The artificial intelligence sector, once dominated by Graphics Processing Units (GPUs), is undergoing a fundamental architectural shift. New data suggest that the increasing complexity of AI workloads—particularly those involving autonomous agents—is rapidly elevating the critical role of Central Processing Units (CPUs). Major industry players, including AMD, Nvidia, Arm, and Intel, are responding to this shift by issuing dramatically increased market forecasts, pointing to a potential market valuation exceeding $120 billion.
The Shift to Agentic AI Workloads
The demand surge is attributed to the maturation of AI from simple chatbot interactions to highly complex, autonomous agentic applications. Unlike non-agentic tasks, which handle straightforward requests, agents are capable of managing hundreds of concurrent tasks and reasoning through problems with limited human direction. This structural difference mandates a higher reliance on CPU processing power.
Analysis by Intel and Georgia Tech highlights the technical necessity of this shift. According to their research, “tool-dominated agentic AI workloads are significantly bottle-necked” by the CPU, which can consume up to 88% of the total end-to-end latency. The paper further suggests that “with better quality GPUs, the bottleneck can swiftly shift more towards CPUs.” This indicates that for agentic AI to scale efficiently, CPU orchestration capacity must match GPU reasoning capacity to minimize lag and prevent underutilization.
CPUs are critical because they manage the orchestration—the process of calling tools, directing API requests, and coordinating tasks among multiple independent AI agents. While GPUs continue to handle the core reasoning and inference, the CPU acts as the central conductor, directing the resources allocated by the GPUs.
Market Forecasts and Demand Projections
Industry analysts and manufacturers are adjusting their long-term forecasts to reflect the dramatic increase in CPU demand. These predictions indicate a massive, multi-year growth period for the server CPU market.
- AMD recently adjusted its server CPU market forecast, nearly doubling its expected Compound Annual Growth Rate (CAGR) to 35%. The company now projects the market will surpass $120 billion by 2030.
- Arm made a similar projection in March, estimating the total addressable market (TAM) for data center CPUs will exceed $100 billion by its fiscal year 2031. This represents more than a 4X increase over its previous TAM estimate of $24 billion, equating to a 33% CAGR.
The anticipated scale of this change is reflected in multiple analyst reports:
- UBS projects the market will grow from $31 billion in 2025 to $170 billion in 2030, or a 40.6% CAGR, noting that AI CPUs alone are expected to rise from $7 billion to $125 billion, or an 88% CAGR.
- Bank of America forecasts a TAM expansion from $43 billion in 2026 to $125 billion in 2030, with a CAGR of 30.6%.
- Citi anticipates the overall market growing from $29.3 billion in 2025 to $132 billion in 2030, projecting a 35% CAGR. Citi specifically estimates that agentic CPU growth will hit $59.4 billion in 2030, representing a massive 185% CAGR.
These high growth rates are unprecedented, as the market historically experienced single-digit annual growth. The sheer scale of the projected growth means that the supply chain was unprepared for the current rate of procurement.
Competitive Landscape and Supply Challenges
Supply Chain Shortages and Pricing Power
The rapid acceleration in demand has already led to noticeable supply constraints. Reports indicate that Intel has a significant backlog of unfulfilled CPU orders, with delivery times stretching as long as six months. AMD products have seen delivery times ranging from eight to ten weeks. Furthermore, electronic equipment distributor Fusion Worldwide estimated that Intel distributors are fulfilling only around 40% of their yearly backlog allocations, citing domestic lead times of 8 to 22 weeks.
These shortages have given CPU vendors substantial pricing power. Reuters reported that Intel and AMD have raised prices by between 10% and 35% quarterly-over-quarter (QoQ). Fusion Worldwide noted that AMD’s EPYC CPUs are estimated to be effectively sold out in 2026.
Vendor Strategies
Leading companies are implementing distinct strategies to capture market share:
AMD
AMD reported record server CPU revenue in Q1 2026, with sales increasing over 50% year-over-year (YOY). The company expects Q2 growth above 70% YOY. AMD plans to push its Venice family of EPYC CPUs, which includes Verano, a processor designed specifically for AI infrastructure. AMD is targeting a 50% market share in the server CPU market by 2030.
Nvidia
While traditionally associated with GPUs, Nvidia is aggressively targeting the CPU market. Its new standalone Vera rack, which contains 256 CPUs, marks a major architectural shift, allowing customers to deploy CPU capacity without proportionally increasing GPU counts. Nvidia stated that the Vera rack opens up a $200 billion CPU TAM for the company, and projects generating nearly $20 billion in CPU revenue this year alone.
Arm
Arm is leveraging its high performance-per-watt advantage with its AGI CPU. The firm claims its AGI CPU can deliver up to 2 times greater performance per watt compared to Intel and AMD’s x86 architecture. The company believes the market demand is even higher than currently estimated, suggesting that the number of CPU cores “probably will” exceed the number of GPU cores.
Intel
Intel is countering the market shift by announcing its intention to deploy rack-scale CPU systems. Intel’s Xeon 6+ chips offer 288 cores per chip. While Intel has a lead in individual core counts, its Xeon 7 ‘Diamond Rapids’ chip, originally expected in late 2026, has been rumored to be delayed until 2027, potentially giving competitors a temporary advantage.
The Future of AI Infrastructure
The overall trend points toward a necessary increase in the CPU-to-GPU ratio within AI clusters to manage the rising complexity of agentic AI. Nvidia CEO Jensen Huang framed CPUs as additive, arguing that while more AI agents increase GPU demand (for inference), they also increase CPU demand (for orchestration).
This transition is fundamentally changing the AI infrastructure buildout. The biggest opportunity is no longer simply adding CPUs as head nodes next to GPU clusters, but moving toward standalone CPU racks. This architectural shift allows for greatly increased orchestration capacity without needing a corresponding increase in GPU units, defining CPUs as the next major bottleneck in the burgeoning AI sector.
