NVIDIA (NASDAQ: NVDA) announced the launch of its Vera CPU, which the company describes as the first processor designed specifically for agentic AI and reinforcement learning applications. The chip delivers twice the efficiency and operates 50% faster than traditional rack-scale CPUs, according to the company's statement.

The Vera CPU builds on NVIDIA's Grace CPU architecture and features 88 custom-designed Olympus cores. Each core can run two tasks using NVIDIA Spatial Multithreading technology. The processor includes a low-power memory subsystem built on LPDDR5X memory, providing up to 1.2 TB/s of bandwidth.

Several major technology companies plan to deploy the Vera CPU, including Alibaba, ByteDance, Meta, and Oracle Cloud Infrastructure. Cloud service providers CoreWeave, Lambda, Nebius, and Nscale are also collaborating with NVIDIA on deployment.

Manufacturing partners adopting the Vera CPU include Dell Technologies, HPE, Lenovo, and Supermicro, along with ASUS, Compal, Foxconn, GIGABYTE, Pegatron, Quanta Cloud Technology, Wistron, and Wiwynn.

NVIDIA introduced a new Vera CPU rack configuration that integrates 256 liquid-cooled processors to support more than 22,500 concurrent CPU environments. The rack uses the company's MGX modular reference architecture.

As part of the NVIDIA Vera Rubin NVL72 platform, Vera CPUs connect with NVIDIA GPUs through NVLink-C2C interconnect technology, providing 1.8 TB/s of coherent bandwidth.

"The CPU is no longer simply supporting the model; it's driving it," said Jensen Huang, NVIDIA's founder and CEO.

The Vera CPU is in full production and will be available from partners in the second half of 2026.