- InfiniBand’s long dominance is facing real pressure from the open-standards Ethernet movement
- Meta and Nvidia are betting on openness to scale AI networks
- The ESUN project connects industry competitors through shared networking ambitions
the Open Computing Project (OCP) It announced a new initiative known as Ethernet for Scale-Up Networking (ESUN), which aims to develop open standards for high-performance communications within AI clusters.
This collaboration brings together companies like Meta, Nvidia, AMD, Cisco, and OpenAI to explore how Ethernet can compete with existing interconnects like InfiniBand in large-scale data centers.
Other companies joining the collaboration include Arista, ARM, Broadcom, HPE Networking, Marvell, Microsoft, and Oracle.
Opening up networks for AI groups
InfiniBand has long dominated the market for high-speed AI networks, accounting for about 80% of communications infrastructure Graphics processing units and accelerators.
However, the ESUN Group believes that Ethernet’s maturity, cost-effectiveness and interoperability make it a strong candidate for scaling AI clusters.
Unlike proprietary systems, widespread knowledge of Ethernet among engineers can help reduce the complexity in managing massive AI workloads.
Proponents argue that using Ethernet as an open standard will allow operators to scale infrastructure while lowering costs.
OCP Group’s new AI tools initiative builds on previous work under the SUE-Transport (SUE-T) program, which explored Ethernet transport for multiprocessor systems.
ESUN participants will meet regularly to define standards for switching behavior, including protocol headers, error handling, and lossless data transfer.
The group will also study how network design affects load balancing and memory arrangement within GPU-based systems.
It plans to coordinate with the Ultra Ethernet Consortium and the IEEE 802.3 standards body to ensure compatibility across the broader Ethernet ecosystem.
Several companies have already developed Ethernet-based products aimed at scaling AI – for example, Broadcom’s Tomahawk Ultra switch supports up to 77 billion packets per second, and Nvidia’s Spectrum-X platform combines Ethernet and hardware AI cluster accelerators.
However, Meta, which co-founded OCP in 2011, views ESUN as a natural extension of its efforts for open hardware within data centers.
However, observers point out that replacing existing InfiniBand networks will require Ethernet to prove itself under more demanding AI workloads, where latency and reliability are critical.
ESUN’s success will depend on striking a balance between openness and performance. Proponents see a future where AI systems run on interoperable hardware using standardized Ethernet technologies.
However, given the scale and sensitivity of AI infrastructure, it remains uncertain whether industry momentum will shift decisively away from private interconnections.
For now, ESUN represents an ambitious effort, and it remains to be seen if it can match InfiniBand’s performance.
Follow TechRadar on Google News and Add us as a favorite source Get expert news, reviews and opinions in your feeds. Make sure to click the follow button!
And of course you can too Follow TechRadar on TikTok To get news, reviews and unboxings in video form, and get regular updates from us on WhatsApp also.