Delivering Vera: NVIDIA’s First CPU Built for Agents Is Shipping Now

AI NEWS

Delivering Vera: NVIDIA’s First CPU Built for Agents Is Shipping Now

NVIDIA has officially shipped its first custom CPU, the 'Vera,' designed specifically for the emerging era of agentic AI. The hardware was hand-delivered to major partners including AWS, Oracle Cloud Infrastructure (OCI), OpenAI, Anthropic, and SpaceXAI. Unlike traditional CPUs focused on core density, Vera features 88 custom 'Olympus' cores optimized for the high-throughput reasoning and tool-calling workloads required by AI agents that act rather than just answer.

THE NEWS

What happened

NVIDIA has officially shipped its first custom CPU, the 'Vera,' designed specifically for the emerging era of agentic AI. The hardware was hand-delivered to major partners including AWS, Oracle Cloud Infrastructure (OCI), OpenAI, Anthropic, and SpaceXAI. Unlike traditional CPUs focused on core density, Vera features 88 custom 'Olympus' cores optimized for the high-throughput reasoning and tool-calling workloads required by AI agents that act rather than just answer.

CONTEXT

Why it matters

NVIDIA is launching a new era with the Vera CPU, its first custom processor designed specifically for agentic AI. The chip was recently hand-delivered to major cloud providers and AI labs like AWS and Oracle. With 88 custom Olympus cores, it handles the heavy lifting of AI agents that act and reason, offering up to 1.8x faster performance and significantly better energy efficiency.

AT A GLANCE

Key facts

  • NVIDIA delivered its first Vera CPU server to AWS in Seattle alongside the new Vera Rubin GPU.
  • Vera is NVIDIA's first custom CPU purpose-built for agentic AI workloads, featuring 88 Olympus cores.
  • The chip delivers up to 1.8x faster per-core performance on agentic AI tasks compared to traditional designs.
  • Oracle Cloud Infrastructure plans to deploy hundreds of thousands of Vera CPUs starting in 2026.
  • Vera serves as the host processor for the NVL72 system, pairing with Rubin GPUs via second-generation NVLink-C2C.
  • The architecture supports a unified memory design that improves energy efficiency by 2x over traditional infrastructure.

SOURCE

Original source

This article is based on information published by NVIDIA AI.