Analyst: Ethernet is Beating InfiniBand in AI Networks

In a recently published report, networking market research and analyst firm Dell’Oro Group said Ethernet “is winning the war against InfiniBand in AI backend networks.” The report, titled AI Networks for AI Workloads, reveals the Ethernet-InfiniBand battle may drive nearly $80 billion in data center switch sales over the next 5 years.

“When we first initiated our coverage of AI backend networks in late 2023, the market was dominated by InfiniBand, holding over 80% share,” said Dell’Oro Group vice president Sameh Boujelbene. “Despite its dominance, we have consistently predicted that Ethernet would ultimately prevail at scale. What is notable, however, is the rapid pace at which Ethernet gained ground in AI backend networks. As the industry moves toward 800 Gbits/sec and beyond, we believe Ethernet is now firmly positioned to overtake InfiniBand in these high-performance deployments.”

Boujelbene added the key question is not whether or not Ethernet will surpass and hold off InbiniBand over the long term, but rather, “How much of the Ethernet opportunity will be captured by NVIDIA versus other switch vendors? In 2024, Celestica, NVIDIA, and Huawei led the Ethernet segment of the AI backend switch market. However, we anticipate significant share shifts in 2025 as large-scale Ethernet deployments at Meta, Microsoft, Oracle, and other GPU-as-a-service providers ramp up—creating opportunities for additional switch vendors such as Accton, Arista, Cisco, Juniper/HPE, Nokia, and others to grain traction.”

The AI networks report, published in July, also projects that growth from GPU-as-a-service providers including CoreWeave, Lambda Labs, and Vultr will outpace that of Tier 1 cloud service providers in the next 5 years.

Also this year (2025), the majority of switch ports deployed in AI backend networks will be 800 Gbits/sec, but by 2027 1.6 Tbits/sec will represent the majority and 3.2 Tbits/sec will lead market share by 2030.

Additionally, copackaged optic adoption in AI clusters will begin to materialize in the next few years, led by NVIDIA, the report concludes.

You can find more information about the report AI Networks for AI Workloads here.