WhiteFiber has announced that initial research and development (R&D) testing for its new cross-data-centre networking solution achieved 111.2 terabits per second (Tbps) across 83km of dark fibre.
It recorded a round-trip latency of 0.9 milliseconds, which is within 8% of the physical speed-of-light limit for fibre over that distance.
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
The data throughput reached in these tests is approximately double that of previously published full-spectrum field trials, despite utilising only part of the available fibre spectrum.
The networking solution links two geographically separated data centres, allowing them to function as a unified graphics processing unit (GPU) supercluster.
WhiteFiber stated that this architecture enables enterprises to process AI workloads with greater scale, resilience, and compliance than could be managed by a single facility. Patent applications have been submitted for the underlying implementation.
Technology partners involved in the R&D included DriveNets, which supplied the network fabric linking both sites, and WEKA NeuralMesh that provided data and memory infrastructure across the cluster.
WhiteFiber CEO Sam Tabar said: “These results validate what we set out to prove: that geographic distance does not have to be a constraint on AI infrastructure. This is the foundation for a new class of AI compute, one that delivers the performance of a single supercluster with the resilience and flexibility of a distributed system. We are excited to bring this capability to market.”
DriveNets AI vice-president product management Yossi Kikozashvili said: “What WhiteFiber has built demonstrates what becomes possible when you pair the right networking architecture with genuine engineering ambition.
“Our Ethernet-based fabric supports the highest performance in the most demanding, high-bandwidth, low-latency environments, ensuring optimised utilisation of GPU and power resources in a scale-across supercluster architecture. These results are a direct reflection of that.”
WhiteFiber owns high-performance computing data centres and offers cloud, colocation, and hosting services tailored for generative AI tasks.
Full-fibre testing is planned before commercial launch, which is expected in the third quarter of 2026.
