News Summary
- The NVIDIA Vera CPU delivers twice the efficiency and 50% higher performance than traditional CPUs, enabling faster, more cost-effective computing for demanding workloads.
- NVIDIA is working with clients in the global cloud, AI, and enterprise sectors to deploy the Vera CPU.
- Manufacturers across the industry have already started using the Vera CPU in their systems.
At GTC, NVIDIA launched the Vera CPU, the world’s first processor for agentic AI and reinforcement learning, offering twice the efficiency and 50% more speed than traditional rack-scale CPUs.
As reasoning and agentic AI improve, the infrastructure behind these models becomes more important for scaling performance and cost. This includes systems that plan tasks, run tools, interact with data, execute code, and check results.
The NVIDIA Vera CPU builds on the NVIDIA Grace CPU, allowing organizations of any size to create AI factories that use agentic AI at scale. Vera offers top single-thread performance and per-core bandwidth, making it ideal for large-scale AI services such as coding assistants and both consumer and enterprise agents.
Major cloud providers like Alibaba Cloud, Coreweave, Meta, and Oracle Cloud Infrastructure, along with system makers such as Dell Technologies, HPE, Lenovo, and Supermicro, are working with NVIDIA to deploy Vera. This widespread adoption positions Vera as the new standard for critical AI workloads, making AI easier to use and accelerating innovation for developers, startups, institutions, and businesses.
Vera is arriving at a turning point for AI as intelligence becomes agentic, capable of reasoning and acting. The importance of the systems orchestrating that work is elevated, said Jensen Huang, founder and CEO of NVIDIA. The CPU is no longer simply supporting the model. It’s driving it with breakthrough performance and energy efficiency. Vera unlocks AI systems that think faster and expand further.
Configurable for Every Data Center
NVIDIA announced a new Vera CPU rack that holds 256 liquid-cooled Vera CPUs. This setup can support over 22,500 CPU environments running at full performance. At the same time, AI factories can quickly scale up to tens of thousands of instances of agentic tools in one rack.
The new Vera rack uses NVIDIA MGX modular reference architecture, a flexible blueprint for building different server configurations, and is supported by AT partners around the world.
On the NVIDIA Vera Rubin NVL72 platform, Vera CPUs connect to NVIDIA GPUs using NVIDIA NVLink C2C interconnect technology, a high-speed connection that lets CPUs and GPUs share data quickly, offering 1.8 TB/s of bandwidth, which is seven times more than PCIe Gen 6. This allows for fast data sharing between CPUs and GPUs. NVIDIA also introduced new reference designs that use Vera as the main GPU for NVIDIA HGX Rubin NBL8 systems, managing data movement and system control for GPU-accelerated tasks.
Vera system partners offer both dual- and single-socket CPU server setups. These are ideal for tasks such as reinforcement learning, agentic inference, data processing, orchestration, storage management, cloud applications, and high-performance computing.
Across all configurations, Vera systems include NVIDIA, ConnectX, Supermicro cards and BlueField 4 DPUs, delivering high-speed networking, storage, and security. Key benefits for agentic AI: customers can optimize performance with a single software stack across the NVIDIA platform, while high-performance CPU cores, high-bandwidth memory, and an advanced coherency fabric ensure quick responses even under heavy agentic workflows and Reinforcement learning.
Vera has 88 custom NVIDIA-designed Olympus cores that provide strong performance for compilers (software that translates programming code), runtime engines (systems that execute application code), analytics pipelines (processes for analyzing data), agentic tools (AI tools that perform tasks independently), and orchestration services systems that coordinate complex processes. Each core can handle two tasks at once using n-media spatial multi-threading (a technology that lets a single-core CPU execute multiple instructions simultaneously), ensuring steady, predictable performance. This is ideal for AI factors that run many jobs at the same time.
Vera also improves energy efficiency with the second generation of NVIDIA’s low-power memory system, now using LPDDR5X (a high-performance, low-power memory type). This provides up to 1.2 TB of bandwidth, which is twice that of general-purpose CPUs, and uses half the power.
Widespread Ecosystem Support
Cursor, a company focused on AI-native software development, is using NVIDIA Vera to improve performance of its AI coding agents.
We are excited to use NVIDIA Vera CPUs to improve overall throughput and latency so we can deliver faster, more responsive coding agent experiences for our customers, said Michael Truell, the co-founder and CEO of Cursor.
Redpanda, a top streaming data platform and AI platform, is using Vera to greatly improve performance.
Redpanda recently tested NVIDIA Vera running Apache Kafka–compatible workloads and saw dramatically better performance than other systems in a benchmark. It delivered up to 5.5x lower latency, said Alex Gallego, founder and CEO of Redpanda. Vera represents a new direction in CPU architecture with more memory and less overhead per core. This enables our customers to scale real-time streaming workloads further than ever and unlock new AI and agentic applications.
National labs planning to use Vera CPUs include the Leibniz Supercomputing Center, Los Alamos National Laboratory, Lawrence Berkeley National Laboratory, National Energy Research Scientific Computing Center, and the Texas Advanced Computing Center (TACC).
At TACC, we recently tested NVIDIA’s Vera CPU platform as we prepared for launching our upcoming Horizon system and running six of our scientific applications. We saw impressive early results, said John Cazes, Director of High-Performance Computing at TACC. Vera’s per-core performance and memory throughput represent a giant leap forward for scientific computing, and we look forward to bringing Vera-based nodes to our CPU users on Horizon later this year.
Leading cloud service providers planning to deploy Vera CPUs including Alibaba Cloud, ByteDance, Cloudflare, CoreVive, Curso, Lombardo, Nebus, NScale, Oracle Cloud Infrastructure, Together.ai, and Vultr.
Leading infrastructure providers adopting various APIs include Aivres, ASRock, Rack, Asus, Compel, Cisco, Dell, Foxconn, Gigabyte, HPE, HYVE, InventTech, Lenovo, MiTAC Computing, MSI, Pegatron, Quanta Cloud Technology, QCT, Supermicro, Wistron and Wiwynn.
Source: NVIDIA Launches Vera CPU, Purpose-Built for Agentic AI
