NVIDIA Vera: The First CPU Built Specifically for AI Agents
NVIDIA's new Vera CPU has been delivered to Anthropic, OpenAI, SpaceXAI, and Oracle Cloud — the first processor purpose-built for the age of agentic AI infrastructure
A New Era in AI Infrastructure
May 2026 marks a pivotal moment in computing history: the arrival of the first CPU purpose-built for agentic AI workloads. NVIDIA founder and CEO Jensen Huang introduced the standalone Vera CPU at GTC San Jose in March as NVIDIA's next multibillion-dollar business. Within months, that CPU moved from NVIDIA's labs into customer hands — a remarkably fast transition that underscores the urgency of the moment.
The first NVIDIA Vera CPUs arrived at three of the world's leading AI labs on Friday — Anthropic in San Francisco, OpenAI in Mission Bay, SpaceXAI in Palo Alto — followed by a delivery to Oracle Cloud Infrastructure in Santa Clara on Monday. NVIDIA Vice President of Hyperscale and High-Performance Computing Ian Buck hand-delivered them personally, an unusual gesture that signals how strategically important these first deployments are for NVIDIA.
"Agentic AI is creating a new CPU moment in the AI factory — as models move from answering to acting, Vera is purpose-built to keep that work moving at scale," Buck said. This is a historic inflection point: for the first time, a CPU is architected entirely around the demands of AI agents rather than general-purpose computing.
Why AI Agents Demand a Different Kind of CPU
AI agents don't run on GPUs alone. Every agentic sandbox, every tool call, every orchestration layer, every long-context retrieval operation — that is CPU work. Traditional processors, designed over decades for general-purpose workloads, simply cannot keep pace with the demands of agentic AI. Vera reimagines this flow from the ground up.
Consider the lifecycle of a single agentic task: an LLM generates a plan, orchestration decomposes it into sub-tasks, tool calls are dispatched, results retrieved, context managed across thousands of tokens — all in sub-second timeframes. Vera's architecture handles this concurrent, real-time workload naturally.
Vera's Architecture: 88 Olympus Cores Deep Dive
At the heart of Vera lies NVIDIA's custom silicon engineering. The chip packs 88 custom-designed Olympus cores, 1.2 TB/s of memory bandwidth, and delivers 50% faster per-core performance under full load. These aren't repurposed server cores — they are purpose-designed for agentic workloads, with specialized instruction paths for concurrent orchestration and data movement.
Deliveries to the Labs: Anthropic, OpenAI, SpaceXAI
At Anthropic, James Bradbury noted: "Scaling compute is an important accelerant for the growth of models. We're excited to see Vera emerge as a promising part of the ecosystem when solving for agentic workloads." At OpenAI, Sachin Katti examined the system as Buck demonstrated its internals — at one point retrieving a screwdriver to reveal the system's architecture. SpaceXAI is evaluating Vera for reinforcement learning workloads and agent-based simulation pipelines.
Oracle Cloud: Deploying Hundreds of Thousands
OCI plans to deploy hundreds of thousands of NVIDIA Vera CPUs beginning in 2026. Karan Batta stated: "Vera's architecture is purpose-built for high-throughput reasoning workloads, delivering the efficiency, density and footprint OCI needs to power the next generation of enterprise AI." OCI is the first cloud provider to deploy Vera at hyperscale.
Industry Significance
NVIDIA CEO Jensen Huang declared that "demand is going parabolic, utterly parabolic." Agentic AI inference at one-tenth the cost per token with Vera Rubin NVL72. Agent sandboxes run 50% faster on Vera than traditional CPUs. For the Georgian market and agencies like SiTech, this means AI agents are moving from experiments to production infrastructure, creating new opportunities for AI-powered services.