At GTC Taipei, NVIDIA announced that the global technology sector is aligning to adopt NVIDIA Vera, marking the debut of the world’s first central processing unit (CPU) engineered from the ground up for the age of agentic AI. Now entering full production, this innovative processor class is built to deliver up to 1.8x faster task completion compared to traditional x86 CPUs. By accelerating multi-step workloads, including agentic AI, reinforcement learning, and data processing, the platform is designed to generate increased data center token revenue for enterprise infrastructures.
The launch builds upon the market success of NVIDIA Grace™ CPUs, which have reached nearly 2.5 million shipments to date. The introduction of Vera establishes new benchmarks for CPU performance and energy efficiency within modern data centers. This paradigm shift addresses a major structural evolution in corporate computing, where autonomous AI agents are moving beyond basic conversational queries to run independent code, execute complex planning loops, interact with sandboxed data, and independently validate computational outcomes.
Early evaluation and adoption of the Vera CPU architecture spans prominent global AI innovators like OpenAI, Anthropic, and SpaceXAI, alongside financial infrastructure leaders such as the New York Stock Exchange (NYSE). Major hyperscalers including ByteDance, CoreWeave, Lambda, Nebius, Nscale, and Oracle Cloud Infrastructure (OCI) are actively planning integration, supported by a production ecosystem of top-tier hardware manufacturers including Dell Technologies, HPE, Lenovo, and Supermicro.
“AI agents will be the largest users of computing,” said Jensen Huang, founder and CEO of NVIDIA. “Vera is the first CPU designed for that future built to run agentic AI at hyperscale with extraordinary performance, efficiency and programmability.”
Modern Real-Time Infrastructures and Hyperscale Supercomputing
For systems managing vast amounts of telemetry and high-velocity transactional data, the reliance on high-throughput, low-latency infrastructure is a necessity.
“At the NYSE, our focus is to optimize the latency, throughput and reliability of the systems underpinning our unrivaled infrastructure,” said Lynn Martin, president of NYSE Group. “The NYSE processes more than 1.1 trillion messages per day, and in collaboration with Redpanda and HPE, using NVIDIA Vera CPUs, we will be scaling our capacity while further optimizing latency to power a high-performance, resilient and AI-ready market infrastructure.”
Similarly, cutting-edge AI labs are turning to dedicated processing hardware to handle the heavy algorithmic loads required by advanced large language models and autonomous coworkers.
“Scaling compute is an important accelerant for the growth of models,” said James Bradbury, head of compute at Anthropic. “We’re excited to see Vera emerge as a promising part of the ecosystem when solving for agentic workloads.”
Also Read: HPE Unveils Next-Gen ProLiant Server Powered by NVIDIA Vera CPU to Drive Agentic AI Infrastructure
To satisfy the surging industrial requirements for enterprise training, inference, and real-time operations, cloud providers are integrating these processors into high-density architectures to optimize core operational pipelines.
“Oracle Cloud Infrastructure is rapidly scaling AI infrastructure to meet surging demand for training, inference and agentic AI,” said Mahesh Thiagarajan, executive vice president of Oracle Cloud Infrastructure. “By deploying NVIDIA Vera CPUs, OCI will support high-throughput reasoning and data processing workloads across next-generation AI environments.”
Eliminating the Agentic Sandbox Bottleneck
According to independent benchmark evaluations by Phoronix, a comprehensive open-source benchmarking suite, the NVIDIA Vera CPU recorded the fastest overall industry performance across critical agentic workloads including intensive Python executions, Java environments, code compilation, and multi-tier database processing.
These operations frequently act as the critical path within modern AI factories. Because autonomous agents utilize iterative loops to write code, invoke software tools, and run sandboxed tests, localized CPU speeds dictate overall agent throughput and interactivity. By accelerating these computational bottlenecks by up to 1.8x over conventional x86 alternatives, Vera dramatically maximizes token output and efficiency.
Beyond standalone applications, Vera provides system-level efficiency as the baseline host CPU for next-generation multi-node architectures, including the NVIDIA Vera Rubin NVL72 and HGX™ Vera Rubin NVL8 platforms. The chip actively balances GPU computing nodes by managing data ingestion pipelines (ETL), executing key-value (KV) cache management, and governing complex network orchestrations.
Architectural Foundations of the Vera Platform
The performance metrics of the Vera architecture are anchored by two foundational hardware innovations:
- NVIDIA Custom Olympus Cores: The processor features 88 custom-engineered Olympus cores tailored for latency-sensitive, control-heavy enterprise logic. To ensure high single-thread execution speeds, the design introduces NVIDIA Spatial Multithreading. This framework creates 176 logical threads utilizing strictly partitioned core resources, preventing performance degradation and ensuring predictable execution throughput at scale.
- High-Bandwidth LPDDR5X Memory Subsystem: Vera achieves up to 1.2 terabytes per second (TB/s) of aggregate memory bandwidth by integrating low-power double data rate 5X (LPDDR5X) memory. This architecture provides twice the operational bandwidth at half the power consumption of traditional enterprise memory solutions. Supporting capacities up to 1.5 TB, the subsystem maintains ultra-responsive conditions for thousands of parallel sandbox environments, data analytics pipelines, and high-performance computing (HPC) workflows.
Expanding its utility across the full data center lifecycle, the processor architecture also extends directly to high-capacity storage environments through integration with the newly announced NVIDIA Vera BlueField®-4 STX platform.
