Industry analysis shows Ethernet-based networking is displacing InfiniBand as the preferred GPU interconnect standard for AI datacenters.

According to the recently published AI Networks for AI Workloads report by Dell'Oro Group, Ethernet is winning the war against InfiniBand in AI back-end networks, potentially driving nearly $80 B in data center switch sales over the next five years. When Dell'Oro Group first initiated its coverage of AI Back-end Networks in late 2023, the market was dominated by InfiniBand, holding over 80 percent share. Despite this dominance, the firm had consistently predicted that Ethernet would ultimately prevail at scale.

"What is notable, however, is the rapid pace at which Ethernet gained ground in AI back-end networks," said Sameh Boujelbene, Vice President at Dell'Oro Group. "As the industry moves to 800 Gbps and beyond, we believe Ethernet is now firmly positioned to overtake InfiniBand in these high-performance deployments." Although Ethernet is gaining ground, a key question remains regarding market capture: how much of the Ethernet opportunity will be seized by NVIDIA versus other switch vendors?

In 2024, Celestica, NVIDIA, and Huawei led the Ethernet segment of the AI Back-end Switch market. However, Dell'Oro Group anticipates significant share shifts in 2025 as large-scale Ethernet deployments at Meta, Microsoft, Oracle, and other GPU-as-a-Service Providers ramp up. These deployments are expected to create opportunities for additional switch vendors such as Accton, Arista, Cisco, Juniper/HPE, Nokia, and others to gain traction in the market.

Dell'Oro Group's AI Networks for AI Workloads Advanced Research Report explores the use cases that optimize InfiniBand versus Ethernet in AI back-end networks, as well as the choices made by large Cloud Service Providers. The report provides a 5-year market forecast on a worldwide basis, segmented by customer type, by InfiniBand versus Ethernet, and by port speed.