper integration across its network stack, and its biggest play yet for the virtual machine (VM) segment.

CEO Antonio Neri acknowledged event host Barcelona by drawing comparisons with its city planning, building out a “a forward looking network built to handle growth” when it was bursting at the seams in the 19th century, becoming “a city designed to anticipate the future, not to react to it.”

HPE, in his view, is doing the same, while also solving the elephant in the room: 95% of all generative AI pilots are doomed to fail, as MIT research revealed this year.

According to HPE, that solution comes with the spoils from its Juniper acquisition. Announced at HPE Discover Barcelona was the HPE Juniper Networking QFX5250 switch, which connects GPUs with a Ultra Ethernet Transport-ready switch, powered by Broadcom Tomahawk 6 silicon. The offering, which touts 102.4 Tb/s bandwidth, was described by Rami Rahim, EVP, president and GM of HPE Networking as “ideal for next-generation front end and back end,” building out next-gen AI infrastructure for enterprises today.

Addressing the growing AI inference use case is another new bit of kit, the HPE Juniper Networking MX301 multiservice edge router. The compact 1RU router delivers 1.6 Tb/s performance and 400G connectivity across inference, multiservice, metro, mobile backhaul, and enterprise routing environments.

“It brings AI inference out of the cloud, closer to where the data sits. There’s nothing like it in the market today,” claimed Rahim.

That edge-focus follows an expansion of HPE and Nvidia’s AI factory solutions to include HPE Juniper Networking edge on-ramp and data center interconnect (DCI), as revealed earlier this week, alongside the debut of MP X10000 Data Intelligence Nodes in HPE Alletra Storage, based on the Nvidia AI Data Platform, providing an active, real time data layer for AI pipelines across all environments.

The third quarter of 2026 will also see the release of a dual platform WiFi access point which runs on both HPE Juniper Networking Mist and Aruba sensors, backing Rahim’s assertion that “eventually the difference between [HPE and Juniper] will disappear.”

Virtual takeover

HPE also announced updates to its software lines with a subtle jab at a certain embattled VM giant.

Without naming VMware, Neri noted recent challenges around rising virtualization costs, rigid architectures, upcoming deadlines, and the tough adaptation of single vendor-based runtime models.

In response, HPE Morpheus Enterprise Software touts full support for Kubernetes and containerized workloads on top of the hardware VM (HVM) hypervisor.

Resilience gets a boost through stretched cluster technology in Morpheus, meaning synchronous replication for VMs on the HVM hypervisor, promising automatic failover across distributed environments.

HPE’s data protection Zerto Software is also added to HPE Morpheus for VM data safeguarding within both HPE Morpheus VM Essentials and HPE Morpheus Enterprise. The former is also integrated with Veeam Data Platform v13 for hypervisor-based image-level backup and prompt recovery for VMs across private cloud environments.

And with Rahim stating it was “impossible for a networking company not to be a security company,” zero-trust security has come to HPE Morpheus Software, alongside Apstra Data Center Director capabilities via Juniper, allowing automated switch configuration and secure connectivity between VMs and hosts.

HPE’s boldened security stance was revealed earlier at Discover Barcelona with integration of STIG-hardened, FIPS-enabled Nvidia AI Enterprise in HPE Private Cloud air-gapped environments for added security and data sovereignty rail guards.

Also launched were the hybrid HPE StoreOnce 5720 solution and the all flash 7700 variant, which both integrate directly with HPE Alletra Storage MP and HPE SimpliVity for data backup. The 7700 boasts 300 TB/hour ingest capacity for prompt protection of critical workloads, with Nvidia claiming it can have recovery time in half and thus avoid up to $1 million per incident.