What Does an AI Agent Need a CPU For? Why Nvidia Wants In on the Business Dominated by AMD and Intel

Since the start of the artificial intelligence (AI) revolution, the story has centered around the graphics processing unit (GPU). Nvidia (NASDAQ: NVDA) built the world's fastest, and the tech giant's sales skyrocketed. But as AI transitions from answering questions to completing tasks on its own, another chip is stepping into the spotlight: the central processing unit (CPU).

Advanced Micro Devices (NASDAQ: AMD) and Intel (NASDAQ: INTC) have long led this market -- but now, Nvidia wants in. Let's consider why AI agents need CPUs and whether Nvidia can dominate in this market too.

Image source: Getty Images.

What an AI agent does with a CPU

There's been a lot of talk about agentic AI in recent months -- but what exactly is this technology? It's software that works through various steps to reach a particular goal. Nvidia chief Jensen Huang, speaking during an earnings call earlier this year, described an agent as a "harness" built around an AI model, and the harness manages inputs and outputs, memory, tool use, and more. All of that runs on CPUs, while GPUs power the model's "thinking."

Huang predicts the world will eventually have billions of AI agents, and all of this should result in more use for GPUs and CPUs.

AMD chief executive officer Lisa Su also expects the CPU market to take off. She says CPUs used to be paired with GPUs at a ratio of one to four or one to eight, but that ratio is moving toward one to one, according to TrendForce. AMD now expects server CPU market growth of more than 35% annually, topping $120 billion by 2030, the market intelligence firm added.

Nvidia seizes a fresh opportunity

Nvidia, recognizing the opportunity, is now rolling out the Vera Rubin platform, featuring the company's first stand-alone CPU. Designed specifically for agentic AI, it aims for speed, delivering up to 1.5x faster performance per core compared to CPUs based on x86 architecture, Nvidia says.

The tech giant is on track to generate $20 billion in stand-alone CPU revenue this year and says the market opportunity equals $200 billion.

Here's how that compares to the company's current data center business. Nvidia's data center revenue reached $89 billion in the second quarter. Using that as a reference, $20 billion is only about three weeks of data center sales. But for a business Nvidia didn't previously address, that's a pretty impressive amount. And we're likely in the early days of this growth story. It's also important to note that this $20 billion is over just a few months, since rollout began recently.

The hurdles ahead

Though Nvidia is the GPU giant, this doesn't mean the company will unseat Intel and AMD in CPUs. And this is for two major reasons. First, AMD and Intel have decades of customer relationships and confirmed leadership in CPUs. Second, customers running software designed for x86 chips may not quickly switch over to Nvidia, which is based on a different architecture.

That said, Nvidia doesn't necessarily have to unseat the two market leaders to win. If the AI agent market is the next big growth area in AI, there should be plenty of room for all three companies to be successful.

What does this mean for Nvidia and its shareholders? The emergence of CPUs in the AI story offers the company a fresh opportunity for growth -- whether it becomes a leader or not. Nvidia has broadened its reach across the AI space in recent years, through partnerships and the development of new technologies, and that strengthens its position as an AI leader.

Investors should look to see how much CPU market share Nvidia may capture and how that translates into revenue in the coming quarters. Considering Nvidia's commitment to innovation over time, there's reason to be optimistic that CPUs may become a significant growth driver for the company in the years to come.

Adria Cimino has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Advanced Micro Devices, Intel, and Nvidia. The Motley Fool has a disclosure policy.