The personal AI agent boom brings AMD and intel New life after years of corporate surveillance Nvidia And its GPUs dominate the artificial intelligence market.
Central processing units (CPUs) were the primary workhorse of computer servers before Nvidia’s graphics processing units took center stage with the launch of ChatGPT and the beginning of the generation AI era. Now, CPUs are back and boosting the fortunes of market leaders AMD and Intel. Their chips specialize in powering long-running agents that can work alone for hours.
Last week, OpenAI released the AI agent Dots. Metadebuted Muse in early September. Muse quickly gained popularity and reached the top of the Apple App Store in less than two weeks. Long before that, AMD and Intel were already big winners in the stock market this year, largely due to a surge in CPU demand, but they gained momentum in the last month, rising 32% and 21%, respectively, outpacing all big-cap stocks in tech companies.
AMD is gaining strength in both CPUs and GPUs, pushing the company into the $1 trillion market cap club.
“As more agents are created and developed, and more people start using them for more tasks, workloads will start to shift from GPUs to CPUs,” said Ryan Shrout, president of Signal65, an AI hardware and cost consulting firm.
Users learned about the power behind the new agent by asking questions. Although Muse tells users it runs on AMD-powered computers, Dots said it runs on virtual computers with AMD’s EPYC brand CPUs.
A Meta spokesperson told CNBC that the company designed the system to be able to use any type of CPU available.
“We take a diverse approach to hardware, with a primarily CPU-agnostic design that gives us the most flexibility in acquiring capacity,” a Meta spokesperson said in a statement.
OpenAI uses multiple CPU providers as well. This means AMD isn’t the only one in the game.
Most servers in hyperscalers — Amazon, googlemeta and microsoft — Uses an Intel or AMD CPU. However, nearly all major cloud providers also develop their own custom chips, typically: arm technology. Arm announced its own CPUs to distributors in March with Meta as its debut customer, and the stock has more than doubled since then.
Earlier this year, Nvidia released Vera, a completely redesigned central processor, with only CPUs in entire racks. The semiconductor giant said it developed its new Vera CPU specifically for agents, and expects CPUs to become a $200 billion market by 2030.
But analysts say this is one part of the chip industry where Nvidia is not currently the favorite.
“This is just a new product and it’s going to be competing with a lot of different companies,” Mizuho Securities analyst Jordan Klein said in an interview, adding that he believes AMD currently has the “best product.”
CPUs are driving a healthy portion of AMD’s growth. Revenue from the company’s data center business more than doubled to $6.7 billion in the quarter that ended in June, accounting for nearly 60% of total sales.
“The conversation has changed quite a bit in the last eight to 10 months when it comes to using agents,” said Dan McNamara, AMD’s senior vice president and general manager of compute and enterprise AI. McNamara added that CPU sales will “really increase.”
GPU thinks, CPU executes
Coding software like Muse, Dots, and even OpenAI’s Codex use AI models to plan tasks that can be performed autonomously. Futurum Group CEO Daniel Newman explained that agents require dedicated computers that can run in the background for hours or even days.
GPUs are still required for inference and AI model processing, but agents also require CPUs.
“The CPU is actually running the workflow, and the GPU is doing the thinking,” Newman says.
Meta and OpenAI use virtual machines and divide the server into multiple lightweight virtual computers, so a single server can support a large number of users.
Some AMD EPYC processors have up to 192 CPU cores. According to benchmarks and the agent itself, Muse’s virtual machine uses 2 cores and Dots uses 9 cores.
CPUs are cheaper than GPUs.
Benchmarks and queries show that Dots uses an AMD EPYC 9V74 CPU. This CPU is available from resellers for less than $3,000. The EPYC 9D25 that Meta for Muse reportedly uses is even cheaper on the secondary market. Nvidia GPUs can cost more than 10 times the price of a single chip and are typically sold in clusters of hundreds or thousands of chips.
Investors are curious to see what Meta plans to offer Muse users after it has received more than 5 million downloads since its launch last month, according to Sensor Tower. morgan stanley estimates that Muse’s service costs range from $3 to $130 per month, with an average of $37 per user, depending on inference usage.
The company says Meta distributors could account for 20% of AMD’s 2026 chip sales.
Advanced Micro Devices (AMD) CEO Lisa Su speaks at a forum held at the Taipei Marriott Hotel in Taipei on May 22, 2026.
Chen Yihua | AFP | Getty Images
Due to the large profits available in CPUs, growth forecasts for this segment are accelerating.
Futurum predicts total CPU sales in 2027 to be $118 billion, nearly double the company’s previous forecast in May.
In July, AMD predicted that the total CPU market revenue would reach $220 billion in 2030, exceeding its previous forecast for 2025 of $60 billion. AMD said in July that it expects to capture more than half of the market.
Mercury Research announced in August that AMD currently holds about 46% of the x86 CPU unit market. Demand for Intel is also rapidly increasing.
But AMD has a foothold with hyperscalers, and that’s where many of the virtual machines for agents will be deployed. It also has the benefit of using x86 architecture for its chips, which allows agents to use older software without worrying about compatibility issues.
“AMD is probably the biggest beneficiary of this because they have the largest market share in CPUs and the largest share in the cloud hyperscaler world,” Klein said, adding, “We’re not going to exclude Nvidia.”
“Any company that can supply both types of chips is in the best position,” Klein said.
WATCH: Helios is AMD’s first AI system to rival Nvidia Vera Rubin — we got an exclusive first look
