Compute Cloud@Customer introduces GPU support

To build a scalable platform for AI and graphics intensive applications with Compute Cloud@Customer, add a GPU expansion rack to your deployment. The minimum configuration contains a single GPU node with 4 NVIDIA L40S GPUs. Up to 12 GPU nodes can be added incrementally by connecting two full GPU expansion racks to the base installation.

GPU nodes are added to the same 3 fault domains, but the server families operate separately. Users must deploy compute instances with a dedicated shape to take advantage of the GPUs. These instances do not support live migration.

For more information, see GPU Expansion.