Oracle's HPE Juniper Network Expansion: What MX, PTX and QFX Do

Key takeaways

• HPE describes the Oracle agreement as a potential multi-year deployment, not a fixed rollout schedule or a disclosed purchase volume.
• In OCI today, MX and PTX routers support parts of the edge network, while QFX switches operate across multiple layers of the regional data center fabric.
• HPE says its latest QFX switches are intended to provide dense connectivity, dynamic load balancing and congestion management for large RoCEv2 AI back-end networks.
• EX switches are included in the potential global deployment, but HPE does not assign them a specific current OCI role in the announcement.
• HPE issued Oracle warrants to buy HPE shares, but the announcement does not give their quantity, exercise price or other terms.

What Oracle and HPE announced

On September 2, 2026, HPE announced an expanded collaboration with Oracle around networking for Oracle Cloud Infrastructure. The companies are building on more than a decade of engineering work between Oracle and Juniper Networks.

The announcement covers the potential multi-year deployment of HPE Juniper Networking PTX and MX routers and QFX and EX switches across Oracle data centers globally. It also includes multi-year networking support services. The word potential matters: HPE did not publish a firm site list, equipment count or deployment calendar.

That distinction makes the agreement relevant without turning it into a number that HPE has not disclosed. It identifies the network families Oracle may use as OCI expands, while leaving the eventual volume and timing open.

What is already running in OCI

HPE separates OCI's current network footprint from the planned expansion. Its release says MX and PTX routers support portions of OCI's edge network today. QFX switches, meanwhile, power multiple layers of OCI's regional data center fabric.

Those are the most specific role assignments in the announcement. HPE does not say that every OCI edge location uses both router families, and it does not map MX and PTX to separate traffic classes. It likewise does not identify which QFX models are installed at particular regions.

Why QFX is singled out for AI back-end networks

HPE gives the latest QFX switches a more detailed forward-looking role. It says they will provide high-density connectivity, dynamic load balancing and advanced congestion management for large AI back-end networks using RoCEv2, or RDMA over Converged Ethernet version 2.

That description ties QFX to the east-west network inside large AI environments, where many accelerators exchange data. It does not reveal the port speeds, switch models, cluster sizes or topology Oracle will use. Those design details would require separate product or deployment documentation.

The release also explains why the wider network matters. Large GPU clusters raise requirements for bandwidth, latency, congestion management and fault recovery. OCI's regional and edge networks must also connect those clusters to services, customers and external networks. HPE's announcement therefore spans both the AI back-end fabric and the broader cloud network around it.

Telemetry is part of the collaboration

Oracle and HPE are also working on intelligent telemetry for greater visibility into device and fabric behavior. HPE says earlier detection of queue buildup, packet loss, traffic imbalances and component degradation can help OCI respond faster and reduce the effect on GPU utilization and network availability.

This is a monitoring and troubleshooting commitment, not a published service-level guarantee. HPE does not provide benchmark results or quantify how much the telemetry will improve utilization or uptime.

Support, financing and the warrant

The collaboration includes multi-year networking support services, and HPE says the broader relationship includes financing capabilities. Separately, HPE states that it issued Oracle warrants to purchase HPE common stock as part of the agreement.

The announcement does not say that the warrants pay for the networking deployment or offset Oracle's infrastructure costs. It also does not disclose the exercise price, number of shares, vesting conditions or expiration date. Without those terms, the warrant should be treated as a disclosed element of the commercial relationship, not as a way to calculate the deal's value.

What remains unknown

HPE's release establishes the vendors, product families and broad network roles, but it leaves several material questions unanswered:

• how many Oracle data centers will receive new equipment;
• which specific MX, PTX, QFX or EX models Oracle will deploy;
• when deployments will start or finish;
• the contract value and financing structure; and
• the terms of the warrants issued to Oracle.

The clearest reading is therefore a framework for expanding an existing relationship. Oracle already uses HPE Juniper routing and switching in its edge and regional networks, and the new agreement creates a path to extend that footprint as OCI's AI infrastructure grows. The scale of that extension will only become measurable when either company publishes orders, site deployments or financial terms.

Sources

This article was researched and fact-checked against the following sources:

• HPE and Oracle deepen networking collaboration to accelerate gigawatt-scale AI infrastructure (hpe.com)