NVIDIA launches Vera CPU designed for agentic AI and reinforcement learning

NVIDIA introduced the Vera Central Processing Unit (CPU), a processor specifically crafted to support agentic Artificial Intelligence (AI) and reinforcement learning workloads. The new CPU offers twice the efficiency and completes tasks 50% faster compared to standard rack-scale CPUs, according to the company.

The Vera CPU addresses growing demands for infrastructure capable of managing AI models that perform reasoning, planning, tool use, data interaction, coding, and result validation. By enhancing single-thread performance and bandwidth per core, it aims to improve AI throughput, responsiveness, and energy use in large-scale AI services including coding assistants and automated agents.

The architecture integrates 88 custom Olympus cores supporting simultaneous multithreading, a High Bandwidth Memory (HBM) subsystem built on LPDDR5X memory delivering up to 1.2 TB/s bandwidth, and NVIDIA's second-generation Scalable Coherency Fabric. These design features intend to support the high utilization conditions typical of agentic AI and reinforcement learning applications.

NVIDIA unveiled a new Vera CPU rack that combines 256 liquid-cooled CPUs, capable of sustaining over 22,500 concurrent CPU environments operating independently at full performance. The modular design employs the NVIDIA MGX reference architecture and features NVLink-C2C interconnect technology providing 1.8 TB/s of coherent bandwidth between CPUs and GPUs. System configurations include single- and dual-socket servers suited for workloads such as reinforcement learning, agentic inference, data processing, orchestration, storage management, cloud applications, and High performance computing (HPC).

Organizations collaborating on or planning Vera deployment include Alibaba, ByteDance, Meta, Oracle Cloud Infrastructure, CoreWeave, Lambda, Nebius, and Nscale. Hardware manufacturers adopting Vera comprise Dell Technologies, HPE, Lenovo, Supermicro, ASUS, Compal, Foxconn, GIGABYTE, Pegatron, Quanta Cloud Technology, Wistron, and Wiwynn. Additional user feedback from Redpanda and the Texas Advanced Computing Center highlights performance improvements in latency and per-core metrics.

Jensen Huang, NVIDIA founder and CEO, said, “Vera is arriving at a turning point for AI. As intelligence becomes agentic — capable of reasoning and acting — the importance of the systems orchestrating that work is elevated. The CPU is no longer simply supporting the model; it’s driving it. With breakthrough performance and energy efficiency, Vera unlocks AI systems that think faster and scale further.” The Vera CPU entered full production with availability through partners expected in the latter half of the year.