Why Intel and AMD Dominate CPUs and What Opportunities Exist for Chinese Chipmakers

Global CPU Market Overview

Intel and AMD together control more than 80% of the worldwide CPU market in 2021, with Intel holding about 80% and AMD rapidly gaining share. Other manufacturers together account for less than 7%.

In the data‑center segment, Intel’s share fell from 81% in 2021 to 71% in 2022, while AMD rose to 20% and new players such as Amazon’s custom silicon and Ampere captured around 5%.

GPU Market Landscape

The discrete GPU market remains an oligopoly. Nvidia dominates the consumer‑PC segment with ~85% share, AMD holds ~9%, and Intel has entered the market with ~6% as of Q4 2022.

Other emerging vendors (e.g., Ampere, custom ASICs) are increasing their presence, especially in data‑center accelerators.

Key CPU Performance Parameters

CPU performance is primarily determined by clock frequency, instructions‑per‑cycle (IPC), bus widths, process node, packaging, and supply voltage. Higher frequency and IPC deliver more instructions per second, while wider address and data buses improve memory access and throughput.

Domestic Chinese CPUs often match foreign designs in clock speed, core count, and memory type, but lag in IPC, which is the main performance gap.

Instruction‑Set Architecture and Ecosystem Lock‑in

Instruction sets define the binary interface between software and hardware. The market is split between CISC (x86) and RISC families. x86, dominated by Intel and AMD, remains the standard for desktops and servers. ARM‑based RISC designs (including RISC‑V) are prevalent in mobile, embedded, and increasingly in servers due to low power and scalability.

RISC‑V is an open‑source ISA with a BSD‑style license, attracting many companies, but its ecosystem is still immature compared to x86/ARM.

AI Chip Design Trends

AI workloads require massive linear‑algebra computation, high memory bandwidth, and energy efficiency. Training and inference demand different optimizations: training emphasizes throughput (Peta‑FLOPS) and scalability, while inference focuses on TOPs/W efficiency.

Emerging trends include cloud‑based ASICs (e.g., Google TPU), edge‑device heterogeneous SoCs that combine AI accelerators, CPUs, GPUs, DSPs, and ISPs, and “software‑defined” chips that can reconfigure functionality at runtime.