cd /news/ai-infrastructure/ualink-ultra-accelerator-link-open-s… · home › topics › ai-infrastructure › article
[ARTICLE · art-154203] src=en.wikipedia.org ↗ pub= topic=ai-infrastructure verified=true sentiment=· neutral

UALink – Ultra Accelerator Link – Open-Source/Hardware Alternative to NVLink

The UALink Consortium published version 2.0 of the Ultra Accelerator Link (UALink) open specification in April 2026, following the public release of the 1.0 (UALink_200) specification in 2025. UALink is a die-to-die interconnect and serial bus for AI accelerators co-developed by Alibaba, AMD, Apple, Astera Labs, AWS, Cisco, Google, Hewlett Packard Enterprise, Intel, Meta, Microsoft and Synopsys, positioning it as an open alternative to Nvidia's NVLink. The 1.0 specification runs at 200Gbps over the IEEE P802.3dj PHY layer, and UALink Switches (ULS) connect up to 1024 accelerators within an AI pod, each assigned a unique 10-bit routing identifier.

read2 min views2 publishedOct 21, 2026
UALink
Ultra Accelerator Link
Year started 2024
First published 23 October 2024
Latest version 2.0 7 April 2026
Organization UALink Consortium
| Related standards | [Infinity Fabric](https://en.wikipedia.org/wiki/Infinity_Fabric) ,[Ultra Ethernet](https://en.wikipedia.org/wiki/Ultra_Ethernet?action=edit&redlink=1) | 
| Website | [ualinkconsortium](http://ualinkconsortium.org) | 

Ultra Accelerator Link (UALink) is an open specification for a die-to-die interconnect and serial bus between AI accelerators. It is co-developed by Alibaba, AMD, Apple, Astera Labs, AWS, Cisco, Google, Hewlett Packard Enterprise, Intel, Meta, Microsoft and Synopsys.<sup>[1]</sup> The UALink Consortium officially incorporated as an organization and electronics industry consortium in 2024, for promoting and advancing UALink.

Its first specification will provide interconnectivity specifically for a scalable network. The initial 1.0 version 200Gbps UALink specification, is based on the IEEE P802.3dj 200 Gb/s (Ultra Ethernet) PHY Layer. Each system node is made up of a host and as many accelerators as needed, and is managed by one OS image. The accelerators can be connected to the host using a variety of interconnect protocols (CXL, PCIe, XGMI, CHI c2c, Infinity Fabric). UALink Switches (ULS) connect up to 1024 accelerators within an AI 'pod', where each Accelerator is assigned a unique 10-bit routing identifier. Each UALink Switch port connects to a distinct Accelerator.<sup>[2]</sup> AMD Infinity Fabric is expected to be the main shared-memory protocol.[3]

The specification was due to be available to Contributor Members in 2024.<sup>[4]</sup> The 1.0 version (UALink_200) was released to the public in 2025. Members of the public can download an "evaluation copy" after providing contact information.[5]

Version 2.0 of the specification was published in April 2026.[6]

References #

[
[edit](https://en.wikipedia.org/w/index.php?title=UALink&action=edit§ion=1)]

1. [↑](#cite_ref-1)["UALink Members UCIe"](https://www.ualinkconsortium.org/members) .*ualinkconsortium.org* . Retrieved 2024-11-01.
  1. ↑"UALink 1.0 White Paper v3" (PDF). Archived fromthe original (PDF) on 2025-08-09.
  2. ↑ Morgan, Timothy Prickett (30 May 2024)."Key Hyperscalers And Chip Makers Gang Up On Nvidia's NVSwitch Interconnect" . The Next Platform. Retrieved 19 May 2026.
  3. ↑"ABOUT UALINK" .UALink Consortium . Retrieved 2025-01-20.
  4. ↑"Ultra Accelerator Link Consortium, Inc.™ UALink_200 Rev 1.0 Specification" .ualinkconsortium.org .pdf
  5. ↑ Cooter, Maxwell (7 April 2026)."New v2 UALink specification aims to catch up to NVLink" . Network World. Retrieved 19 May 2026.

See also #

[
[edit](https://en.wikipedia.org/w/index.php?title=UALink&action=edit§ion=2)]
[
[edit](https://en.wikipedia.org/w/index.php?title=UALink&action=edit§ion=3)]

- [Ultra Ethernet Consortium](https://ultraethernet.org) (has a draft-sharing relationship with IEEE P802.3dj)
── more in #ai-infrastructure 4 stories · sorted by recency
── more on @ualink consortium 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/ualink-ultra-acceler…] indexed:0 read:2min 2026-10-21 · —