UALink 2.0 Adds In-Network Compute, Chiplet Support, and 200G Performance
The UALink Consortium released a new set of specifications that expand its open scale-up interconnect framework for AI infrastructure, adding support for ...
In Focus: UALink
🌐 UALink Open Scale-Up Fabric for AI Accelerators | |
| UALink is an open, high-speed interconnect designed for scale-up communication between AI accelerators within servers and rack-scale systems. Developed by the UALink Consortium, the technology provides a standards-based alternative to proprietary GPU interconnects, enabling coherent, low-latency communication across multi-vendor AI platforms. | |
| Why It Matters | Future AI systems require hundreds of accelerators to communicate efficiently within a server or rack before traffic reaches the larger data center network. UALink addresses this scale-up challenge, while technologies such as Ultra Ethernet and InfiniBand provide scale-out connectivity between servers and racks. |
| Category | Networking Technology |
| Typical Deployment | Rack-scale AI systems • Accelerator clusters • Multi-GPU servers • AI factories • Open accelerator platforms |
| Key Capabilities | Low-latency accelerator communication • High-bandwidth scale-up fabric • Memory coherency • Open specification • Multi-vendor interoperability • Rack-scale scalability |
| Topics Covered | AI Infrastructure • Scale-Up AI • AI Factories • GPU Networking • Rack-Scale Computing • UALink 1.0 • Accelerator Fabrics |
| Key Companies | AMD • Astera Labs • Broadcom • Cisco • Google • HPE • Intel • Meta • Microsoft |
| Related Technologies | NVLink • PCI Express (PCIe) • Compute Express Link (CXL) • Ultra Ethernet • InfiniBand |
| Standards & Industry Groups | UALink Consortium |
Q&A: UALink 2.0, In-Network Compute, and the Future of Open AI Interconnects
Kurtis Bowman, Chairman, UALink Consortium AI infrastructure is entering a new phase defined by massive GPU clusters, memory bandwidth constraints, and ...
Keysight Targets Scale-Up Bottlenecks with UALink 200G, PCIe 7.0 and CXL 3
Keysight Technologies unveiled a new portfolio of scale-up validation solutions aimed at AI data center operators grappling with bandwidth, latency, ...
Keysight launches 200 Gbps UALink transmitter test solution for AI data center interconnects
Keysight Technologies has introduced its UALink 1.0 Transmitter Compliance Test Solution, enabling automated validation of 200 Gb/s PAM4 transmitters used ...
#AIInfraSummit25: UALink – Accelerating AI Interconnect Innovation
AI Infrastructure Summit 2025 in Santa Clara, California, we sat down with Kurtis Bowman, Chairman of the UALink Consortium, to ...
Hot Interconnects: UALink for Rack-Scale AI Interconnects
At the IEEE Hot Interconnects conference, Sharada Yeluri of Astera Labs outlined why UALink (Ultra Accelerator Link) is emerging as ...
Marvell Debuts Custom UALink Interconnect for Rack-Scale AI Systems
Marvell has introduced a custom UALink scale-up solution aimed at enabling rack-scale AI infrastructure with high compute utilization, low latency, ...
UALink 1.0 Released for Low-Latency Scale-Up AI Accelerators
The UALink Consortium officially released the UALink™ 200G 1.0 Specification, establishing an open industry standard for low-latency, high-bandwidth interconnects designed ...
UALink: Networking Massive GPU Clusters
What is Ultra Accelerator Link (UALink™) and how will it be used to build massive clusters of GPUs? Kurtis Bowman, ...
Data Center Networking for AI and Cloud
The data center networking landscape is experiencing unprecedented transformation driven by the explosive growth of generative AI workloads. Major industry ...
Astera Labs Predictions: Driving Million-Node Clusters
What's driving the evolution of AI infrastructure in data centers for 2025? Thad Omura, Chief Business Officer from Astera Labs ...
A private dossier for networking and telecoms