AI Agent Era Drives CPU Renaissance, 2030 Market Space Projected at $223 Billion

Overview

This report explores the impact of the generative AI paradigm shift from 1.0 (chatbots) to 2.0 (autonomous agents/Agentic AI) on semiconductor hardware architecture. The core view is that this shift will greatly increase demand for server CPUs, leading to a renaissance of CPUs in AI data centers. Institutions raised the 2030 Server CPU Total Addressable Market (TAM) from a previous $137 billion to $223 billion. Against this backdrop, Arm architecture, with its superior performance-per-watt, becomes the largest beneficiary, while x86 players AMD, Intel, and China-based Hygon Information Technology will also benefit from the overall market expansion. The report correspondingly raises target prices for Arm, SoftBank, AMD, Intel, and Hygon Information Technology.

Core views

Demand-side logic reconstruction: Changes in AI workload nature are the root cause of CPU demand explosion. In traditional LLM training and simple inference, GPUs are the absolute core, with CPUs playing an auxiliary role; the CPU:GPU ratio was as low as 1:8. However, in the Agentic AI era, AI systems need to execute complex task orchestration, tool calling, memory management, and multi-step reasoning loops. If CPU performance is insufficient, expensive GPUs will sit idle waiting. Therefore, in next-gen AI infrastructure, the CPU:GPU ratio will return to 1:1 or higher. Institutions, based on a $3.5 trillion AI data center CapEx assumption for 2030, calculate that 2030 Server CPU TAM will reach $223 billion, with $174 billion coming from Agentic AI workloads. Arm's Structural Advantages and Business Model Leap: Arm architecture, with high energy efficiency and high core density, perfectly fits the strict power and space limits of AI data centers. AWS, Azure, Google, and other cloud giants have deployed self-developed Arm CPUs on a large scale. More importantly, Arm is undergoing a major business model transformation, shifting from a pure IP licensor to a direct silicon supplier (Arm AGI CPU). Institutions predict that by FY31, Arm's AGI CPU business revenue will reach $22 billion, contributing approximately $7.7 billion in operating profit. Although this transformation may dilute gross margin percentage-wise, it will significantly expand the absolute profit pool. Core Target Beneficiary Logic: Arm is the core beneficiary of this CPU renaissance, expected to hold a $123 billion share of the server CPU market by 2030. AMD, leveraging product strengths in the x86 domain and market share growth trajectory, will continue to benefit from strong server demand growth. Although Intel faces competitive pressure, the recovery of its server business will help stabilize it. In the China market, Hygon Information Technology benefits from localized substitution of x86 demand and accelerating AI investment, expected to exceed 35% share of China's x86 server CPU market by 2030. NVIDIA, primarily viewed as a GPU giant, plays a key role in Agentic AI systems via its Grace/Vera CPU series through NVLink high-speed interconnect technology, further consolidating its ecosystem.

Analysis framework

The institution adopted an analytical framework combining top-down and bottom-up approaches. First, a top-level TAM model was built through macro assumptions (AI CapEx, GPU market size, CPU:GPU ratio, average CPU selling price), yielding a base-case prediction of $223 billion. Second, cross-validation was conducted by breaking down downstream market demands, classifying server CPU demand into four categories: NVIDIA-related Arm CPUs, hyperscaler self-developed CPUs, Arm's self-developed AGI CPUs, and other commercial Arm CPUs. This dual-track validation ensured the robustness of the forecast. Additionally, the report used sensitivity analysis to discuss the TAM range under different CapEx levels and ratio assumptions (Bear case $137 billion, Bull case $330 billion).

Methodology notes

• Industry/Industrial Analysis FrameworkSupply and Demand Framework

  Hardware ratio changes caused by AI workload evolution

  The research report revealed the rising bottleneck status of CPUs in the system by analyzing the process of AI evolving from simple computing to complex task orchestration, thereby deriving supply-demand structure changes where the CPU:GPU ratio returns from 1:8 towards 1:1.

• Industry/Industrial Analysis FrameworkQuantity-Price Decomposition

  TAM model quantity-price driver decomposition

  When calculating the $223 billion market size, the report decomposed total scale into multiple driving factors including total AI data center CapEx, GPU proportion, quantity ratio between CPU and GPU, and average selling price (ASP) of CPUs, facilitating investor understanding of how each variable affects the final market size.

• Competition and Strategy FrameworkValue chain analysis

  Arm vertical integration from IP licensing to silicon sales

  The report points out that Arm extends from an upstream IP provider to a downstream chip supplier by launching self-developed AGI CPUs, aiming to capture a larger value pool. Although this may bring structural changes to gross margins, it can increase absolute profit amounts.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Arm Holdings (ARM.US)

  Core beneficiary, Agentic AI drives increased penetration rate of Arm architecture at the server side, and the company's transition to self-developed chips opens new revenue growth poles.

  Strengths

  Extremely high energy efficiency, extensive cloud vendor ecosystem support, business model expanding from IP to silicon.

  Weaknesses

  Self-developed chips may cause customers' concerns about competitive relationships, gross margins may be diluted due to hardware business.

  Comparison

  Compared to x86 vendors, significant advantages in energy efficiency and customization.

  Risks

  Customer self-developed chip progress fails to meet expectations, geopolitical risks.

• Advanced Micro Devices (AMD.US)

  Main beneficiary in x86 camp, continuously seizing server market share with excellent product performance.

  Strengths

  Strong EPYC processor product line, providing CPU and GPU solutions simultaneously in AI data centers.

  Weaknesses

  Facing Intel's competitive pressure and long-term erosion by Arm architecture.

  Comparison

  Outperforms Intel in general server market, but needs to compete with NVIDIA in AI-specific domains.

  Risks

  Intensified market competition, IT spending decline due to economic downturn.

• Hygon Information Technology (688041.SS)

  Core target for localized substitution of China's x86 server CPUs, benefiting from domestic AI investment and supply chain security needs.

  Strengths

  x86 instruction set license, good compatibility with domestic AI chips, solid government and SOE customer base.

  Weaknesses

  Advanced process supply may be limited, reliance on domestic market.

  Comparison

  Has monopoly advantage in local Chinese x86 market, expected share over 35% by 2030.

  Risks

  Supply chain sanction risks, technology iteration speed lagging behind international giants.

• SoftBank Group (9984.JP)

  Indirect beneficiary, as a major shareholder of Arm, its net asset value increases with Arm valuation rise.

  Strengths

  Holds large stake in Arm, diversified investment portfolio.

  Weaknesses

  Stock price fluctuation heavily influenced by Arm performance and other investment assets.

  Comparison

  Compared to holding Arm stock directly, provides leverage effect and diversification.

  Risks

  Arm stock price fluctuation, poor performance in other investment sectors.

Key data

• 2030 Server CPU TAM (Base)$223 BillionUp 6x from $37 billion in 2025, assuming $3.5 trillion AI CapEx and CPU:GPU ratio of 1:1
• Arm Target Price (New/Old)$500 / —Based on 42x P/E, implying 21% upside
• AMD Target Price (New/Old)$600 / —Maintains Outperform rating
• Hygon Information Technology Target Price (New/Old)CNY 450 / CNY 280Based on 2028 EPS of CNY 6.3 and 71x P/E
• Arm AGI CPU FY31 Revenue Forecast$22 BillionPreviously management guidance was $15 billion, report predicts more aggressively
• 2030 Arm Share of Server CPU Market~55%Corresponding to $123 billion market share, where NVIDIA-related demand accounts for $58 billion

Impact & implications

For the industry, this means semiconductor investment focus will expand from single GPU compute to system-level capabilities for CPU-GPU synergistic optimization. For cloud service providers, adopting Arm architecture will become a key strategy to reduce energy consumption and Total Cost of Ownership (TCO). For investors, companies capable of providing high-efficiency CPU solutions in the Agentic AI era should be watched, especially Arm and its ecosystem partners. At the same time, led by Hygon Information Technology, domestic Chinese chip manufacturers are expected to achieve excess growth through localized substitution against the background of restricted access to the global x86 supply chain.

Risks

• Wafer fab and memory capacity may not be sufficient to support rapid CPU demand growth.
• Hyperscalers may bypass integrators like NVIDIA and purchase HBM directly from memory suppliers, changing value chain allocation.
• Arm's self-developed chip business may trigger concerns about "refereeing and playing as a player" interest conflicts.
• Geopolitical factors may lead to supply chain disruptions in the China market or limitations on technology acquisition.

What to watch

• Deployment progress and penetration rate changes of Arm CPUs by major cloud providers (AWS, Azure, Google).
• Release schedule and market acceptance of NVIDIA Vera CPU and subsequent products.
• Actual revenue realization and profit margin performance of Arm AGI CPU business.
• Policy dynamics regarding advanced process acquisition for domestic x86 CPU manufacturers (such as Hygon) in China.