Nvidia Takes on AMD and Intel With Vera AI Server CPU

Nvidia is expanding its AI hardware strategy beyond GPUs with Vera, a data center CPU designed for agentic AI and reinforcement learning. The step puts the chipmaker in direct competition with AMD and Intel. It also allows Nvidia to supply more of the technology used inside AI servers.

Nvidia representatives stated that Vera chips were delivered to Anthropic, OpenAI, and SpaceX in June 2026. OpenAI plans to deploy the processors in large quantities starting this quarter. However, Oracle is currently the only major cloud provider named among Vera's listed partners.

Vera features 88 Nvidia-designed Olympus cores and supports Armv9.2. Its spatial multithreading technology enables 176 total threads. The CPU supports up to 1.5 TB of LPDDR5X memory and delivers up to 1.2 TB/s of memory bandwidth. It also features a 3.4 TB/s scalable coherency fabric and 1.8 TB/s of NVLink-C2C bandwidth. Power consumption ranges from 250 watts to 450 watts.

Nvidia says Vera delivers 50% better performance for AI agent workloads than x86 processors. Its architecture prioritizes single-core speed, memory bandwidth, and low latency.

Meanwhile, vice president of hyperscale and high-performance computing at NVIDIA, Ian Buck, pointed out that agents have made CPUs "much more integral," particularly in how quickly they can answer individual requests.

Nvidia Vera Enters a Server CPU Market Where AMD is Gaining Ground

AMD reached 33.2% of x86 server CPU unit share in the first quarter (Q1) 2026, according to data cited by Tom's Hardware. Its revenue share climbed to 46.2%, while Intel retained 66.8% of unit shipments.

Nvidia estimates the server CPU opportunity could eventually reach $200 billion USD. Bernstein valued the mature market at about $37B in 2025. Wolfe Research expects Vera to average roughly $5K per chip and reach 1.3M shipments this year (2026).

Nvidia has not disclosed pricing yet.

Vera will be available independently and in single and dual-socket configurations. Nvidia will also offer a liquid-cooled rack with up to 256 CPUs, capable of running more than 22.5K concurrent environments. Vera will also serve as the host CPU in Vera Rubin systems. The NVL72 configuration combines 36 Vera CPUs with 72 Rubin GPUs.

The launch tests whether Nvidia can extend its AI accelerator leadership into CPUs. AMD is gaining server share, while Intel continues to lead in unit shipments.