Nebius, Nscale and Verda build AI infrastructure with Nvidia Vera CPUs

Nvidia said on October 20 that Nebius, Nscale and Verda are building infrastructure around its Vera processors. The cloud providers are targeting the CPU workloads that support AI models, including processing data and executing software tools.

Hardware partners include Bull, Dell Technologies, HPE, Lenovo and Supermicro. Their systems can use Vera independently or combine the processors with Nvidia GPUs. Nvidia also named software companies testing or adopting Vera, including Atlassian, which is evaluating it for preparing indexes used in semantic search.

AI agents create computing demands beyond generating responses. They may compile software, run tests and evaluate results before taking another action. Nscale said slow CPU execution can delay those steps and leave GPUs waiting for the next task.

Nscale had already opened orders for Vera infrastructure in a September 30 announcement. It plans to combine CPU capacity with GPUs, networking and the software needed to coordinate workloads. Nvidia's October announcement did not provide a common deployment timetable for the participating providers.