Oracle is expanding the capabilities of its Oracle Cloud Infrastructure (OCI) with a new generation of cloud networking architecture designed to deliver faster, more efficient, and more secure performance across workloads ranging from enterprise databases to large-scale AI and high-performance computing.
The newly introduced Oracle Acceleron suite marks a significant step in how Oracle handles data movement in the cloud – streamlining the network stack and collapsing layers of complexity to improve speed, reliability, and cost-efficiency.
According to Oracle, Acceleron represents more than a decade of networking innovation within OCI. It integrates a series of software and hardware advancements, including dedicated fabric networks, multi-planar routing, converged network interface cards (NICs), and host-level zero-trust packet routing. Together, these enhancements are designed to deliver predictable high-bandwidth, ultra-low latency performance with built-in security and resiliency.
The approach is aimed at addressing one of the most pressing challenges in today’s cloud environments: how to efficiently manage massive data movement for increasingly complex workloads – particularly AI and analytics – without sacrificing security or scalability. Oracle claims that the new architecture offers up to twice the network processing capacity and storage IOPS, with line-rate encryption throughput and microsecond-level latency. Importantly, all of this performance improvement comes at no additional cost to customers.
“Our customers want cloud infrastructure to help them innovate faster, operate more efficiently, and scale with confidence,” said Clay Magouyrk, CEO of Oracle Cloud Infrastructure. “For more than a decade, Oracle has led foundational cloud networking innovation, and these latest advances extend Oracle Acceleron to deliver uncompromised performance, scale, and security for any cloud workload.”
Acceleron’s new capabilities operate across three primary layers of Oracle’s networking architecture. At the fabric level, dedicated network fabrics isolate traffic for specific workloads – such as Exadata databases, HPC clusters, or large AI training environments – enabling predictable latency and throughput even at hyperscale. The multi-planar network design connects each customer NIC to multiple isolated planes, automatically shifting traffic to alternate routes in the event of a network issue. This results in greater resiliency, reduced tail latency, and minimal job restarts – critical for time-sensitive AI workloads.
Higher Throughput
The new fabric accelerator technology also removes intermediaries from network paths, allowing packets to flow through the most direct route possible. This “disintermediated” approach minimizes hops, reduces bottlenecks, and improves overall consistency and efficiency.
At the host level, Acceleron introduces a converged NIC that combines customer and provider network planes while maintaining isolation between them. This SmartNIC-based design supports NVMe-over-TCP for storage acceleration, line-rate encryption, and live patching without downtime. Oracle says this configuration can deliver up to 2x the throughput of prior-generation dual-NIC designs while reducing hardware costs and complexity.
Security has also been re-engineered at the packet level. The new Zero-Trust Packet Routing (ZPR) system enforces least-privilege access at the first packet, ensuring that only authorized traffic ever reaches sensitive endpoints. ZPR includes new deny-based identity and access management (IAM) policies that prevent unauthorized inspection paths, protecting data at the host before it traverses the network.
These developments reflect Oracle’s broader strategy of building a high-performance cloud optimized for data-intensive applications and artificial intelligence. With AI clusters scaling to tens of thousands of GPUs, and enterprise applications demanding deterministic network performance, Oracle is positioning Acceleron as the backbone for its next phase of OCI growth.
Industry partners are already aligning with Oracle’s vision. Arista Networks president and CTO Ken Duda said the collaboration continues to “pioneer innovation” for customers, highlighting Arista’s readiness to deliver next-generation AI networking through Oracle Acceleron. AMD, which provides Pensando data processing units (DPUs) for OCI, also emphasized the long-standing collaboration. “OCI launched with AMD Pensando DPUs, and we’ve worked together to deliver breakthrough performance and security for all workloads,” said Forrest Norrod, executive vice president and general manager of AMD’s Data Center Solutions Business Group. He added that AMD-powered converged NICs for Oracle Acceleron are expected to debut early next year.
