Oracle plans to expand its use of HPE Juniper Networking across data centers worldwide as it builds larger artificial intelligence infrastructure, according to a September 2 announcement from Hewlett Packard Enterprise.
The companies described the arrangement as a potential multiyear deployment covering routing and switching equipment, support services and financing capabilities. HPE did not disclose the contract value, the number or location of participating data centers, deployment milestones or the terms of the financing. The announcement was distributed to Asian media on September 9.
Networking for larger AI clusters
HPE said Oracle Cloud Infrastructure already uses Juniper PTX and MX routers in parts of its edge network, while QFX switches operate across several layers of its regional data center fabric. The expanded collaboration could add PTX and MX routing platforms and QFX and EX switching systems across Oracle facilities globally.
Oracle is increasing the scale of its AI superclusters, which places additional pressure on the network connecting graphics processing units, storage, cloud services and customers. Large training and inference systems need to move data with low latency while limiting packet loss and congestion that can leave costly accelerators underused.
The latest QFX switches are intended to provide high-density links, dynamic load balancing and congestion management for backend networks using RoCEv2, or RDMA over Converged Ethernet version 2. The protocol allows systems to move data directly between memory resources over Ethernet with less processor involvement, although its performance depends on careful traffic engineering.
HPE said a common Juniper networking foundation could simplify engineering and operations across Oracle’s AI clusters, regional facilities and edge network. Oracle executive vice president Mahesh Thiagarajan said the equipment already plays a role in those environments, framing the planned expansion as an extension of an existing technical relationship rather than a new deployment from scratch.
Telemetry and an equity-linked agreement
The companies will also work on telemetry designed to give Oracle deeper visibility into devices and network fabrics. The planned tools would look for queue buildup, packet loss, uneven traffic and signs of component degradation, helping operators identify problems before they materially reduce GPU utilization or network availability.
HPE also issued Oracle warrants to purchase HPE common stock as part of the agreement. The public announcement did not state the number of shares, exercise price, vesting conditions or other warrant terms. That omission makes it difficult to assess the economic value of the incentive or the level of Oracle purchasing that may be required.
The collaboration was announced alongside HPE’s fiscal third-quarter results. Reuters reported on September 2 that HPE raised its fiscal 2026 and 2027 forecasts after quarterly revenue climbed 33.7 percent to $12.21 billion, supported by demand for AI servers and networking equipment. The report also noted supply constraints affecting memory, NAND, processors and drives.
Scope remains to be defined
The announcement gives HPE a prospective role in one of the technology industry’s largest AI infrastructure buildouts, while giving Oracle another supplier for the network layer surrounding its compute systems. It does not, however, quantify how much equipment Oracle has committed to buy or whether every contemplated deployment will proceed.
HPE’s use of the phrase “potential multiyear deployment” is therefore important. The companies have confirmed an expanded commercial and engineering relationship, existing Juniper deployments inside OCI, joint telemetry work and the issuance of warrants. Capacity, geography, schedule and financial commitments remain undisclosed and should not be inferred from Oracle’s broader AI infrastructure plans.
Oracle and AMD expand partnership to help customers achieve next-generation AI scale