The Pulse
Vera CPU Runs 1.6x More Agents per Watt Than the Fastest x86 CPU, NVIDIA Says
NVIDIA’s Vera CPU supports 1.6 times more agents per watt than the fastest x86 CPU, the company says, in a comparison based on a 100-megawatt data center. Results from six early adopters, including tests of agentic coding, orchestration, Cl
AI.info Team · Written with AI · fact-checked ·
NVIDIA’s Vera CPU can run 1.6 times more agents per watt than the fastest x86 CPU, the company says. In a comparison based on a 100-megawatt data center, Vera also delivers twice the agent speed of the most energy-efficient x86 CPU while maintaining leading energy efficiency, according to NVIDIA. The company’s November 5, 2026, blog post includes results from six early adopters: Perplexity, DeepInfra, ClickHouse, DuckLabs, Daytona and Modal.
CPU work between GPU reasoning steps
Agents, in NVIDIA’s description, are systems that do more than generate answers. They search for information, query databases, write and test code, and take action. Each of those steps relies on a CPU, with agents moving between GPU reasoning and CPU execution throughout a task.
As businesses put more agents to work, NVIDIA says AI factories need CPUs that can execute those steps quickly and support many agents at once, within available power. Providers need to support more agents without making each one wait longer to get work done, NVIDIA says. NVIDIA frames power use as a constraint on data-center operations, with the goal of maximizing the work done for every watt.
Vera’s speed and capacity claims
Vera combines fast agent performance and high capacity per watt in a single CPU, NVIDIA says. In its comparison, Vera enables 1.6 times more agents per watt than the fastest x86 CPU. Compared with the most energy-efficient x86 CPU, NVIDIA says Vera delivers twice the agent speed with leading energy efficiency.
The company presents one CPU for a range of agent workloads as an alternative to mixing fast and dense x86 parts to balance speed and capacity. Its comparison is framed around a 100-megawatt data center, where NVIDIA says a single Vera CPU performs better than both the fastest x86 CPU and the most energy-efficient x86 CPU.
Early-adopter results
Perplexity ran agentic coding tasks 1.5 times faster with Vera than on traditional CPUs, according to NVIDIA. Perplexity plans to deploy Vera in an upcoming production system. For agentic orchestration on its production agent workload, DeepInfra delivered up to 2.2 times the speed of the best x86 processor. That result was the fastest among the four architectures tested, NVIDIA says.
ClickHouse’s standard binary, without Vera-specific tuning, scaled near-linearly across 176 cores and topped the ClickBench hardware leaderboard. DuckLabs achieved about 1.5 times the median TPC-H SF1000 performance of an Intel Xeon 6 baseline.
Daytona brought 1,000 agent sandboxes online in 10.8 seconds with Vera, compared with 88.4 seconds with x86. At 2,000 concurrent sandboxes, it achieved 12.7 completed jobs per second using Vera. Modal achieved up to twice the sandbox throughput per core compared with comparable x86 hardware, with gains that NVIDIA says grow under heavy load.
The examples cover coding, orchestration, database benchmarks and sandbox throughput. NVIDIA presents them as early-adopter results alongside its broader comparison of agent speed and capacity per watt; the reported measures differ across the workloads, from task speed and benchmark performance to the time needed to start sandboxes.