With hyperscale adoption planned from 2026, NVIDIA positions Vera as a foundational processor for enterprise and cloud AI deployments.
NVIDIA has officially begun deploying its new Vera CPU, a processor purpose-built to meet the growing computational demands of agentic artificial intelligence. First introduced by CEO Jensen Huang at the company’s GTC conference, Vera represents NVIDIA’s entry into custom CPU design tailored specifically for AI systems capable of reasoning, planning, and executing complex tasks autonomously.
While GPUs remain central to AI training and inference, agentic AI introduces a new layer of computing requirements handled primarily by CPUs. Tasks such as orchestration, tool execution, long-context memory management, reinforcement learning environments, simulation workflows, and large-scale data coordination rely heavily on CPU performance. Vera was designed from the ground up to manage these persistent, real-time workloads efficiently.
The first Vera systems have already been delivered to leading AI organizations including Anthropic, OpenAI, and SpaceXAI, with further deployment taking place at Oracle Cloud Infrastructure. Oracle plans to scale deployments significantly beginning in 2026, highlighting growing industry demand for CPU architectures optimized for AI factories and hyperscale environments.
Technically, Vera features 88 custom Olympus CPU cores, delivers 1.2 terabytes per second of memory bandwidth, and achieves roughly 50% higher per-core performance under sustained workloads. These capabilities allow AI systems to process concurrent operations faster, improving responsiveness and overall infrastructure efficiency.
Beyond standalone servers, Vera also acts as a host processor within NVIDIA’s broader accelerated computing ecosystem, coordinating data flow and feeding GPUs more effectively through unified memory and high-speed interconnect technologies. This design improves utilization while reducing energy consumption compared with traditional heterogeneous systems.
As AI evolves toward autonomous agents performing continuous reasoning and action-based workflows, CPUs are becoming central to AI infrastructure again. With Vera, NVIDIA aims to redefine the role of the CPU as a core engine powering scalable, production-grade agentic AI deployments worldwide.
