DriveNets Announces Commercial Deployment of Long-Range Scale-Across AI Supercluster

Lod, Israel – July 9, 2026 – DriveNets, a leader in large-scale networking solutions, today announced the commercial deployment of an AI supercluster with long-range scale-across AI networking. As part of WhiteFiber Inc., a leading AI infrastructure solutions provider, launching 'Project Redwood' [https://www.whitefiber.com/news?id=39], DriveNets AI Fabric has connected two WhiteFiber H200 GPU clusters located 52 miles apart into a single logical GPU supercluster, with verified performance of 111.2 Tbps bandwidth and 0.9 milliseconds of guaranteed latency. While scale-across architecture has been widely discussed across the industry, DriveNets has now demonstrated this concept not in a lab, but at production scale, delivering it as a commercially deployed network.

Figure: Conceptual illustration of scale-across architecture enabled by DriveNets AI Fabric. Two sites up to 80 km apart are interconnected via dedicated leaf switches, forming a single unified cluster.

Addressing AI Power Constraints with Scale-Across Networking

Building AI infrastructure is increasingly constrained not by compute power, but by the power and space available at a single site. Scale-across architecture removes this limitation. Instead of being bound by the power envelope of one facility, AI infrastructure builders can now extend clusters to remote sites, operating distributed GPUs not as two separate environments, but as one integrated system. This enables larger clusters, higher fault tolerance, and the freedom to build where power is available—without sacrificing performance.

Extending clusters across distance presents a far more complex networking challenge than simply laying cables between two sites. Links connecting remote sites typically have lower bandwidth than intra-facility fabrics, leaving little margin to absorb bursty traffic before congestion occurs. AI training intensifies this challenge. It generates a small number of extremely large flows that arrive in synchronized bursts, rather than many small, steady flows—traffic patterns not anticipated by traditional data center load balancing and buffering techniques. Without a fabric designed to absorb these bursts and manage congestion in real time, latency spikes and packet loss occur, causing jobs to stall and leaving GPUs idle on both sides of the cluster. Solving this over long distances—without performance degradation—is precisely why scale-across architecture, and the switching, buffering, and congestion management technologies that enable it, are critical for the next phase of AI infrastructure growth.

Commercial Deployment of Scale-Across Solution

WhiteFiber's Project Redwood connects two geographically separated GPU clusters into a single logical GPU supercluster, with the DriveNets AI Fabric solution providing the high-performance network linking the two sites.

Ido Susan, Co-Founder and CEO of DriveNets, stated: "Power availability has been a major constraint on AI infrastructure growth, but with this proven deployment, that constraint is now eliminated. Together with WhiteFiber, we've moved scale-across from concept to commercial reality, demonstrating that two remote data centers can function as a single high-performance supercluster. We believe much of the next generation of AI infrastructure will be built this way."

Sam Tabar, CEO of WhiteFiber, said: "DriveNets AI Fabric played a critical role in proving that Project Redwood can deliver single-site cluster performance and reliability across two locations. This milestone shows that the scale of AI infrastructure we build no longer needs to be limited by geography."

As part of the validation process, performance between GPU racks within a single site was compared to performance between GPU racks with one rack at the primary site and the other at the remote site. Detailed methodology and results are available in DriveNets' white paper (https://drivenets.com/resources/white-paper/scaling-ai-workload-clusters-across-multi-site-deployments/).

Lossless Performance Beyond the Data Center Walls

Traditional Data Center Interconnect (DCI) links are not designed for AI workloads that generate bursts of traffic intolerant to jitter or packet loss. Even minor losses during training can delay job completion and waste expensive GPU cycles. DriveNets' 9300F, 5300R, and 5301R switches, powered by Fabric Scheduled Ethernet (FSE) technology, extend AI fabrics beyond a single data center through cell-based load balancing, end-to-end Virtual Output Queuing (VOQ), and interconnected deep buffers that absorb AI traffic bursts before congestion occurs. The result is predictable, lossless inter-site connectivity that keeps GPU utilization high across the entire cluster, as if all components were under one roof. Achieving both industry-leading performance and zero packet loss is no accident—it is a direct result of DriveNets' purpose-built architecture, which is why DriveNets AI Fabric is the optimal networking solution for geographically distributed AI clusters.

For more information on scaling AI clusters across multi-site deployments, visit www.drivenets.com.

About DriveNets

DriveNets is a leader in large-scale networking software for AI infrastructure and service providers. The company transforms the economics of large networks while maximizing performance, availability, and operational efficiency.