en.Wedoany.com Reported - WhiteFiber, Inc. has achieved 111.2 Tbps of inter-data center network bandwidth over 83 km of dark fiber, approximately double the performance of comparable full-spectrum field trial results, while utilizing only a portion of the available fiber spectrum. The guaranteed round-trip latency was measured at 0.9 milliseconds, within 8% of the physical limit of light traveling through fiber at that distance. Full fiber illumination testing is scheduled to be completed before the commercial launch in the third quarter of 2026.
This inter-data center network solution connects two geographically separated data centers into a logical single GPU supercluster, enabling the cluster to operate as a unified system. The architecture is designed to support enterprises running AI workloads that exceed the scale, resilience, and compliance capabilities of any single facility. WhiteFiber has filed patent applications for the underlying implementation.
The distributed architecture was built in collaboration with DriveNets, which provides the high-performance network fabric connecting the two sites, and with WEKA NeuralMesh, which delivers high-performance data and memory infrastructure across the cluster. Beyond research and demonstration purposes, the technology's applications include telecommunications, edge computing, and sovereign AI.
WhiteFiber CEO Sam Tabar stated that these results validate that geographic distance does not have to be a constraint for AI infrastructure. Chief Design Engineer Tobias Ford noted that achieving 111.2 Tbps with sub-millisecond latency over 83 km is the result of months of precise and rigorous engineering. WEKA Co-founder and CEO Liran Zvibel remarked that WhiteFiber's distributed cluster technology has the potential to reshape how AI infrastructure is deployed. DriveNets Vice President of AI Product Management Yossi Kikozashvili believes that combining the right network architecture with engineering ambition can deliver high-performance architectures.
WhiteFiber stated that details regarding the commercial launch plan, availability, architecture design, and pricing of the solution will be announced in the third quarter of 2026.
