The Zhitong Finance App learned that Bank of America pointed out in the latest semiconductor industry research report released on November 12 that with the accelerated evolution of artificial intelligence (AI) applications from traditional big model training and inference to intelligent AI, the importance of central processing units (CPUs) in data centers is increasing significantly. The bank raised the global server CPU potential market forecast for 2030 from about 210 billion US dollars to more than 240 billion US dollars, and believes that AMD (AMD.US) is expected to be the main beneficiary of this growth trend with its leading product portfolio and supply capacity.
Bank of America analyst Vivek Arya and his team anticipate that the global server CPU market will increase from about 35 billion US dollars in 2025 to 64.1 billion US dollars in 2026, and reach 24.02 billion US dollars in 2030, an increase of nearly 6 times over 2025. The compound annual growth rate from 2026 to 2030 will reach 39.2%, higher than the previous forecast of about 36%.
Although the market is concerned that AI capital expenditure may slow down after 2028, Bank of America pointed out that judging from GPU spot rental prices and memory spot prices, AI computing power demand is currently at an all-time high, while the popularity of smart AI is still low, which means that demand for related infrastructure is still in the early stages of growth.
Bank of America believes that the core issue for future growth in the server CPU market is whether AI demand can continue to drive CPU shipments and average sales prices to rise simultaneously. The bank is optimistic about this. It is estimated that global server CPU shipments will increase from 38.6 million units in 2026 to 86.2 million units in 2030, with a compound annual growth rate of 22.3%; in the same period, the average price will rise from 1,660 US dollars to 2,787 US dollars, with a compound annual growth rate of 13.8%.
Looking at specific application areas, the Bank of America predicts that by 2030, the CPU market for traditional enterprise servers and cloud computing infrastructure will reach about 30 billion US dollars, while the CPU market for AI computing cluster management and coordination of GPU operation will exceed 105 billion US dollars, and the CPU market size dedicated to scheduling and execution of intelligent AI tasks will also reach about 105 billion US dollars.
This means that AI-related applications will contribute about 210 billion US dollars in server CPU requirements, accounting for about 87% of the overall market size. Among them, the market size of smart AI-related CPUs is expected to increase from about 6.8 billion US dollars in 2026 to 105 billion US dollars in 2030, with a compound annual growth rate of close to 99%.
Unlike traditional chatbots, intelligent AI can not only answer user questions, but can also independently plan tasks, call external tools, perform multi-step operations, and continuously process and manage information. These functions require the CPU to take on a large amount of task scheduling, memory management, and data processing tasks, thereby further increasing the value of the CPU in AI infrastructure.
Bank of America pointed out that currently the CPU in AI servers is mainly responsible for coordinating GPU operation, managing computing clusters, and transmitting data. However, as intelligent AI applications become more complex, independent CPU servers and cabinets dedicated to performing related tasks will become a new area of growth. For example, Nvidia (NVDA.US)'s Vera CPU dedicated cabinet is a reflection of this trend.
According to Bank of America's estimates, under the current resource allocation model for smart AI applications, every 1 billion users may bring about 2.7 million new CPU requirements. However, as intelligent AI gradually becomes an important tool in the daily work of individuals and enterprises, the computing resources required by individual users will increase significantly, and the CPU requirement for every 1 billion users may rise to about 21 million in the future.
The bank predicts that by 2030, the number of global smart AI users is expected to reach 1 billion to 2 billion, and the annual shipment volume of CPUs dedicated to smart AI will increase more than tenfold from about 2.5 million in 2026 to 28.7 million in 2030.
Notably, Bank of America emphasized that the increase in CPU demand does not mean that GPUs will become less important. On the contrary, the development of intelligent AI will further expand the market size of the entire AI computing system, and CPUs and GPUs will play complementary roles in different types of computing tasks.
The bank estimates that by 2030, the global data center system market will reach about 2.8 trillion US dollars, of which the AI data center system market is about 2.4 trillion US dollars and the AI accelerator market is about 1.6 trillion US dollars. The share of server CPUs in the total market value of data center systems is expected to rise from less than 7% to about 9% from 2024 to 2025.
In terms of individual stocks, Bank of America continues to list AMD as the preferred target in the server CPU sector, maintaining its “buy” rating, and raising the target price from $650 to $700. The valuation is based on a price-earnings ratio of 28 times the expected earnings per share in 2028, which is 26 times higher than before.
Bank of America believes that AMD has a broad portfolio of server CPUs and has leading advantages in terms of both high operating frequency and high number of cores. The former is beneficial to improving the operating efficiency of AI computing clusters, while the latter helps to process more intelligent AI tasks simultaneously.
Specifically, AMD's fifth-generation EPYC Turin server processor has a maximum of 192 cores, while the next generation Venice processor based on the Zen 6 architecture and 2 nm process will have up to 256 cores and 512 threads. Bank of America believes that AMD's advantages in product performance, advanced manufacturing processes, and supply assurance are expected to help it further expand its server CPU market share.
The bank expects AMD server CPU sales to increase from US$18.5 billion in 2026 to US$84.4 billion in 2030, with a compound annual growth rate of 46.2%, and its share of sales in the global server CPU market will rise from about 29% to 35%.
In contrast, Bank of America is relatively cautious about the short-term outlook for Intel (INTC.US). Although the bank believes that Intel's foundry business still has potential growth opportunities, its CPU supply prospects for the fourth quarter of this year and 2027 are still uncertain enough. As a result, Bank of America maintained Intel's “neutral” rating and lowered its price target from $100 to $80.
Bank of America expects Intel's share of sales in the global server CPU market to drop from about 34% in 2026 to 22% in 2030. However, benefiting from overall market expansion and rising average product prices, Intel server CPU sales are still expected to increase from US$21.9 billion in 2026 to US$53.1 billion in 2030, with a compound annual growth rate of about 25%.
Meanwhile, ARM architecture-based server CPUs are expected to become the product category with the fastest market share growth. Bank of America predicts that with Nvidia Vera processors, AGI CPUs independently developed by Arm (ARM.US), and Qualcomm (QCOM.US) server CPUs, and chips developed by major cloud computing service providers, the sales share of ARM architecture processors in the global server CPU market will rise from 37% in 2026 to 43% in 2030, corresponding to a market size of approximately US$102.8 billion.
However, Bank of America believes that more of the value created by the growth of the ARM architecture server CPU market will be obtained by technology authorized customers such as Nvidia and large cloud computing service providers, and that Arm's own server CPU products are not expected to achieve a certain scale of shipment until 2028 to 2030. The future of Qualcomm's server CPU business is also uncertain, mainly because it faces multiple competitors among key customers.
Regarding Nvidia, Bank of America believes that with its overall competitive advantage in the field of AI computing systems, the company is accelerating the expansion of the Vera CPU deployment scale and is expected to become an important player in the server CPU market. The bank expects Nvidia server CPU sales to increase from US$20.4 billion in 2026 to US$60.6 billion in 2030.
However, considering the high volatility Nvidia has faced recently, Bank of America lowered its target price from $350 to $280 and lowered the valuation multiplier from 22 times to 18 times the expected earnings per share in 2027.
The latest industry data also shows that the competitive landscape in the server CPU market is changing. According to Mercury Research data quoted by Bank of America, in the third quarter of 2026, global server CPU shipments increased 31% year-on-year and 6% month-on-month, which is significantly better than the PC CPU market performance.
Among them, Intel server CPU sales increased 14% month-on-month to US$5.9 billion, a growth rate lower than AMD's 18% and the 28% forecast for Arm-based processors. Due to restrictions on the supply of wafers and packaging substrates, Intel's server CPU sales share fell 1.9 percentage points month-on-month to 34.3%, while AMD and ARM architecture processor shares reached 28.8% and 36.9%, respectively, which are close to the highest levels in history.
Bank of America believes that the development of intelligent AI is opening a new growth cycle for the server CPU market. As AI applications gradually shift from simple model training and content generation to agents that can perform complex tasks autonomously, the shipment volume, average price of CPUs, and their value in data center systems are expected to continue to increase. Under this trend, chip manufacturers with leading product performance, sufficient supply capacity, and perfect technology ecosystem will have a competitive advantage, and AMD is still the bank's most promising investment target for server CPUs.