NVIDIA announced its new Vera CPU designed specifically to handle the demanding CPU-bound tasks of AI agents, such as searching databases, executing code, and managing multi-step workflows alongside GPUs. According to the company, AI factories face strict power constraints, requiring infrastructure that maximizes throughput per watt without compromising processing speed.

NVIDIA stated that in production-style tests and a 100-megawatt data center comparison, a single Vera CPU outperformed both the fastest and most energy-efficient x86 CPUs. Early adopters have already begun deploying the processor in production environments to streamline agentic workloads without mixing x86 variants.

Why it matters

  • Provides data center operators high agent throughput per watt, optimizing power-constrained AI factory capacity.

  • Simplifies server architecture by substituting mixed x86 deployments with a single CPU type tailored for agentic workflows.

  • Reduces execution latency for CPU-heavy AI tasks like code execution and database queries.

Source: blogs.nvidia.com