← Back to stack

Compute

Close behindCore: IndustrialHigh: AI training, AI inference, Robotics, EV / auto, Consumer / IoT

CPU

China has meaningful domestic server and embedded CPU activity, but the global frontier remains difficult across performance, ecosystem, and ISA licensing.

The simple version

The CPU is the general-purpose brain of a computer. It runs operating systems, coordinates work, and handles the many messy tasks that are not specialized enough for a GPU or accelerator.

What the technology does

  • A CPU is judged by performance, power use, software compatibility, and the surrounding server or device ecosystem.
  • The instruction set matters: x86 dominates PCs and many servers, Arm dominates phones and is growing in servers, and RISC-V is the open alternative China likes.
  • Server qualification is slow because customers need reliability, security, compiler support, and years of software confidence.

Market scale

  • Intel and AMD are the x86 references. Arm is the licensing model to understand, because many companies build chips around Arm designs.
  • Apple showed what custom Arm CPUs can do in consumer devices; cloud companies are doing similar work in servers.
  • Chinese CPU firms are meaningful for domestic systems, but still far from replacing the global CPU ecosystem across commercial markets.
Global references
  • IntelUnited States · INTC
  • AMDUnited States · AMD
  • ArmUnited Kingdom · ARM
  • AppleUnited States · AAPL

China angle

Loongson, Phytium, Hygon, Alibaba T-Head, and Huawei Kunpeng map to different strategies: domestic ISA, licensed or inherited x86 paths, Arm-like server ecosystems, and RISC-V. Procurement policy can create volume, but global software depth is harder.

China references

How we got here

China has tried several CPU routes for two decades. The current phase is more serious because government, cloud, and AI infrastructure buyers have stronger incentives to accept domestic alternatives.

  1. 2000s

    Early domestic CPU efforts focus on strategic independence more than commercial competitiveness.

  2. 2010s

    Arm, x86-linked, and indigenous approaches develop in parallel.

  3. 2020-2024

    Government procurement and cloud sovereignty make domestic server CPUs more relevant.

  4. Next phase

    Watch real server deployments, RISC-V maturity, and whether software vendors support the platforms.

Why it matters

CPUs sit underneath AI servers, edge devices, robots, industrial controllers, and almost every computing system.

Main bottleneck

Performance per watt, software ecosystem, server qualification, and dependence on advanced processes.

Where this sector shows up

Core
Industrial
High
AI training, AI inference, Robotics, EV / auto, Consumer / IoT

What to watch

  • Server adoption
  • RISC-V progress
  • Government procurement
  • Linux ecosystem support
  • Process node disclosures

Chip pricing

15 chips from this sector are in the pricing dataset — domestic parts and their global references, each with dated, source-linked price observations. Latest figure shown per chip.

Full generation tables →