Hewlett Packard Enterprise: HPE Discover 2026 Highlights an Expanding AI Portfolio with Networking Stealing the Show

研报英文原文证据摘录

Hewlett Packard Enterprise: HPE Discover 2026 Highlights an Expanding AI Portfolio with Networking Stealing the Show

Samik Chatterjee, CFA AC North America Equity Research

(1-212) 622-0798 16 June 2026 J P M O R G A N

samik.x.chatterjee@jpmorgan.com

earlier to secure supply. The company continues to be prudent in its outlook, with supply

availability expected to remain severely constrained in FY27 and memory in switches

representing the main bottleneck, although HPE expects to upgrade towards less-

constrained memory and potentially even HBM. On the silicon side, management

underscored its position as Broadcom’s largest OEM partner alongside two dedicated

in-house silicon roadmaps for routing (Trio and Express), plus a separate in-house silicon

track across Aruba campus portfolio. Net-net, FY26 supplier capacity is already fully

allocated, while FY27 long-term agreements are still being negotiated as multi-year

(rather than annual) commitments and have not yet been locked in.

• Introduces new Private Cloud AI features focused on managing agentic workflows,

in partnership with Nvidia. The additional features are intended to take advantage of

full-stack, AI factory deployments, including those based on Nvidia’s reference

architecture and HPE ProLiant Compute, in order to control token costs, maximize GPU

utilization, and enable long-term scalability. HPE has integrated Nvidia’s Agent Toolkit,

which includes the Nemotron open models and NemoClaw, among other tools for

running agentic workflows. HPE also announced new security and governance

capabilities, addressing secure agent registration, rogue agent detection and clean-state

rewind via Zerto Software, and extended support for model context protocol (MCP)

through the HPE Data Fabric Software suite.