Oracle AI Networking Update Puts Switch Market in Focus

Oracle Adds RDMA and Failover Networking Features for Distributed AI Workloads

LAS VEGAS, United States — October 28, 2026 — Oracle announced new networking capabilities for AI workloads on Tuesday at its Oracle AI World conference. The additions include high-performance connectivity and mechanisms intended to keep distributed AI systems running through network failures. The announcement draws attention to the AI networking switch market, where fabric performance and reliability affect how well large accelerator clusters are used.

Oracle said the additions extend Oracle Acceleron, the networking and infrastructure suite within Oracle Cloud Infrastructure (OCI), across AI data movement, distributed AI communication and secure multi-tenant infrastructure. Pradeep Vincent, executive vice president for Compute and AI Infrastructure at OCI, said that when thousands of accelerators work together, a single congested or failed network path can have consequences beyond one connection.

The company said Acceleron combines dedicated network fabrics, more direct data paths and a converged network interface card, with host-level zero-trust routing and multi-planar designs. Oracle noted that the timing and features of the capabilities it described remain subject to change.

According to Next Move Strategy Consulting, the AI networking switch market covers Ethernet, InfiniBand and proprietary scale-up switch systems used to interconnect accelerators in AI training and inference clusters. NMSC valued the market at USD 22.40 billion in 2025 and estimates USD 38.60 billion in 2026. It forecasts USD 165.00 billion by 2035, a compound annual growth rate of 17.5% from 2026 to 2035.

Key Highlights:

• Oracle Acceleron High-Performance Fabric brings RDMA networking to OCI front-end compute for inference, storage and high-performance computing workloads, Oracle said.

• Expanded NVMe block-storage attachment options span bare metal and virtual machine shapes, with performance improvements of up to 3x, according to Oracle.

• The Oracle Collective Communications Library spreads traffic across OCI’s backend multiplanar network to keep distributed AI infrastructure running when links, ports or network cards fail, the company said.

• Transparent Front-End Multiplanar Networking adds path selection, failover and congestion control for front-end RDMA traffic, according to Oracle.

Analyst Insight:

According to analysts at Next Move Strategy Consulting, small amounts of packet loss reduce accelerator utilisation, which is why cluster load balancing, congestion control and telemetry software matter across the market. NMSC analysts note that switch spending now scales with accelerator shipments, so fabric technology, port speed and buyer type explain most of the movement in revenue.

NMSC analysts add that Ethernet is the dominant fabric technology, at about 54% of 2025 revenue, and that high-speed Ethernet switches at 800G held the largest share by port speed. NMSC’s analysis indicates that platforms at 1.6T and above will account for about 55% of 2035 revenue, up from about 5% in 2025, which makes the 2027 to 2029 qualification cycle decisive for vendor rankings.

Industry Outlook:

The announcement concerns network software and architecture on OCI rather than switch hardware, but it places congestion handling and failure recovery among the design priorities for large AI clusters. NMSC projects the market for AI networking switches to grow at a 17.5% CAGR through 2035, with North America holding about 63% of revenue in 2025. Customer adoption of the new Oracle capabilities, and when they become available, will indicate how AI data center networking requirements develop.

Source: Oracle

For More Information: Download FREE Sample on AI Networking Switch Market

Prepared By: Sanyukta Deb