NVIDIA Vice President of Hyperscale and High-Performance Computing Ian Buck delivered the first Vera CPUs to Anthropic, OpenAI, and xAI on Friday, signaling a broader shift in NVIDIA’s strategy toward CPU development for artificial intelligence workloads.

Rather than focusing mainly on increasing core counts, the new Vera CPU features 88 custom NVIDIA-designed Olympus cores and delivers roughly 50% faster per-core performance. The processor is designed to meet the growing computational demands of “agentic AI,” where AI systems increasingly perform actions, coordinate tools, retrieve data, and manage complex workflows instead of simply generating responses.

Figure 1. Vera CPU Boosts AI Performance With 50% Faster Per-Core Speed

According to Ian Buck, Vera is specifically engineered to support large-scale AI operations by handling the expanding CPU-side workloads associated with AI agents, while working alongside GPU acceleration systems. Figure 1 shows Vera CPU Boosts AI Performance With 50% Faster Per-Core Speed.

NVIDIA Vera CPU Designed for the Demands of Agentic AI

NVIDIA designed the new Vera CPU with 88 custom Olympus cores, reflecting a strategic move away from simply increasing core counts toward an architecture tailored for the demands of agentic AI systems.

The processor is intended to support AI models that move beyond answering questions and instead perform actions, coordinate workflows, generate code, and interact with tools in real time. These workloads place heavy pressure on CPUs, especially in environments requiring fast, concurrent processing. NVIDIA says Vera addresses this challenge with up to 50% faster per-core performance, enabling more efficient handling of complex AI-driven tasks.

Oracle Cloud Infrastructure is already preparing to deploy hundreds of thousands of Vera CPUs, citing their suitability for large-scale AI reasoning workloads. According to OCI product leader Karan Batta, the architecture is optimized for the efficiency, density, and sustained performance required for enterprise AI systems.

A key capability highlighted by NVIDIA is rapid code generation. As Ian Buck explained, AI models increasingly need to dynamically generate and execute code, such as Python scripts, in order to complete tasks and produce accurate results. Integrated with NVIDIA’s broader AI ecosystem — including Rubin GPUs — Vera is positioned as a core component of next-generation AI infrastructure.

Olympus Core Architecture and Performance Specifications

The rise of agentic artificial intelligence is increasing demand for specialized computing hardware, and NVIDIA designed the Vera CPU specifically to meet those needs. Rather than focusing primarily on maximizing core density, Vera uses 88 custom-built Olympus cores optimized for the CPU-intensive workloads created by AI agents.

These workloads include tool orchestration, data retrieval, long-context processing, and coordination tasks that work alongside GPU acceleration rather than replacing it. NVIDIA says the processor delivers roughly 50% faster per-core performance, helping reduce bottlenecks caused by the continuous and highly concurrent operations required by agentic AI systems.

The CPU also features 1.2 terabytes per second of memory bandwidth, enabling rapid data access and sustained real-time performance under heavy workloads. According to NVIDIA, nearly every agentic sandbox and long-context retrieval operation depends heavily on CPU resources.

James Bradbury of Anthropic noted that expanding compute capacity is a major driver of AI model advancement. Meanwhile, Ian Buck highlighted Vera’s ability to efficiently generate Python code, an increasingly important capability for AI agents handling complex reasoning and task execution.

Vera CPU Deployment Across Initial Labs and Oracle Cloud Infrastructure

Ian Buck personally delivered the first Vera CPUs to Anthropic in San Francisco, OpenAI in Mission Bay, and xAI in Palo Alto before continuing the rollout to Oracle Cloud Infrastructure in Santa Clara.

The hands-on delivery highlights NVIDIA’s focus on supporting early adopters developing agentic AI systems and large-scale AI infrastructure. At xAI, evaluations of the Vera CPU are expected to focus on reinforcement learning and agent-based simulation pipelines [1]. According to the report, Elon Musk asked detailed questions about the processor’s core count, memory architecture, and cooling design.

Buck also emphasized that agentic AI systems frequently require dynamic code generation during operation, making CPU performance increasingly important alongside GPU acceleration.

reference:

1. https://quantumzeitgeist.com/faster-per-core-performance-vera-cpu/

Cite this article:

Janani R (2026), New Vera CPU Delivers Major Performance Gains for AI Computing, AnaTechMaz, pp.975