--- title: "AMD, with a market capitalization of $700 billion, is the new king of the data center." type: "News" locale: "en" url: "https://longbridge.com/en/news/286714717.md" description: "AMD has transformed into a leading player in the data center market, achieving $10.25 billion in revenue in Q1 2026, with data center revenue reaching $5.775 billion. This surpasses Intel's $5.1 billion, marking a significant shift in the x86 server market. Analysts have upgraded AMD's stock, predicting a multi-year revenue growth trajectory. AMD's innovative Chiplet architecture and energy-efficient designs position it to potentially capture 50-70% of the server CPU market, solidifying its status as the 'King of Data Centers.'" datetime: "2026-05-18T04:41:48.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/286714717.md) - [en](https://longbridge.com/en/news/286714717.md) - [zh-HK](https://longbridge.com/zh-HK/news/286714717.md) --- # AMD, with a market capitalization of $700 billion, is the new king of the data center. Author: Su Yang, Tencent Technology “Su Ma” is shaping an AMD for the AI ​​era. It’s inaccurate to describe AMD solely as a hardware company now. When AMD released its Q1 2026 financial report, it essentially announced that it had completed its complete transformation from a chip manufacturer to a “computing power arms dealer” for the AI ​​era. In the first fiscal quarter of 2026, AMD achieved $10.25 billion in revenue, with its data center business further increasing from $5.38 billion in the fourth fiscal quarter of 2025 to $5.775 billion in that quarter. It must be clearly stated that this is a very crucial juncture. In Q1 of fiscal year 2024, Nvidia also reached this juncture, after which its data center business revenue began to surge, embarking on a path of quarterly sequential growth exceeding tens of billions of dollars. Wall Street sensed AMD's transformation early on. Over the past 52 weeks, AMD's stock price has surged from a low of approximately $106.98 to a high of $469.22 over the past year, with a more than 80% increase in the past 30 days alone, pushing its market capitalization above $700 billion. Bernstein analyst Stacy Rasgon upgraded AMD to "outperform," doubling his price target from $265 to $525; JPMorgan analyst Harlan Sur also pointed out that AMD's forecasts point to "a much higher-than-expected multi-year CPU revenue trajectory." The data is cold, the investors' enthusiasm is fiery, but three structural signals behind it are clear: the king of CPUs, the change in the GPU:CPU ratio from 8:1 to 1:1, and the dual-engine of GPU+CPU—this is the most significant "assignment" that Lisa Su, who has served as CEO for nearly 12 years, has delivered to the company's board of directors and Wall Street. The New King of the Data Center For decades, Intel held an absolute dominant position in the data center CPU market. Until 2017, AMD emerged as a challenger, launching its first-generation data center brand EPYC, with the flagship product at the time being the AMD EPYC 7601. For nearly a decade afterward, AMD continued its pursuit. In its most recent fiscal quarter, AMD's data center revenue reached $5.775 billion, consistently surpassing and widening the gap with Intel, whose data center revenue was $5.1 billion that quarter. This surpassing was actually foreshadowed in the past two quarters: in the third quarter of 2025, Intel's data center revenue was $4.1 billion, while AMD's was $4.3 billion. Digitime not only emphasizes that this is the first time AMD has significantly surpassed Intel in data center revenue in a single quarter, but also stresses that this trend may signal the beginning of AMD's long-term dominance, marking a historic turning point in the x86 server market landscape, a position Intel may find difficult to challenge for some time. Many call AMD the "King of CPUs," but it should more accurately be called the "King of Data Centers." In Mercury Research's Q1 2026 research report, AMD achieved a record 46.2% revenue share in the server market, compared to 39.5% in the same period of 2025. Rather than looking solely at revenue share, I prefer to consider it in conjunction with shipment share—the Mercury Research report also mentions that in Q1 2026, AMD's shipment share in the data center CPU market was 33.2%. A 33.2% shipment share corresponds to a 46.2% revenue share, which is a topic of market share efficiency. Bank of America Securities, in a May 2026 report, indicated that AMD is expected to capture approximately 50% of the server CPU market. In a simple proportional calculation, AMD could potentially capture 70% of the server CPU market share in the future. What will propel it towards becoming the "King of the Data Center" is its more flexible Chiplet architecture and a more stable roadmap—AMD's advantages in single-core performance, energy efficiency, and total cost of ownership. AMD previously set a goal of improving the energy efficiency of its AI training and high-performance computing (HPC) server processors/accelerators by 30 times between 2020 and 2025. In June 2025, AMD disclosed data showing that its combination of fifth-generation AMD EPYC CPUs and Instinct GPUs had successfully achieved a 38-fold improvement in energy efficiency compared to the 2020 baseline system. This equates to a 97% reduction in energy consumption for the same amount of computation. Meanwhile, AMD also proposed that by 2030, a typical AI model that currently requires more than 275 racks to train can be trained using only one rack, while reducing power consumption by 95% and carbon emissions from approximately 3,000 metric tons to 100 metric tons. Lower total cost of ownership corresponds to higher adoption rates. AWS, Google Cloud, and Microsoft Azure have all announced new and expanded fifth-generation EPYC cloud instances, including Google Cloud H4D virtual machines and Azure cross-workload optimized instances. Oracle announced that it will deploy its first 50,000 MI450 GPUs in the third quarter of 2026, becoming the world's first publicly available hyperscale AMD computing cluster. With widespread customer adoption, the only thing that might stop AMD is production capacity. At the Morgan Stanley conference in March, when asked if CoWoS packaging capacity was sufficient, Lisa Su replied, "We absolutely have enough CoWoS capacity. The best answer I can give you is that we have capacity, technology, strong customer relationships, and data center service providers have already allocated space for it." CPUs Take Center Stage At the "Advancing AI" event in June 2025, "Mama Su" presciently mentioned the explosive growth in inference demand. However, at that time, CPU demand was still overshadowed by the inference performance of GPUs and ASIC chips. With the emergence of agent products like Openclaw since the beginning of the year, Lisa Su has also updated her perspective on the CPU narrative. At the Morgan Stanley conference in March, she said, "Even hyperscale cloud vendors are surprised. The CPU computing demands accompanying AI may be a severely underestimated variable." AMD's performance this quarter is noteworthy not only for the growth in its revenue figures but also for the growth potential corresponding to management's assessment of market trends. During the first fiscal quarter conference call, Lisa Su said, "In the past, the CPU to GPU ratio was basically a 1:4 or 1:8 head-node configuration. Now, this ratio is approaching 1:1." This change in quantity essentially reflects that AI computing infrastructure is moving away from a single "accelerator model" and into a "balanced computing era." Goldman Sachs, in its latest research brief, analyzed that the essence of agents is "action" rather than "prediction," and this shift in focus from \*\*vector computing to logical reasoning\*\* directly leads to a structural return of computing load to the CPU. What changed this narrative logic was the explosive growth of Agents. Previously, the standard configuration of a data center was one CPU paired with four to eight GPUs. The CPU was merely a "gatekeeper" responsible for scheduling and management, a practice that persisted from ChatGPT to large-scale model training. With the advent of Agents, AI assistants shifted from "answering questions" to "autonomously completing tasks," requiring continuous planning, tool usage, result checking, and replanning. These intensive logical judgments and task orchestrations necessitate the use of CPUs. While the demand for Agent inference is an objective factor, Wall Street, overshadowed by the GPU narrative, also needed a new story. A Bank of America research report predicts that the global data center CPU market will surge from $27 billion in 2025 to $110 billion in 2030. \*\*AMD's management, during its first fiscal quarter earnings call, aggressively doubled its 2030 total potential market target for server CPUs from a CAGR of 18% to over 35%, aiming for $120 billion.\*\* An Evercore ISI report introduced the concept of a "CPU renaissance." Similar discussions have emerged recently, essentially referring to the CPU returning to center stage. The market share is changing, and so is the market size. AMD, having just become the king of the data center and possessing a lower total cost of ownership, is poised for its best.\*\* Dual-engine acceleration is driving rapid CPU demand growth, gradually propelling AMD towards a trillion-dollar market capitalization. However, even with adjustments in demand, CPU growth doesn't necessarily mean replacing GPUs. Within AMD's ecosystem, they are more like two engines accelerating simultaneously. During the executive presentation of the financial report, Lisa Su clarified this dual-driving approach, stating, "Strong demand for EPYC server CPUs and Instinct GPU shipments will continue to climb." The simultaneous acceleration of both CPU and GPU businesses is rare in the semiconductor cycle. It signifies that AMD is no longer relying solely on the performance of individual chips to compete in the market, but rather providing system-level solutions through a combination of CPUs and GPUs. This represents a transformation from a single computing power provider to a full-stack computing solution provider, optimizing the total cost of ownership for computing power customers. The deep collaboration with Meta is a prime example of this strategy. In February 2026, Meta and AMD signed not just a simple GPU procurement agreement, but a bundled order for 6 gigawatts of Instinct MI450 GPUs with sixth-generation EPYC server CPUs. Meta chose this deep bundling solution primarily for end-to-end optimization of the trillion-parameter Llama 4 and its subsequent agent clusters. The CPU+GPU bundle effectively reduces system-level instruction latency. For example, based on Infinity Fabric interconnect technology, it breaks down the "memory wall" problem of traditional architectures, enabling seamless, ultra-low-latency data exchange and unified memory sharing between EPYC processors and Instinct accelerators. Lisa Su explained in a conference call, "We design chips based on actual workloads." AMD, based on the operating characteristics of the Llama model, achieved customized development from the underlying architecture to the system level. Meanwhile, Goldman Sachs also pointed out in a research report that the "CPU+GPU" package not only increases the value of a single order, but also, through deep architectural binding, makes it difficult for cloud vendors to switch to other solutions after use. This trend is also reflected in the contract with OpenAI for 6GW of GPU computing power: the CPU deployment plan is simultaneously incorporated into its GPU supply agreement. "Customers are already planning the deployment of CPUs and accelerators simultaneously, which is a key market signal," said Lisa Su. Financial data confirms the correctness of this logic. Server CPU revenue grew by over 50% year-on-year in the first quarter, and the guidance for the second quarter was revised upwards to over 70% growth. At the same time, the Instinct GPU series is in the ramp-up phase of shipments, and the MI450 will enter a large-scale mass production phase in the second half of the year. These two engines are working together to amplify the growth potential of AMD's data center business. In the past, NVLink and CUDA built a closed ecosystem, while AMD has always adhered to the concept of openness. In this new track defined by agents, AMD is attempting to leverage its existing advantages in the CPU market, combined with its GPU platform portfolio, to deliver a more universal and difficult-to-replicate "dual-engine standard," thus breaking down Nvidia's closed system. Taking UALink as an example, as an open interconnect standard jointly promoted by AMD and its global industry partners, it has already attracted more than 70 members worldwide, including several Chinese companies. Therefore, if asked what the difference is between AMD and Nvidia, this might be a concrete example of that difference. ## Lisa Su's Second Half With the three stories of becoming the king of data centers, the rise of CPUs, and the dual-engine of GPU+CPU, Wall Street and investors have given AMD a high score of $700 billion in market value, which is also a positive feedback on Lisa Su's nearly 12-year CEO career. In a recent interview, when faced with the question of the "glass cliff," she said: "When I took over, the outside world was generally pessimistic, thinking that AMD was about to go bankrupt, but what I saw was a company with a great engineering foundation, just with unstable execution. I didn't see it as a trap, but as my dream job." If fully integrating AMD into the AI ​​era was the goal of the first half, then the goal of the second half should be the explosive growth in performance after the transformation, and the most recent milestone reflected in market value is $1 trillion. At the November 2025 Financial Analyst Day, Lisa Su stated: "We aim for an average annual growth rate of over 60% over the next three to five years, pushing annual data center revenue to $100 billion, and achieving annual earnings per share exceeding $20 within our strategic timeframe." Based on this quarter's $5.77 billion, AMD's annualized data center revenue is approximately $23 billion. To reach $100 billion within five years, this figure needs to more than quadruple. Supporting these goals is a product chain with no room for error: from EPYC Venice and MI450 to Helios rack systems, and then to the ROCm ecosystem. Regarding ROCm, there's a rather interesting twist. In December 2024, Dylan Patel, founder of SemiAnalysis, published a report specifically pointing out numerous vulnerabilities in AMD's ROCm software stack. Less than 24 hours after the report was released, Lisa Su personally contacted Patel. The next morning at 7 a.m., she sat there, listening to the SemiAnalysis engineering team report on every bug and improvement suggestion for a full 90 minutes. It was precisely because of this attitude that SemiAnalysis rewrote the article in April 2025, reversing the conclusion from "failed to become an effective competitor" to "AMD has entered wartime mode, and the MI450X is expected to directly compete with Nvidia in the second half of 2026." Lisa Su also invited OpenAI President Greg Brockman, Luma AI CEO, Liquid AI CEO, and AI guru Fei-Fei Li to the CES stage. Wherever computing power flows, Lisa Su invites the people from that area to the stage. ### Related Stocks - [AMD.US](https://longbridge.com/en/quote/AMD.US.md) - [AMDL.US](https://longbridge.com/en/quote/AMDL.US.md) ## Related News & Research - ['It's Just the Beginning,' Says Investor About AMD Stock](https://longbridge.com/en/news/288194764.md) - [AMD’s new pitch: our old tech is so good you should just keep using it](https://longbridge.com/en/news/288188926.md) - [AMD, Nvidia, and Intel Face a New AI Battle — Wolfe Research Picks Its Biggest Winner, and It's Not NVDA](https://longbridge.com/en/news/287927847.md) - [Advanced Micro Devices (NASDAQ:AMD) Sets New 1-Year High - Time to Buy?](https://longbridge.com/en/news/287893397.md) - [What's Going On With AMD Stock Thursday?](https://longbridge.com/en/news/287905158.md)