Vera Delivers the Best of Both Worlds: Highest Throughput and Fastest Agents

Agents do not live on the GPU alone. They search, hit databases, write and test code, then bounce back to model reasoning. That bounce runs on CPUs, and AI factories are power-capped.

NVIDIA says its Vera CPU beats both the fastest and the most efficient x86 options in a 100-megawatt data-center comparison: 1.6× more agents per watt than the fastest x86 part, and 2× faster agent speed than the most efficient x86 part at leading efficiency. Early adopters are already quoting production-style wins.

If those claims hold outside NVIDIA's slides, the CPU choice inside AI factories becomes a first-class product decision, not a commodity afterthought.

We're watching the infrastructure story shift from "more GPUs" to "keep the agent fed without cooking the rack." Vera is NVIDIA's answer for the CPU half of that loop — one part instead of mixing fast and dense x86 SKUs to balance latency and capacity.

Partner anecdotes are the interesting part. Perplexity reports agentic coding 1.5× faster and plans a production deploy. DeepInfra cites up to 2.2× the best x86 on agentic orchestration. Daytona brought 1,000 agent sandboxes online in 10.8 seconds on Vera versus 88.4 seconds on x86, and hit 12.7 completed jobs per second at 2,000 concurrent sandboxes. Modal claims up to 2× sandbox throughput per core. ClickHouse's untuned binary scaled across 176 cores to the top of ClickBench hardware. DuckLabs saw about 1.5× median TPC-H SF1000 versus an Intel Xeon 6 baseline.

I take vendor benchmarks with a raised eyebrow and still think the direction is right. Agent fleets fail on orchestration latency and sandbox churn as often as on token throughput. A CPU that collapses that wait under a power cap is useful infrastructure. The downside if the comparisons are cherry-picked: buyers standardize on a NVIDIA-shaped rack that is hard to exit, and x86 vendors answer with their own agent-tuned SKUs that muddy the "one CPU" pitch. For now, the signal is that agent economics are rewriting the bill of materials.

NVIDIA has been extending beyond GPUs into networking, software, and now agent-oriented CPU positioning as customers build power-limited AI factories. Independent replication will matter more than launch-week quotes.