WhiteFiber has introduced Continuum, a distributed GPU supercluster architecture that ties together compute resources across separate data centers. The company frames it as a networking solution that removes bandwidth as a bottleneck for AI infrastructure, allowing enterprises to scale workloads beyond a single facility. The product is now generally available, according to the announcement.
Continuum is aimed at organizations that need large-scale AI training or inference but cannot or do not want to build everything in one physical location. By pooling GPUs over a cross-data-center network, the architecture treats remote resources as if they were local. WhiteFiber claims this changes the tradeoffs around where and how AI infrastructure is deployed.
The announcement does not include technical specifications or performance benchmarks, so the practical limits of the system remain unclear. Still, the move signals growing commercial interest in distributed AI compute, particularly for enterprises that want flexibility in siting GPU capacity without sacrificing cluster coherence.