The Resurgence of CPUs in the AI Ecosystem
The rapid advancement of personal artificial intelligence (AI) agents is giving advanced computing rivals AMD and Intel a significant boost, allowing them to outperform larger technology companies this year and over the past month. These companies, which are primary developers of central processing units (CPUs), are finding a crucial and growing role in the AI market, a sector that had previously been dominated by Nvidia’s graphics processing units (GPUs).
The initial focus on generative AI, kicked off by the launch of tools like ChatGPT, centered heavily on GPUs. However, as AI agents become more sophisticated and are utilized for diverse tasks, the workload is beginning to migrate from GPUs to CPUs, according to Ryan Shrout, president of Signal65. Shrout stated, “As more agents are created and developed, and more people start to use them for more tasks, it’s going to start to shift the workload away from GPUs and onto CPUs.”
Technology Behind the AI Agents
The current proliferation of personal AI agents has revitalized the market standing of Advanced Micro Devices and Intel. Historically, the CPU was the main component in a server, but the recent AI boom shifted attention to GPUs. Now, CPUs are regaining prominence because they are specialized in powering autonomous agents that can operate continuously for extended periods.
The trend was highlighted by industry debuts, such as OpenAI’s AI agent, Dots, and Meta‘s Muse, which launched in early September. Muse gained rapid traction, topping the Apple App Store in under two weeks. Already, AMD and Intel were strong performers in the stock market this year due to rising CPU demand, but their momentum accelerated in the last month. As of Monday’s close, AMD and Intel saw jumps of 32% and 21%, respectively, surpassing all major tech mega-cap stocks.
This positive trend continued through Tuesday, with AMD rising nearly 3% to reach a high of $649.42. AMD’s increasing strength across both its CPU and GPU offerings has positioned the company within the trillion-dollar market capitalization category.
The technical shift is evident in how the agents operate. Meta’s Muse reportedly runs on an AMD-powered computer, while OpenAI’s Dots uses a virtual computer powered by AMD’s EPYC-branded CPU. A Meta spokesperson confirmed that the company’s design is intentionally flexible: “We take a diverse approach to our hardware and are largely CPU-agnostic by design, which gives us the most flexibility in acquiring capacity.”
Competition and Future Growth Projections
While AMD is a key player, the market remains competitive. Most servers operated by major cloud providers—including Amazon, Google, Meta, and Microsoft—utilize processors from either Intel or AMD. However, nearly all major cloud service providers are also developing proprietary chips, typically based on Arm technology. Arm itself debuted a dedicated CPU for agents in March, with Meta serving as its first customer, and its stock price has since more than doubled.
Nvidia recently introduced Vera, a completely redesigned central processor that comes integrated with a rack of dedicated CPUs. The chip giant anticipates that the CPU market will reach $200 billion by 2030. However, industry analysts suggest that Nvidia is not the undisputed leader in this specific segment. Jordan Klein, an analyst at Mizuho Securities, noted that, “It’s just a new product and they’re going to be competing against a lot of different players,” adding that he believes AMD currently possesses the “best product.”
AI experts clarified the roles of the components: Daniel Newman, CEO of Futurum Group, explained that while GPUs are still necessary for inference (the processing of an AI model), the agents require a CPU to perform the actual workflows while the GPUs handle the thinking. He specified, “GPUs are still needed for inference, or processing an AI model, but an agent also needs a CPU,” and added, “CPUs are actually performing the workflows while GPUs are doing the thinking.”
Financial Performance and Market Potential
AMD’s financial results show a clear link to this trend. The company’s data center segment revenue more than doubled in the quarter ending in June, reaching $6.7 billion, which constituted nearly 60% of its total sales. Dan McNamara, AMD’s senior vice president and general manager of compute and enterprise AI, stated that the conversation around agentic usage has changed significantly over the last eight to 10 months, predicting that CPU sales are going to “really ramp.”
The market potential for CPUs is accelerating. Futurum estimates that total CPU sales will hit $118 billion in 2027, which is almost double the firm’s earlier forecast from May. AMD itself predicted that the entire CPU market would reach $220 billion in sales by 2030, significantly higher than its previous forecast of $60 billion for 2025.
Morgan Stanley estimates that Meta’s Muse agent could account for 20% of AMD’s chip sales in 2026. The firm also estimates Muse’s serving costs, ranging between $3 and $130 per month, with an average cost of $37 per user, depending on the inference volume.
AMD maintains a strong position, reportedly commanding about 46% of the market for x86 CPU units, according to Mercury Research data from August. Furthermore, AMD benefits from its use of the x86 architecture, which allows the agents to operate using older software without compatibility concerns. Analysts noted that AMD holds an advantage due to its large market share and strong presence within the cloud hyperscaler infrastructure.
