NVIDIA Vera CPU Targets Faster, More Efficient AI Agent Workloads
NVIDIA says its Vera CPU can combine high-throughput capacity with fast performance for AI agents, which move repeatedly between GPU reasoning and CPU-based execution. In a 100-megawatt data center comparison, the company says Vera outperformed both the fastest and the most energy-efficient x86 CPUs on its measures.
- NVIDIA says Vera delivers 1.6x more agents per watt than the fastest x86 CPU.
- The company reports 2x faster agent speed than the most energy-efficient x86 CPU.
- Perplexity ran agentic coding tasks 1.5x faster with Vera than on traditional CPUs.
- Daytona launched 1,000 sandboxes in 10.8 seconds using Vera.
- Modal reported up to 2x sandbox throughput per core versus comparable x86 hardware.
A CPU built for agent workloads
AI agents perform more than answer generation. According to NVIDIA, they search for information, query databases, write and test code, and take actions. Each step relies on a CPU, while an agent can move between GPU reasoning and CPU execution during a task.
As businesses use more agents, AI infrastructure providers need CPUs that can handle many workloads at once without making each agent wait longer. NVIDIA also describes AI factories as power limited, making work completed per watt a central consideration for operators.
The company positions Vera as a single CPU for a range of agent workloads. That approach could reduce the need to combine fast and high-density x86 parts when infrastructure providers are balancing responsiveness against capacity.
NVIDIA reports performance and efficiency gains
NVIDIA says Vera delivered industry-leading agent speed while enabling 1.6x more agents per watt than the fastest x86 CPU in its 100-megawatt data center comparison. The company also says Vera delivered 2x faster agent speed, with leading energy efficiency, compared with the most energy-efficient x86 CPU.
These figures are NVIDIA's own comparison claims. The source does not identify the x86 processors used by name, but frames the results around two different infrastructure priorities: maximum speed and energy-efficient capacity.
Early adopters test Vera in production-style workloads
NVIDIA says partners have already tested Vera across agentic coding, orchestration, databases, and sandboxed workloads. Perplexity ran agentic coding tasks 1.5x faster with Vera than on traditional CPUs and plans to deploy Vera in an upcoming production system.
DeepInfra reported up to 2.2x the speed of the best x86 processor on agentic orchestration for its production agent workload, making Vera the fastest of the four architectures it tested. ClickHouse scaled its standard binary, without Vera-specific tuning, near-linearly across 176 cores and topped the ClickBench hardware leaderboard.
DuckLabs achieved about 1.5x the median TPC-H SF1000 performance of an Intel Xeon 6 baseline. Daytona brought 1,000 agent sandboxes online in 10.8 seconds and achieved 12.7 completed jobs per second at 2,000 concurrent sandboxes, compared with 88.4 seconds using x86. Modal reported up to 2x sandbox throughput per core versus comparable x86 hardware, with gains that increased under heavy load.
For infrastructure providers and users running agents, the results point to a tradeoff NVIDIA says Vera is designed to address: faster task execution alongside greater capacity within power limits. The reported partner tests also suggest that the CPU is being evaluated across varied workloads rather than a single benchmark.
