NVIDIA has entered the CPU market with Vera, its first processor purpose-built for agentic AI workloads. The company shipped initial units to leading AI labs including Anthropic, OpenAI, and Spacebase AI, signaling aggressive momentum in securing early adopter relationships. According to NVIDIA CEO Jensen Huang, speaking at Dell Technologies World, Vera delivers agentic AI inference at one-tenth the cost per token compared to incumbent solutions—a critical metric for enterprises deploying reasoning-heavy agent systems at scale. The Vera NVL72 variant also accelerates agent sandboxes by 50% relative to traditional CPUs while enabling enterprise data queries up to 3x faster.
The launch reflects a strategic shift in NVIDIA's business model beyond reliance on GPU accelerators. Rather than leaving CPU optimization to partners, NVIDIA is now controlling the full stack for emerging workload patterns. Already, 5,000 enterprises including pharmaceutical company Lilly, Samsung, and Honeywell are running AI workloads on NVIDIA infrastructure, creating fertile ground for Vera adoption. The timing aligns with what Jensen Huang described as demand that is going 'parabolic, utterly parabolic'—indicating market conditions favorable for expanding hardware product lines into new categories.
Vera's positioning underscores NVIDIA's broader ecosystem strategy. In parallel, the company continues deepening developer relationships through initiatives like its joint program with Google Cloud, which serves over 100,000 developers through curated learning paths and hands-on labs. By offering purpose-built silicon, NVIDIA tightens developer lock-in to the CUDA ecosystem while establishing new hardware revenue streams. As agentic AI workloads become production-critical, controlling specialized CPUs alongside dominant GPU market share positions NVIDIA to capture significant infrastructure spending from enterprises deploying reasoning-intensive agents.