Key takeaways: The personal intelligent agent boom is driving strength in AMD and Intel shares.

This year and over the past month, both companies have outperformed other large technology giants.

As leading developers of central processing units (CPUs), these two long-time rivals have found an important new position in an AI market dominated by Nvidia.

Signal65 President Ryan Shrout said: "As more intelligent agents are developed and more people use them to handle various tasks, the computing load will gradually shift from GPUs to CPUs."

For years, Nvidia and its GPUs have dominated the AI market; now the rise of personal AI intelligent agents is injecting new vitality into AMD and Intel.

The generative AI era began with ChatGPT, placing Nvidia GPUs at center stage, while CPUs were once the main computing units in servers.

Now CPUs are returning, driving upward performance for leaders AMD and Intel.

These chips excel at powering intelligent agents that run autonomously for long periods, working independently for hours on end.

Last week, OpenAI followed Meta in launching AI agent Dots.

Meta released Muse in early September, and the product quickly went viral, topping Apple's App Store download chart.

Before that, driven by surging CPU demand, AMD and Intel shares had already risen sharply this year; over the past month, gains accelerated further, up 32% and 21% respectively, surpassing all large technology stocks.

AMD is growing simultaneously in both CPUs and GPUs, with its market capitalization officially entering the trillion-dollar club.

Ryan Shrout, president of AI hardware and cost consulting firm Signal65, said: "As more intelligent agents are developed and more people use them to handle various tasks, the computing load will gradually shift from GPUs to CPUs."

Users can directly ask intelligent agents and learn about the underlying computing power: Muse tells users it runs on servers equipped with AMD chips; Dots says it relies on virtual computers powered by AMD EPYC CPUs.

A Meta spokesperson said the company's system design supports various CPUs.

"Our hardware strategy is diverse, and at the design level we are basically not tied to a specific CPU, giving us maximum flexibility in computing scale-up," Meta said in a statement.

OpenAI also uses multiple CPU suppliers, indicating AMD is not the only choice.

Hyperscale cloud providers such as Amazon, Google, Meta, and Microsoft mostly use Intel or AMD CPUs in their servers.

But almost all major cloud service providers are developing their own chips, mostly based on the ARM architecture.

ARM released its own CPU for intelligent agents in March this year, with Meta as the first customer, and ARM shares have more than doubled since then.

Earlier this year, Nvidia launched a newly restructured central processor called Vera, along with a supporting software ecosystem.

The chip giant said it built the Vera CPU because it is bullish on the CPU sector, expecting the CPU market to reach $200 billion by 2030.

But analysts believe that in the CPU race, Nvidia is not currently the market's first choice.

Mizuho Securities analyst Jordan Klein said in an interview: "This product has just been released and needs to compete with many rivals." He believes AMD currently has the "strongest product."

CPUs are an important growth engine for AMD.

In its fiscal 2026 second-quarter results through June, AMD's data center business revenue more than doubled year over year to $6.7 billion, accounting for nearly 60% of total sales.

Dan McNamara, senior vice president and general manager of AMD's Compute and Enterprise AI business, said: "Over the past 8 to 10 months, the industry's view of intelligent agent applications has undergone a huge shift." He added that CPU sales are about to "ramp up significantly."

GPUs handle thinking, CPUs handle execution

Muse, Dots, and OpenAI's coding tool Codex rely on AI models to plan tasks that can be executed autonomously.

Daniel Newman, CEO of Futurum Group, explained that these intelligent agents need independent computing environments and can run continuously in the background for hours or even days.

Inference (AI model computation) still depends on GPUs, but intelligent agents also rely on CPUs.

Newman said: "GPUs handle the model's thinking and computation, while CPUs execute the entire workflow."

Meta and OpenAI use virtual machine technology to divide a single server into multiple lightweight virtual computing instances, allowing one server to serve many users simultaneously.

Some AMD EPYC processors have up to 192 CPU cores.

Benchmark tests and information disclosed by the intelligent agents themselves show that Muse virtual machines use 2 cores, while Dots uses 9 cores.

CPUs are cheaper than GPUs.

Benchmark tests and query information show that Dots uses an AMD EPYC 9V74 CPU, with a dealer price of less than $3,000.

The EPYC 9D25 used by Meta for Muse is even cheaper on the second-hand market.

A single Nvidia GPU can cost more than ten times that, and GPUs are usually deployed in clusters of hundreds or thousands.

Sensor Tower data shows that Muse has exceeded 5 million downloads since launching last month, and investors are closely watching how Meta will support Muse users.

Morgan Stanley estimates that Muse's monthly service cost per user ranges from $3 to $130, with an average of $37, depending on inference usage.

Morgan Stanley believes this Meta intelligent agent could contribute 20% of AMD's 2026 chip sales.

On May 22, 2026, AMD CEO Lisa Su delivered a speech at a forum at the Taipei Marriott Hotel.

The CPU sector offers huge business opportunities, and industry growth forecasts continue to be raised.

Futurum predicts that total CPU sales will reach $118 billion in 2027, nearly double its previous forecast in May.

AMD predicted in July that the overall CPU market will reach $220 billion by 2030, up from its earlier 2025 forecast of $60 billion; AMD expects to capture more than half of the market share by then.

According to Mercury Research data, AMD currently holds about 46% of x86 CPU shipment market share.

Intel is also seeing a surge in demand.

AMD has already deeply penetrated major hyperscale cloud providers, and a large number of intelligent agent virtual machines will be deployed on these cloud platforms.

At the same time, AMD chips have an obvious advantage with the x86 architecture, as intelligent agents can directly run legacy software without compatibility issues.

Klein said: "AMD is likely the biggest beneficiary of this round of gains, as it has the highest CPU market share among cloud providers. But I would not rule out Nvidia."

"Companies that can supply both types of chips (CPU + GPU) are in the most favorable position."