CES® 2027 NEWS
ROAD TO CES 2027Live now
ROAD TO CES · AN INTELITOYS EVENTJAN 6–9 · VEGAS
Home AI Servers

NVIDIA Says Vera CPU Runs AI Agents Faster While Cutting Data Center Power

NVIDIA says a single Vera CPU beats both the fastest and the most energy-efficient x86 parts on agent workloads, with partners reporting 1.5x to 2.2x gains in production-style tests.. NVIDIA says its Vera CPU is built for the CPU work behind agentic AI — tool use, code execution, sandboxing and orchestration — delivering 1.6x more agents per watt than the fastest x86 CPU and 2x…

InteliToys Editors ? NVIDIA Vera CPU
NVIDIA Vera CPU rack (image: NVIDIA)

NVIDIA is arguing that the CPU, not just the GPU, has become the bottleneck for agentic AI. In a post the company says AI agents search for information, query databases, write and test code and take action, and each of those steps depends on CPU execution between bursts of GPU reasoning. As businesses put more agents to work, NVIDIA says AI factories need processors that can run many agents at once within fixed power budgets.

The company says its answer is NVIDIA Vera, a CPU it describes as delivering both fast agent performance and high capacity per watt in a single part. NVIDIA says that lets providers standardize on one CPU across agent workloads instead of mixing fast and dense x86 processors to balance speed against capacity. In a 100-megawatt data center comparison, NVIDIA says a single Vera CPU performs better than both the fastest x86 CPU and the most energy-efficient x86 CPU.

NVIDIA quantifies the claim two ways. It says Vera enables 1.6x more agents per watt than the fastest x86 CPU, and delivers 2x faster agent speed with leading energy efficiency compared with the most energy-efficient x86 CPU. The company also says Vera delivers up to 2x faster performance on critical agentic workloads, which it frames as a way for AI factories to generate more tokens per dollar.

Ride the Las Vegas Monorail to CES - stops at the Las Vegas Convention Center

The silicon behind those numbers is built around 88 custom NVIDIA Olympus cores, which NVIDIA says are designed for the branch-heavy, latency-sensitive work behind agentic AI and reinforcement learning. NVIDIA says high single-thread performance helps agent runtimes, tool calls, code execution and evaluation loops finish faster, while its Spatial Multithreading technology creates 176 threads with partitioned core resources for predictable throughput at scale. Vera pairs those cores with up to 1.2 terabytes per second of LPDDR5X memory bandwidth, which NVIDIA says uses significantly lower power than traditional server memory, plus low-latency NVIDIA Scalable Coherency Fabric, NVLink-C2C connectivity, confidential computing and full Arm compatibility.

NVIDIA says early adopters are already seeing similar results in production-style tests. According to the company, Perplexity ran agentic coding tasks 1.5x faster with Vera than on traditional CPUs and plans to deploy Vera in an upcoming production system. NVIDIA says DeepInfra delivered up to 2.2x the speed of the best x86 processor on agentic orchestration for its production agent workload, the fastest of four architectures tested. ClickHouse, NVIDIA says, scaled its standard binary near-linearly across 176 cores with no Vera-specific tuning and topped the ClickBench hardware leaderboard.

NVIDIA also points to infrastructure vendors building around the part. The company says Vera serves as the CPU powering standalone Vera servers, NVIDIA Vera Rubin systems and Vera BlueField-4 STX AI storage platforms, and that it extends NVIDIA's AI platform from GPU to CPU. NVIDIA says the chip is in full production, with system manufacturers including Dell Technologies, HPE, Lenovo and Supermicro building standalone Vera CPU systems at scale.

Source: NVIDIA.

InteliToys is editorial. We do not sell NVIDIA Vera CPU. Watch the source tape in the branded lightbox below.