Compute Cluster
A supercomputing facility composed of large numbers of GPUs or TPUs connected through high-speed interconnects, used for large-scale AI model training and inference, with a single cluster's cost potentially reaching billions of dollars.
Timeline (last 90 days)
A compute cluster refers to a supercomputing facility composed of tens of thousands or even hundreds of thousands of GPUs/TPUs connected by high-speed interconnects, with a single cluster costing up to billions of dollars, and only a few entities worldwide can afford it.
Building a gigawatt-scale compute cluster requires committing enormous capital expenditure (CapEx) upfront.
All Facts (2)
A compute cluster refers to a supercomputing facility composed of tens of thousands or even hundreds of thousands of GPUs/TPUs connected by high-speed interconnects, with a single cluster costing up to billions of dollars, and only a few entities worldwide can afford it.
50%UnverifiedBuilding a gigawatt-scale compute cluster requires committing enormous capital expenditure (CapEx) upfront.
50%