With a 1:1 Ratio to GPUs, AMD Earnings Report Reveals CPUs Are Undergoing a Valuation Reassessment

Recently, global semiconductor giant AMD released its latest earnings report. Not only did its financial results comprehensively exceed expectations, but through highly forward-looking market forecasts, the report also powerfully validated that the value of CPUs in the era of AI is undergoing a rapid revaluation.This earnings report did more than merely trigger intense market volatility—such as AMD’s share price surging by over 16% in after-hours trading; crucially, it also revealed a structural seismic shift in current demand for computing power.

AMD Earnings Report Validates Soaring CPU Value

According to the Q1 2026 earnings report released by AMD, revenue for the quarter reached $10.25 billion—a 38% year-over-year increase—surpassing the $9.89 billion revenue previously projected by analysts.

Breaking down the results, Data Center revenue reached $5.8 billion, up 57% year-over-year, serving as the primary growth driver. Client and Gaming revenue totaled $3.6 billion, a 23% year-over-year increase; within this segment, the Client business grew by 26%, driven primarily by Ryzen CPUs. Additionally, Embedded revenue stood at $873 million, up 6% year-over-year.

AMD projects second-quarter revenue to be approximately $11.2 billion—representing a year-over-year increase of roughly 46%, with a margin of error of plus or minus $300 million—exceeding the average analyst estimate of $10.5 billion compiled by Bloomberg.

In an official AMD press release, CEO Lisa Su directly stated that robust demand for inference computing and intelligent AI agents is driving sales of high-performance CPUs and accelerators; customer orders for the MI450 series and the Helios platform have already exceeded the company’s previous expectations.
During the conference call, Su explicitly noted that server CPU revenue hit a record high for the fourth consecutive quarter, serving as the primary driver behind the growth of the data center business. In essence, AMD is attributing this current surge in data center growth to AI-driven demand for GPUs, which in turn stimulates increased demand for high-performance EPYC CPUs—thereby rapidly amplifying both revenue and profits within the data center segment.Notably, this marks the first time AMD has surpassed its long-standing rival, Intel, in terms of data center revenue. During the same quarter, Intel’s Data Center and AI Group generated $5.1 billion in revenue—a year-over-year increase of 22%. This signifies that the landscape of the global server CPU market is undergoing profound changes.
Interestingly, AMD notes that today’s cloud and internet giants have already adopted EPYC as one of the primary CPUs powering their AI infrastructure. For instance, companies such as Meta, AWS, Google Cloud, Microsoft Azure, and Tencent have all announced—or expanded—cloud instances based on the “5th Gen EPYC” architecture, deploying them across various scenarios including general-purpose computing, memory- and compute-optimized workloads, and High-Performance Computing (HPC).
Meta has also entered into a strategic partnership with AMD, announcing plans to deploy up to 6 GW of AMD Instinct GPUs. The initial 1 GW tranche will feature a custom variant of the MI450 GPU, and Meta is simultaneously becoming one of the first customers for the 6th Gen EPYC processors.
Evidently, this also implies that within large-scale AI deployments, the role of the CPU is not being diminished; rather, it works in tandem with the GPU, serving as the core node responsible for host management, scheduling, and I/O operations.

Meanwhile, AMD also stated that it expects the CPU market to grow by over 35% by 2030, surpassing $120 billion. Furthermore, the company anticipates that its CPU revenue for the second quarter will see growth exceeding 70%, driven primarily by new demand stemming from Agentic AI.
This is because, in the Agentic AI phase, AI is no longer limited to merely generating a single sentence; instead, it is required to autonomously plan tasks, utilize tools, and process complex logic. These specific functions—task scheduling, logical reasoning, and context management—happen to be the very strengths of the CPU.

The value of CPUs is being revalued.

Over the past few years, nearly all investment in AI infrastructure has flowed into GPUs dedicated to training large-scale models, while CPUs have served merely as baseline components; the typical server configuration ratio has been one CPU paired with four to eight GPUs. Today, however—as AI applications move from development into practical deployment—this ratio is rapidly evolving toward a 1:1 balance in order to accommodate the massive demands for logical scheduling and data preprocessing. In certain high-density intelligent agent application scenarios, situations have even emerged where the number of CPUs exceeds that of GPUs.

For instance, in articles related to MLPerf, Intel has stated quite candidly that AI inference is increasingly about more than just GPU throughput; it also depends on the CPU’s ability to drive overall system performance and optimize Total Cost of Ownership (TCO). The CPU is responsible for memory management, task orchestration, workload distribution, security, reliability, and operational continuity—making it the very core of AI infrastructure.
Today, many cloud providers are adopting this same approach; Microsoft, for example, emphasized during its earnings calls that a significant portion of its capital expenditure is directed toward the collaborative interplay between server CPUs and GPUs.

Furthermore, in certain specific scenarios, the user experience provided by a CPU is actually superior. For instance, with lightweight models such as MobileNetV2, once data transfer overhead is taken into account, modern CPUs can achieve lower latency than GPUs, making them better suited for edge deployment. Additionally, in cloud-based scenarios characterized by single requests, low concurrency, and real-time response requirements, CPUs can outperform GPUs in terms of response speed and availability.

I have previously provided a detailed analysis of Intel’s earnings reports, which similarly placed the CPU in a position of paramount importance. Intel CEO Pat Gelsinger went even further, explicitly stating that in recent months, clear indications have emerged showing that the CPU is once again becoming an indispensable foundation in the era of artificial intelligence; the central processing unit now serves as the scheduling layer and critical control plane for the entire AI technology stack.

Lisa Su also noted that, historically, the prevailing view was that the CPU merely served to provide processes for the GPU and handle data transfer; however, with the widespread adoption of AI applications, the CPU must now not only support general-purpose computing but also function as the primary host node for AI accelerators. Without a powerful CPU, GPUs cannot be effectively orchestrated.

Concurrently, AWS plans to commit a record-breaking capital expenditure of approximately $200 billion by 2026, with a primary focus on AI infrastructure. The annualized revenue generated by its in-house chip portfolio—specifically the Graviton CPUs and the Trainium/Inferentia XPUs—has already surpassed $20 billion; furthermore, the company disclosed that the total value of committed orders for its Trainium chips exceeds $225 billion.

AWS has also partnered with Meta, under which Meta will deploy tens of millions of Graviton CPU cores on AWS to power workloads such as Agent AI. This signifies that major industry players have effectively repositioned the CPU as a critical component of AI infrastructure.
Currently, the global CPU market has entered a “super-cycle” characterized by simultaneous growth in both volume and price. The global server CPU market is facing a severe supply-demand imbalance; lead times have stretched from the standard 1–2 weeks to 8–12 weeks, with delivery times for certain models extending to as long as six months. Prices for server CPUs from Intel and AMD have generally risen by 10%–20%, while high-end, specialized AI CPUs have seen price hikes exceeding 25%.
However, regarding future supply, the overall situation appears tight yet manageable. Although supplies of wafers, backend services, and memory remain constrained, AMD has secured long-term supply agreements and is engaged in deep collaboration with its supply chain partners—a strategy that positions the company to support its ambitious growth targets for the 2026–2027 period.

Notably, in the face of challenges from competitors Intel and Arm, AMD has chosen to take a distinct path. Lisa Su emphasized that AMD will no longer offer general-purpose CPUs; instead, it will optimize its processors for specific AI scenarios.
She stated that AMD will focus on CPUs optimized for throughput, power efficiency, and—specifically for AI infrastructure—dedicated AI capabilities (such as the Verona series). She argued that diverse workloads—including general-purpose computing, accelerator host nodes, and agent tasks—necessitate entirely different CPU designs.
Interestingly, compared to AMD’s own projections for the CPU market, UBS holds a more bullish outlook; the firm estimates that by 2030, the total addressable market for server CPUs could reach approximately $170 billion—representing nearly a fivefold increase in growth potential relative to 2025 levels.