Summary
NVIDIA says its Vera CPU delivered 1.6 times more agents per watt than the fastest x86 CPU in its comparison, and twice the agent speed of the most energy-efficient x86 CPU. The company also reports early partner results across coding, orchestration and sandbox workloads.
NVIDIA says its Vera CPU can support both fast AI-agent execution and high agent capacity per watt, citing a company comparison framed around a 100-megawatt data centre. The company also points to early partner tests involving coding, agent orchestration and software sandboxes.
AI agents carry out tasks beyond generating text: they can retrieve information, query databases, write and test code, and take actions through software tools. NVIDIA says those steps rely on CPUs, with workloads moving between GPU-based model reasoning and CPU execution. That makes CPU speed relevant to how quickly an agent completes a task, while capacity per watt matters when an operator is trying to run many agents within a power limit.
NVIDIA’s performance claims
In its 100-megawatt data-centre comparison, NVIDIA says a single Vera CPU performed better than both the fastest x86 CPU and the most energy-efficient x86 CPU it compared against. Against the fastest x86 CPU, NVIDIA reports 1.6 times more agents per watt. Against the most energy-efficient x86 CPU, it reports twice the agent speed while maintaining leading energy efficiency.
These figures use different x86 baselines: one comparison emphasises agent capacity per watt, while the other emphasises speed. They should not be read as a single combined performance ratio. NVIDIA’s announcement gives the headline results but does not name the x86 processors or detail the configuration for the 100-megawatt comparison, so the ratios are best understood as company-reported results for its comparison.
Early partner workload results
NVIDIA says Perplexity ran agentic coding tasks 1.5 times faster with Vera than on traditional CPUs and plans to deploy Vera in an upcoming production system. DeepInfra reported up to 2.2 times the speed of its best x86 processor on its production agent-orchestration workload, and said Vera was fastest among four architectures tested.
For infrastructure tasks, Daytona brought 1,000 agent sandboxes online in 10.8 seconds with Vera, compared with 88.4 seconds using x86. At 2,000 concurrent sandboxes, it reported 12.7 completed jobs per second with Vera. ClickHouse reported that its standard binary, without Vera-specific tuning, scaled near-linearly across 176 cores and topped the ClickBench hardware leaderboard.
The partner results cover different tasks and measurements, rather than one common benchmark. Together with NVIDIA’s comparison, they show the company’s case for Vera as a CPU aimed at both agent response speed and the ability to run many concurrent workloads efficiently.