Nvidia Claims $200 Billion Opportunity in ‘Agentic AI’ with New Vera CPU
Table of Contents
A New Frontier for the GPU King
Jensen Huang has built a reputation not just as a technical visionary, but as one of the most effective corporate promoters in the tech industry. While his optimistic forecasts for Nvidia’s growth often border on the hyperbolic, the company’s financial performance has consistently validated his claims. During Wednesday’s earnings call, Huang unveiled the next major pillar of Nvidia’s expansion: a “brand new $200 billion” total addressable market (TAM) centered around the company’s new CPU, Vera.
The announcement comes on the heels of a staggering quarter for the company, which reported $81.6 billion in revenue and projected another $91 billion for the coming period. However, the introduction of Vera is more than just a numbers game; it represents a strategic move by Nvidia to penetrate a segment of the computing market that has historically been the stronghold of Intel and AMD.
The Shift Toward Agentic Computing
For years, the narrative surrounding AI hardware has been dominated by the GPU. Graphics Processing Units are the engines that power the “thinking” or training phases of large language models. But Huang argues that the next phase of AI—Agentic AI—requires a different architecture entirely. Agentic AI refers to systems that don’t just answer questions, but actively execute tasks, use tools, and operate autonomously to achieve goals.
According to Huang, while GPUs handle the heavy cognitive lifting, these autonomous agents rely on CPUs to perform their assigned tasks. He posits that as the world moves toward billions of active AI agents, these agents will require their own dedicated computing resources—essentially acting as “PCs” for non-human users.
“The world is rebuilding computing for agentic AI and robotic physical AI,” Huang stated during the call. “Nvidia sits at the center of these transitions.”
Engineering for Tokens, Not Cores
The technical distinction of the Vera CPU lies in its specialization. Traditional cloud architecture CPUs are designed around “cores” and the ability to handle multiple simultaneous application instances efficiently. Vera, however, is purpose-built to process tokens as quickly as possible, optimizing the specific data flow required by AI agents rather than general-purpose computing.
This specialization is designed to offer a performance edge over the homegrown chips currently being developed by the very customers Nvidia serves. Recently, Amazon Web Services (AWS) highlighted a massive contract with Meta involving its own AI CPUs, with CEO Andy Jassy suggesting that AWS could eventually match or exceed Nvidia’s capabilities in the CPU space.
Early Traction and Market Reality
Despite the looming threat of in-house silicon from hyperscalers, Huang claims the market is already responding to Vera. He revealed that Nvidia has already sold $20 billion worth of standalone Vera CPUs this year, suggesting that the transition to agentic hardware is already underway.
Vera is currently available both as a standalone product and as part of a bundle with the Rubin GPU, creating a full-stack ecosystem that allows Nvidia to control both the “brain” and the “hands” of the AI agent. By positioning Vera as the foundational hardware for a world populated by billions of digital agents, Nvidia is attempting to ensure that it remains the primary landlord of the AI era, regardless of whether the workload is training a model or executing a complex autonomous task.
