HPE AI System Revenue Crossed $2 Billion in Q2 FY2026
Hewlett Packard Enterprise reported in its Q2 FY2026 earnings (February through April 2026, results published June 3, 2026) that AI system revenue — comprising NVIDIA H100, H200, and B200 GPU-based ProLiant and Cray XD server systems sold to enterprise and government customers for AI training and inference workloads — reached $2.1 billion in the quarter, crossing $2 billion for the first time in HPE’s history and representing a 42 percent year-over-year increase from $1.48 billion in Q2 FY2025, driven by accelerating enterprise adoption of on-premises AI infrastructure and HPE’s expanded GPU system portfolio that now spans from the ProLiant DL380 Gen11 (entry-level single-GPU AI inference server) through the Cray XD6XX supercomputer family (multi-rack AI training system designed for national laboratory and hyperscale enterprise deployments). HPE’s Q2 FY2026 investor filings show total company revenue reaching $7.7 billion in the quarter, up 9 percent year over year from $7.1 billion in Q2 FY2025, with the Server segment (which includes AI systems within the broader server portfolio) contributing $4.1 billion, the Networking segment — now incorporating both Aruba campus and branch networking and the Juniper Networks enterprise and data centre switching portfolio acquired in the March 2024 $14 billion transaction — contributing $1.7 billion, and the HPE Hybrid Cloud segment (GreenLake cloud services and storage) contributing $1.4 billion. HPE’s AI system revenue growth of 42 percent year over year positions the company as the second-largest provider of enterprise AI server infrastructure after Dell Technologies, which reported $10.3 billion in AI server revenue for full fiscal year FY2026 (ending January 2026), and ahead of Lenovo and Super Micro in the enterprise segment of the AI server market that IDC distinguishes from the hyperscaler direct-to-NVIDIA procurement market that CoreWeave and cloud providers access through separate supply relationships. HPE’s competitive differentiation in AI servers relative to Dell and Lenovo operates primarily through the HPC (high-performance computing) and national laboratory segment, where HPE’s Cray supercomputer heritage gives the company a multi-decade relationship with the US Department of Energy, European national computing centres, and defence research laboratories that represent the largest single-system AI procurement decisions in the market — systems exceeding $100 million in individual contract value — and through the HPE GreenLake subscription model that allows enterprise customers to deploy AI server infrastructure on a consumption-based operating expense model rather than a capital expenditure purchase, reducing the budget approval friction that large upfront AI server capital commitments face in enterprise procurement processes. Dell Technologies AI server revenue crossing $10 billion in FY2026 establishes the market leadership context against which HPE’s $2 billion quarterly milestone is measured: Dell’s approximately 17 percent market share in the enterprise AI server market (per IDC Q4 2025 data) compared to HPE’s approximately 11 percent share reflects Dell’s stronger commercial enterprise relationships built through the Dell Direct sales model and PowerEdge brand recognition, while HPE’s higher share of the HPC and government segment reflects the Cray acquisition’s technical differentiation in extreme-scale computing — creating two overlapping but structurally different customer bases between which the AI server market’s growth is distributed.
HPE’s Juniper Networks integration — completed in March 2024 after an 18-month regulatory review — has created a networking business that competes directly with Cisco’s Catalyst and Nexus families in the enterprise campus, branch, and data centre switching segments while adding Juniper’s AI-Native Networking Platform (formerly Mist AI, the AI-powered wireless and wired network management system that Juniper acquired in 2019 for $405 million) to HPE’s Aruba campus networking portfolio. The combined HPE Networking Business Unit — rebranded as HPE Networking in Q1 FY2025 — generates approximately $1.7 billion in quarterly revenue from switching hardware (Aruba CX, Juniper EX and QFX campus and data centre switches), wireless access points (Aruba AP series), and the AI-Native Networking Platform subscription service that replaces traditional network management tools with an AI-driven platform that identifies network anomalies, predicts capacity constraints, and automates remediation actions before user-reported performance degradation occurs. Juniper’s AI-Native Networking Platform subscription revenue — approximately $280 million quarterly in Q2 FY2026, growing at approximately 25 percent year over year — is the highest-margin product in the combined HPE Networking portfolio because the SaaS subscription model delivers ongoing AI-driven network insight without incremental hardware sales, creating a recurring revenue stream attached to the installed base of Aruba and Juniper switching and wireless hardware that any networking customer can access independently of hardware refresh cycles. The AI-Native Platform’s value proposition — providing network operations teams with AI-generated anomaly alerts, capacity utilisation forecasts, and automated ticket creation for incidents that the AI system has correlated across the campus wireless, wired access, and WAN segments — is validated by the customer success metrics that HPE Networking discloses: enterprises using AI-Native for full-stack campus management report a 28 percent reduction in network-related helpdesk tickets and a 41 percent reduction in mean time to resolve network incidents, metrics that translate directly to IT operations cost reductions that enterprise procurement teams cite as the primary economic justification for the subscription fee. IDC’s enterprise networking market sizing for Q2 2026 shows the combined enterprise switching and wireless LAN market at approximately $17 billion annually, with HPE Networking holding approximately 20 percent market share (second to Cisco’s approximately 44 percent) after the Juniper acquisition, a combined position that was previously split between Aruba’s 13 percent campus wireless share and Juniper’s 9 percent enterprise switching share. Marvell Technology’s AI revenue crossing $1 billion in Q1 FY2027 provides the silicon supply layer for HPE’s AI server systems: the custom ASIC interconnect components and Ethernet switching silicon that Marvell supplies to data centre switching vendors are incorporated in HPE’s AI cluster networking fabric (HPE Slingshot interconnect for Cray systems, HPE’s 400G Ethernet fabric for ProLiant AI clusters) and represent the upstream semiconductor supply chain that HPE’s AI system capacity additions depend on alongside NVIDIA GPU supply allocation. Cisco’s AI networking revenue and Nexus Hyperfabric launch establishes the primary competitive reference for HPE Networking: Cisco’s Nexus Hyperfabric AI data centre fabric competes directly with HPE’s AI cluster networking products for the enterprise customer who wants a managed AI data centre networking layer, while Cisco’s Catalyst campus switching competes with HPE’s Aruba CX and Juniper EX portfolio for campus enterprise LAN deployments — making Cisco and HPE the two primary full-stack enterprise networking vendors in a market where Arista Networks and Extreme Networks serve narrower segments.
What HPE GreenLake AI Infrastructure Crossing $1 Billion in Annual Contract Value Signals About On-Premises AI Subscription Models
HPE GreenLake — the consumption-based infrastructure subscription model that allows enterprises to deploy HPE server, storage, and networking infrastructure on a pay-per-use operating expense model rather than a capital acquisition — reached $1 billion in annual contract value (ACV) for AI infrastructure orders in FY2026, a milestone that HPE CEO Antonio Neri cited in the Q2 FY2026 earnings commentary as evidence that enterprise customers are increasingly choosing on-premises AI infrastructure subscription over public cloud GPU rental for production AI workloads at a scale where the TCO comparison favours dedicated on-premises capacity over cloud variable pricing. GreenLake’s AI infrastructure contracts typically span three to five years at a fixed reservation commitment (similar to cloud reserved instance pricing) with a consumption overlay for burst above the committed level, structured to deliver approximately 15 to 25 percent total cost savings relative to equivalent AWS, Azure, or GCP GPU instance pricing for workloads running above approximately 70 percent continuous utilisation — the utilisation threshold at which on-premises infrastructure becomes cheaper than cloud on a per-GPU-hour basis, accounting for the capital cost of the hardware, the data centre space, power, and cooling, and the IT operations overhead that cloud pricing includes implicitly. The $1 billion ACV milestone for GreenLake AI is significant for HPE’s business model transformation because GreenLake contracts convert what would historically be a lumpy, project-based capital equipment revenue stream (one large AI server order per customer per refresh cycle, approximately every 4 years) into a recurring subscription revenue stream that grows with the customer’s AI workload expansion between hardware refresh cycles, creating revenue predictability that capital equipment sales cannot provide and that HPE is using to justify the valuation multiple expansion it has sought as its GreenLake ACV grows as a proportion of total server revenue. ARM Holdings’ server market penetration through AWS Graviton provides the architectural context for the on-premises AI server market that HPE’s GreenLake AI contracts serve: while AWS Graviton represents Amazon’s strategic substitution of ARM-based custom silicon for x86 CPUs in cloud compute, HPE’s GreenLake AI infrastructure is primarily NVIDIA GPU-based, meaning the on-premises AI server market that HPE serves is GPU-capacity-constrained in a way that is fundamentally different from the x86 CPU refresh cycle dynamics that governed enterprise server procurement before the AI infrastructure era, and that makes HPE’s AI server order backlog — approximately $4 billion at the end of Q2 FY2026 — a genuine leading indicator of future revenue rather than a soft commitment that cancels in economic downturns.
