NVIDIA Ships First Vera CPUs to 4 AI Customers, Targeting Oracle's 2026 Hyperscale Rollout
Updated
Updated · NVIDIA Blog · May 19
NVIDIA Ships First Vera CPUs to 4 AI Customers, Targeting Oracle's 2026 Hyperscale Rollout
10 articles · Updated · NVIDIA Blog · May 19
Friday deliveries put NVIDIA's first standalone Vera CPU systems into Anthropic, OpenAI and SpaceXAI, with Oracle Cloud Infrastructure receiving its system on Monday.
Vera is pitched as a CPU built for agentic AI's orchestration-heavy workloads, with 88 custom Olympus cores, 1.2 TB/s memory bandwidth and 50% faster per-core performance.
Oracle said it plans to deploy hundreds of thousands of Vera CPUs starting in 2026, making OCI the first cloud provider to commit to hyperscale rollout.
The launch extends NVIDIA's push beyond GPUs into a new multibillion-dollar CPU business, with Vera also serving as the host processor in Vera Rubin NVL72 systems.
As NVIDIA's new CPU redefines AI infrastructure, what will be the next performance bottleneck for developers?
With AI labs also buying rival chips, can NVIDIA’s integrated ecosystem truly conquer the agentic AI market?
NVIDIA Vera CPUs and Rubin Platform: Transforming Agentic AI Infrastructure in 2026
Overview
In May 2026, NVIDIA launched the Vera CPU, a purpose-built processor designed to power agentic AI. This marks a major shift as advanced AI systems move from concept to large-scale deployment, thanks to specialized hardware tailored for autonomous AI agents. The Vera CPUs are engineered to meet the demanding needs of AI labs, cloud providers, and enterprises. Early adoption is evident, with Oracle showcasing the Vera system in its AI Customer Excellence Center, where executives reviewed the hardware as NVIDIA GPU racks processed customer AI workloads in real time. This launch signals a new era for practical, scalable agentic AI.