NVIDIA Vera CPU Sets New Standards for AI Agent Workloads

NVIDIA's Vera CPU has emerged as a groundbreaking solution for AI infrastructure, combining high throughput with exceptional energy efficiency. Designed for agentic AI workloads, Vera enables businesses to maximize performance while minimizing power consumption, a critical factor for power-limited AI factories.

Performance and Efficiency Benchmarks

In a 100-megawatt data center comparison, NVIDIA reported that Vera delivers 1.6 times more agents per watt than the fastest x86 CPU and achieves twice the agent speed of the most energy-efficient x86 CPU. These results highlight Vera's ability to support more concurrent AI agents without compromising speed or efficiency.

Early adopters have validated these claims in production-style tests. For example, Daytona achieved 1,000 agent sandboxes online in just 10.8 seconds using Vera, compared to 88.4 seconds on x86 CPUs. Additionally, Daytona recorded 12.7 completed jobs per second at 2,000 concurrent sandboxes with Vera, outperforming x86 alternatives. DeepInfra reported up to 2.2 times faster agent orchestration speeds compared to leading x86 processors, while ClickHouse observed near-linear scaling across Vera's 176 cores, placing it atop the ClickBench hardware leaderboard.

Technical Specifications

Vera features 88 custom Olympus cores, supporting 176 threads through NVIDIA Spatial Multithreading. It includes 2 MB of L2 cache per core, 164 MB of unified L3 cache, up to 1.5 TB of LPDDR5X memory, and a memory bandwidth of up to 1.2 TB/s. The CPU operates within a configurable thermal design power range of 250–450 watts, emphasizing sustained per-core execution under full-socket loads.

NVIDIA's second-generation coherency fabric and LPDDR5X subsystem enable more than 90% of peak memory bandwidth utilization, with memory consuming less than 30 watts—significantly lower than DDR5 configurations.

Adoption and Real-World Impact

Several organizations have already integrated Vera into their operations. Perplexity achieved 1.5 times faster agentic coding tasks compared to traditional CPUs and plans to deploy Vera in an upcoming production system. Modal reported up to 2 times sandbox throughput per core versus comparable x86 hardware, with performance gains increasing under heavy workloads. DuckLabs observed 1.5 times the median TPC-H SF1000 performance of an Intel Xeon 6 baseline.

For infrastructure operators, Vera's efficiency translates to reduced hardware requirements. Daytona estimated that sustaining 1,000 jobs per second would require approximately 15% fewer Vera machines than EPYC 9755 machines and 38% fewer than EPYC 9575F machines for the same workload.

Strategic Implications

As AI factories continue to scale, the demand for CPUs capable of balancing speed, capacity, and energy efficiency will grow. NVIDIA's Vera CPU positions itself as a versatile solution for these evolving needs, offering a unified platform for diverse agent workloads without the need for mixed x86 configurations.