Agentic AI

Agentic AI is artificial intelligence that completes multi-step tasks on its own by planning, calling tools, retrieving information, and acting in a loop, rather than answering a single prompt. Because one prompt can trigger a long chain of reasoning and tool calls, agentic AI shifts compute demand toward CPUs and inference.

What is agentic AI?

An AI agent does not just answer; it acts. Given a goal, it breaks the goal into steps, decides which tools or APIs to call, fetches data, runs code, evaluates the result, and repeats until the task is finished. A single user request can therefore expand into many model calls and a great deal of ordinary software work, such as Python runtimes, sandboxed code execution, orchestration logic, and analytics, all of which run between the model calls. That makes agentic AI fundamentally heavier and more sequential than a one-shot chatbot response, and it changes what hardware an AI system needs.

How is agentic AI used in thematic investing?

Agentic AI is the demand thesis behind the host CPU and the inference layers of AI. NVIDIA describes an agent workload as a chain of reasoning, retrieval, tool use, and response generation, and built a CPU specifically for it: the 88-core Vera, which it markets as “the CPU for agents,” claiming 1.8 times faster task completion than x86 and, at the rack level, 10 times the agent throughput of the prior Grace Blackwell generation (NVIDIA, May 31, 2026). The shift toward agents is why the AI CPU and AI inference infrastructure concepts matter: more sequential steps and more CPU work per task raise the value of fast host processors and of cheap, abundant inference capacity. NVIDIA founder and CEO Jensen Huang described the workload directly:

“Agentic AI is a new kind of workload. One prompt can launch a thousand-step journey of reasoning, retrieval, tool use and response generation. Vera Rubin was built for this moment.”

— Jensen Huang, founder and CEO, NVIDIA (NVIDIA, May 31, 2026)

For investors, the rise of agents reframes AI demand from a single forward pass to a long, compute-hungry loop, which lengthens the runway for accelerators, host CPUs, memory, and inference capacity.

Related terms & concepts

Frequently asked questions

How does agentic AI differ from a chatbot?

A chatbot answers a single prompt in one pass. An agent plans a task, calls tools and APIs, retrieves data, checks its work, and loops until the task is done, which can mean dozens of model calls and a lot of non-model code in between.

Why does agentic AI matter for chip demand?

Because agent tasks run many sequential steps and a lot of CPU work, such as Python runtimes, sandboxed code execution, and orchestration, alongside the GPU. NVIDIA built its 88-core Vera CPU specifically for agents, claiming 1.8 times faster task completion than x86.

Sources & references

  1. NVIDIA Unveils Vera, the CPU for Agents · NVIDIA Corporation (Newsroom), 2026-05-31
  2. NVIDIA Vera Rubin Ramps Into Full Production to Power Agentic AI Factories Worldwide · NVIDIA Corporation (Newsroom), 2026-05-31