HPE deepens Oracle tie-up on AI data centre networking

HPE has expanded its collaboration with Oracle on networking for AI data centres, with Oracle planning a multi-year global deployment of HPE Juniper Networking products across its AI infrastructure.

The arrangement extends more than a decade of engineering work between Oracle and Juniper Networks, now part of HPE, and includes support services and financing.

The deployment is intended to span AI superclusters, regional data centre fabrics and edge networks across Oracle's cloud estate. HPE Juniper Networking PTX and MX routing platforms, along with QFX and EX switching platforms, are set to be used across Oracle data centres globally.

The companies are deepening ties as Oracle expands Oracle Cloud Infrastructure for AI workloads, where large GPU clusters place heavy demands on bandwidth, latency, congestion management and fault recovery. Oracle's regional and edge networks must also connect those clusters with services, customers and external networks across multiple markets.

Juniper equipment already supports parts of Oracle's network. The companies said MX and PTX routers are used in sections of the edge network, while QFX switches are deployed in several layers of the regional data centre fabric.

The latest expansion puts greater focus on the backend networks used by AI systems. HPE said its newer QFX switches are designed to provide the high-density connectivity, dynamic load balancing and congestion management needed for large RoCEv2 networks, a form of Ethernet networking commonly used to link AI servers at scale.

Infrastructure push

The move reflects the growing importance of networking in large AI buildouts, where the ability to move data quickly between processors can affect the efficiency of expensive computing clusters. Alongside power use, floor space and operational complexity, network design has become a central issue for cloud providers building larger AI environments.

Rami Rahim, executive vice president, president and general manager of networking at HPE, said Oracle's expansion would be a major test of network architecture for AI infrastructure. "Oracle is building AI infrastructure at extraordinary scale, and HPE is uniquely positioned to support OCI with one of the industry's most comprehensive networking portfolios for AI data centers," Rahim said.

"Together, we are advancing a high-performance networking architecture that helps Oracle address the traffic management, power, space, and operational requirements of increasingly large AI environments," he added.

For Oracle, the agreement gives HPE Juniper Networking equipment a larger role across several layers of its cloud network rather than in isolated segments. That includes AI clusters, broader regional data centre networks and edge systems that connect users and services to core infrastructure.

Mahesh Thiagarajan, executive vice president of Oracle Cloud Infrastructure, said the supplier had become more central to Oracle's design. "Building AI infrastructure at Oracle's scale requires an integrated ecosystem where every layer of the stack works together," Thiagarajan said.

"HPE Juniper Networking has been an important part of our cloud architecture and now plays a key role in our AI clusters, regional data centers, and edge networks with the scale, performance, and reliability our customers expect," he added.

Telemetry work

Alongside the hardware deployment, the companies said they are working together on intelligent telemetry to provide deeper visibility into device and fabric behaviour. The aim is to spot queue build-up, packet loss, traffic imbalances and component degradation earlier, so issues can be addressed before they have a wider effect on AI operations.

That focus highlights a technical challenge for operators of large AI systems: even brief slowdowns in one part of a cluster can reduce overall utilisation. In tightly connected GPU environments, delays measured in microseconds can have knock-on effects across the wider system.

Praveen Jain, senior vice president and general manager of data center and AI networking at HPE, said the telemetry work is intended to improve fault detection and availability. "Every microsecond matters to AI models and agents," Jain said.

"Through high-performance networking and our work with Oracle on deeper fabric visibility, we are helping OCI detect bottlenecks faster, maintain network availability, and improve the efficiency of AI infrastructure," he added.

As part of the agreement, HPE has issued Oracle warrants to purchase shares of HPE common stock.