NVIDIA says its Vera CPU delivers 1.6x more agents per watt than the fastest x86 CPU. The disclosure in the NVIDIA Blog frames Vera as a single socket for both speed and density, and early NVIDIA Vera benchmarks from partners now test that pitch.
NVIDIA compared a 100 megawatt data center with Vera against the fastest x86 CPU and the most energy efficient x86 CPU. Vera delivered industry leading agent speed while enabling 1.6x more agents per watt versus the fastest x86 part, NVIDIA said.
Against the most efficient x86 CPU, Vera delivered 2x faster agent speed with leading energy efficiency. The company positions that as one CPU for mixed agent workloads instead of mixing fast and dense x86 parts.
The partner tests span databases, orchestration and sandboxes. Perplexity ran agentic coding tasks 1.5x faster on Vera and plans deployment in an upcoming production system.
DeepInfra reported up to 2.2x speed over the best x86 processor on agentic orchestration. ClickHouse said its standard binary scaled near linearly across 176 cores and topped the ClickBench hardware leaderboard with no Vera specific tuning.
DuckLabs reported 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 on Vera versus 88.4 seconds on x86.
Daytona also achieved 12.7 completed jobs per second at 2,000 concurrent sandboxes. Modal reported up to 2x sandbox throughput per core versus comparable x86 hardware, with gains that grow under heavy load.
Independent testing tempers the white paper margin. In SPEC CPU 2026 integer throughput testing Vera finished 1.5% ahead of the EPYC 9755 system, compared with the 3% advantage in NVIDIA's white paper, per Cryptonomist.
Tom's Hardware also benchmarked standalone Vera against Granite Rapids-AP and EPYC Turin for data center float and intrate performance. StartupHub.ai data shows inference hardware vendors including Groq and Ampere Computing remain among the most tracked AI hardware companies as of November 2026, with Vera targeting the same power limited data center buyer.
That buyer matters. AI agents shuttle between GPU reasoning and CPU execution for search, database queries and code tests.
Throughput per watt decides how many agents a rack can host without adding wait time. A CPU that lifts both speed and density squeezes incumbents Intel and AMD and Arm server vendors that sell on efficiency or peak single thread.
For startups selling sandboxes and orchestration, Vera raises the bar for per core throughput claims. Modal and Daytona results point to headroom under heavy concurrency, not just peak runs.
The next public check is NVIDIA CEO Jensen Huang's GTC Berlin keynote at 11 a.m. CEST Oct 21.
