Agentic Era Begins: NVIDIA Delivers First 'Vera' CPUs to Top AI Labs
NVIDIA has delivered its first Vera CPUs to OpenAI, Anthropic, and SpaceXAI, marking a shift toward 'Agentic AI.' These chips are designed to power autonomous agents that can reason and execute tasks in physical environments with 10x lower cost per token.
The landscape of Physical AI shifted this week as NVIDIA’s first Vera CPUs—the company's first processors specifically architected for AI agents—arrived at the doorsteps of OpenAI, Anthropic, and SpaceXAI. Moving beyond traditional LLMs, the "Vera" architecture is designed to support Agentic AI: systems that don't just generate text, but act as autonomous entities capable of reasoning, planning, and executing complex workflows in real-world environments.
NVIDIA CEO Jensen Huang characterized the current market demand as "parabolic," noting that the transition from simple chatbots to autonomous agents requires a fundamental redesign of the compute stack. The Vera NVL72 system claims to run agent sandboxes up to 50% faster than traditional CPUs and provides inference at one-tenth the cost per token. For Physical AI applications, this efficiency is critical; it allows robots and autonomous systems to process environmental data and make high-level logic decisions without the latency or energy overhead of general-purpose silicon.
By placing these chips in the hands of major AI labs and SpaceX’s AI wing, NVIDIA is signaling that the next frontier of automation isn't just about "smarter" models, but about the compute density required to let those models inhabit physical forms. As these agents begin to integrate with robotics hardware, we are likely to see a leap in the "common sense" reasoning capabilities of autonomous machines.
Source: NVIDIA Blog
