NVIDIA founder and CEO Jensen Huang recently stated on the company's earnings call that with the launch of Vera, a new CPU product designed for "agent AI", NVIDIA is opening a "new, previously uncovered US$200 billion serviceable market (TAM)." He said that global computing is being reshaped around agents and embodied robots, and Nvidia is at the center of this transformation.

Huang Renxun has always been known for "building momentum for his own company", and he is often compared to the equally optimistic and high-profile Salesforce CEO Marc Benioff. However, as Nvidia delivers financial results that exceed expectations time and time again, this optimism is becoming more and more credible in the eyes of the capital market. In its latest financial quarter, Nvidia recorded revenue of US$81.6 billion and forecast that revenue in the next quarter will further rise to US$91 billion, setting another record.

The Vera CPU, for which he has high hopes, was released as early as March this year. It can be sold alone or combined with Nvidia's Rubin GPU to form a complete platform. Huang described Vera during the conference call as "the world's first CPU designed specifically for intelligent artificial intelligence" and claimed that it has already demonstrated strong sales momentum.

For a long time, Nvidia has dominated the GPU field, while the traditional CPU market is mainly dominated by companies such as Intel and AMD. Although Nvidia has made CPUs in the past, this has never been its core business. At the same time, cloud computing giants have stepped up efforts to develop self-developed AI chips, exacerbating concerns about Nvidia's long-term position. For example, last month Amazon Cloud Services (AWS) announced a high-profile contract with Meta covering millions of self-developed AI CPUs. Amazon CEO Andy Jassy has publicly stated many times that AWS has the ability to be "at least as good as, and possibly even better than, Nvidia" in terms of GPUs and CPUs.

In this context, Vera is regarded by Jen-Hsun Huang as a key weight for Nvidia to open up a new growth curve. He said on the call: "Vera has opened up a new $200 billion TAM for NVIDIA, which is a market we have never entered before. Every major hyperscale cloud service provider and system manufacturer is working with us to deploy it." He further pointed out that the global computing architecture is being rebuilt for "intelligent AI" and "embodied robot AI", and NVIDIA is at the center of this reconstruction.

The logic given by Huang Renxun is: the "thinking" part of today's large models mainly relies on the GPU, and when agents perform various tasks, most of them run on the CPU. These agents will not only call tools to complete specific tasks, but may also have a "dedicated computing environment" similar to a personal computer in the future. Unlike CPUs in traditional cloud architectures that emphasize multi-core parallelism and aim to host as many application instances as possible, Vera is designed to process "tokens" as quickly as possible to adapt to the workload characteristics of the agent when performing tasks.

Facing external doubts: With cloud manufacturers and startups developing their own AI chips, why does NVIDIA think it will become the preferred supplier of "intelligent CPUs"? One of the answers given by Jen-Hsun Huang is order data. He revealed on the call that Nvidia has sold $20 billion worth of independent Vera CPUs this year, and 2026 has just passed a short period of time.

In the picture he painted, the future world will not only have about one billion human users, but also "intelligent agents" of the same order of magnitude or even larger. “My sense is that the world will have billions of intelligent agents, and of course it won’t happen overnight, but it will grow incrementally,” he said. "These agents will all use tools, and these tools are like PCs to them, corresponding to the way we use PCs today." In his view, the direct hardware consequences of this evolution are very clear - "We will need more CPUs."

Although Nvidia's dominance in the GPU field has not been shaken for the time being, Wall Street remains highly sensitive to which force will pull Nvidia off the "altar". Recently, this uneasiness has become more and more concentrated in the CPU battlefield: on the one hand, the self-developed processors of Intel, AMD and cloud manufacturers are gradually maturing, and on the other hand, Nvidia is trying to use Vera as an entry point to replicate its advantages in the GPU field on the CPU side. In Huang Renxun's latest narrative, the intelligent CPU is packaged as a "new $200 billion story" that can support Nvidia's growth for many years to come. What the market will have to see next is whether this narrative can continue to be transformed into real orders and profits.

Related articles:

Nvidia delivers first batch of Vera CPUs to OpenAI, Anthropic, SpaceX and Oracle