Nvidia Vera CPU challenges x86 with 1.8x faster agentic AI

Nvidia's first CPU designed for AI agents enters production, threatening Intel and AMD's server stronghold with 1.8x faster task completion than x86 processors.

Nvidia's Vera CPU, the company's first processor built specifically for agentic AI workloads, delivers 1.8x faster task completion than x86 alternatives, challenging Intel and AMD's decades-long dominance in data center compute as AI factories scale.

"AI agents will be the largest users of computing," Jensen Huang, founder and chief executive officer of Nvidia, said. "Vera is the first CPU designed for that future — built to run agentic AI at hyperscale with extraordinary performance, efficiency and programmability."

Vera packs 88 custom Olympus cores with Spatial Multithreading and a LPDDR5X memory subsystem delivering 1.2TB/s of bandwidth. It succeeds Nvidia's Grace CPU, which has shipped nearly 2.5 million units. Phoronix benchmarks show Vera leading across agentic workloads including code compilation, Python, Java and database processing. The chip serves as the host CPU for Vera Rubin NVL72 systems via second-generation NVLink-C2C interconnect at 1.8TB/s of coherent bandwidth, and also powers the Vera BlueField-4 STX processor for AI storage platforms.

The launch marks Nvidia's most direct assault on Intel and AMD's server CPU franchise, a market valued at roughly $30 billion annually. Vera systems from Dell Technologies, HPE, Lenovo and Supermicro will be available starting this fall, with cloud deployments from CoreWeave, Oracle Cloud Infrastructure and Lambda among the first adopters.

Vera's Architecture Targets the Agentic AI Bottleneck

AI factory economics are shifting from cores per dollar to tokens per dollar, requiring CPUs that complete orchestration and data-processing work faster. Vera's Olympus cores are engineered for Python runtimes, sandboxed code execution and analytics pipelines — the CPU-bound steps that keep GPUs idle. The chip's Spatial Multithreading allows it to anticipate application behavior and move data across large numbers of concurrent environments, reducing latency in multi-step agent workflows where one prompt can launch a thousand-step reasoning journey.

The Vera CPU can be deployed across the full AI factory — from standalone CPU infrastructure to tightly coupled accelerated systems. It extends Nvidia's Confidential Computing at rack scale, encrypting data across high-speed interconnects with hardware-level attestation to ensure tamper-proof execution for proprietary models and regulated data.

Broad Adoption Puts Pressure on Incumbents

Nvidia has lined up extensive support. Customers evaluating or deploying Vera include NYSE, Anthropic, OpenAI, SpaceXAI, ByteDance, CoreWeave, Lambda, Nebius, Nscale and Oracle Cloud Infrastructure. System builders offering Vera-based systems span Aivres, ASRock Rack, ASUS, Compal, Dell, Foxconn, GIGABYTE, HPE, Inventec, Lenovo, MiTAC Computing, MSI, Pegatron, Quanta Cloud Technology, Supermicro, Wistron and Wiwynn — 18 manufacturers in total.

The Vera Rubin platform, which pairs Vera CPUs with Rubin GPUs, delivers 10x agent throughput at scale compared with the prior-generation Grace Blackwell platform. It also introduces Spectrum-X Ethernet Photonics with co-packaged optics, enabling million-GPU AI factories with 5x better power efficiency than traditional transceivers. Nvidia's supply chain spans 150 partners in Taiwan alone across 350 factories in 30 countries.

For investors, the question is whether Vera can capture meaningful share of the server CPU market. Nvidia shares trade at roughly 35x forward earnings. Intel and AMD have dominated this segment for decades, but the shift to agentic AI workloads creates a new architecture requirement that favors purpose-built silicon. Production shipments begin this fall.

This article is for informational purposes only and does not constitute investment advice.