# Ultra Ethernet 1.0 turns commodity Ethernet into an AI back-end fabric
> A 560-page spec gives Ethernet InfiniBand-class RDMA and congestion control; the 2026 work item is Programmable Congestion Management

**Meta:** type: story · date: 2026-06-10 · heads: खामोश बदलाव, लंबी पारी · 9 takes · 3 lenses · 3 regions

## Summary

The [Ultra Ethernet Consortium](/hi/entity/ultra-ethernet) (UEC) shipped Specification 1.0, a 560-page, vertically integrated stack across NICs, switches, optics and cables, giving commodity Ethernet the features that made InfiniBand the AI back-end default: native RDMA, advanced flow control, path diversity, congestion awareness and ultra-low tail latency, while keeping standard Ethernet optics and tooling. It is the scale-**out** complement to [UALink's](/hi/n/ualink-2-0-specs-2026) scale-**up** fabric, and the cost-led alternative to [Nvidia's](/hi/entity/nvidia) InfiniBand for million-GPU clusters. The headline 2026 work item is **Programmable Congestion Management**, code a congestion-control algorithm once in a standard language and run it on any UE-compliant NIC, decoupling the fabric's most contested layer from any single vendor.

## By the numbers

- 560+, pages in the UEC 1.0 specification.
- 2026, first PCM (Programmable Congestion Management) work item.
- 5, stack layers covered (NICs, switches, optics, cables, transport).
- 1.0, current spec generation, released ahead of 2026 refinements.

## Why it matters

InfiniBand's lock on AI back-end networks is Nvidia's quietest moat. An open Ethernet stack with RDMA-class performance lets hyperscalers wire million-GPU clusters on multi-vendor gear at commodity-optics prices, shifting where the networking margin sits and giving Broadcom, Cisco and Arista a credible AI-fabric pitch.

## What to watch

- Programmable Congestion Management landing in shipping NICs.
- UEC-1.0 deployments in named hyperscale clusters vs continued InfiniBand.
- How UEC scale-out and UALink scale-up divide the fabric stack in practice.

## Regional takes (batched by bias / lens)

### unlabelled
- **Ultra Ethernet Consortium** (United States, en) — The consortium's own announcement of Specification 1.0, a 560-page, vertically integrated stack spanning NICs, switches, optics and cables, giving Ethernet modern RDMA, path diversity, congestion control and low tail latency for AI/HPC scale-out.
  Source: https://ultraethernet.org/ultra-ethernet-consortium-uec-launches-specification-1-0-transforming-ethernet-for-ai-and-hpc-at-scale/
- **Converge Digest** (United States, en) — 
  Source: https://convergedigest.com/ultra-ethernet-consortium-releases-specification-1-0/
- **Network World (1.0 launch)** (United States, en) — 
  Source: https://www.networkworld.com/article/4006285/ultra-ethernet-consortium-publishes-1-0-specification-readies-ethernet-for-hpc-ai.html
- **STORDIS** (Germany, en) — 
  Source: https://stordis.com/ultra-ethernet-consortium/
- **Rivos** (United States, en) — 
  Source: https://www.rivosinc.com/resources/blog/ultra-ethernet-specification-1-0-a-game-changer-for-ai-networking
- **Nokia** (Finland, en) — 
  Source: https://www.nokia.com/blog/nokia-celebrates-the-future-of-ai-networking-with-ultra-ethernet-consortium/
- **IntelligentVisibility** (United States, en) — 
  Source: https://intelligentvisibility.com/guides/ultra-ethernet-consortium-ai-networking

### enterprise networking desk
- **Network World** (United States, en) — Reports the UEC's 2026 priorities, chiefly Programmable Congestion Management, letting anyone code a congestion algorithm in a standard language that runs on any UE-compliant NIC, and frames the year as Ethernet groups racing bandwidth and AI demand against InfiniBand and proprietary fabrics.
  > "The 2026 priority is Programmable Congestion Management: code a congestion algorithm once in a standard language and run it on any UE-compliant NIC."
  Source: https://www.networkworld.com/article/4113364/ethernet-groups-keep-2026-focus-on-higher-bandwidth-ai-demands.html

### HPC trade desk
- **HPCwire** (United States, en) — Stresses that UEC 1.0 brings InfiniBand-class features, native RDMA, advanced flow control, congestion awareness, ultra-low tail latency, while keeping compatibility with standard Ethernet optics and tooling, the cost argument for hyperscalers building million-GPU back-end networks.
  > "UEC 1.0 gains InfiniBand-class native RDMA and congestion control while keeping standard Ethernet optics and tooling, the cost case for AI scale-out."
  Source: https://www.hpcwire.com/off-the-wire/ultra-ethernet-consortium-releases-specification-1-0-to-support-scalable-ai-and-hpc-networking/

## Across the graph
- Related: [[ualink-2-0-specs-2026]], [[nvlink-fusion-marvell-2026]], [[nvidia-spectrum-x-photonics-cpo-2026]]
- Entities: Ultra Ethernet, Broadcom, Nvidia, Amd, United States

---
Canonical: https://rbtfl.xyz/hi/n/ultra-ethernet-1-0-2026