Why China’s GPGPU Market Is Poised for Explosive Growth by 2025

In August 2020 the Chinese State Council issued policies to accelerate the integrated circuit and software industries, emphasizing their strategic role in the new wave of technological revolution.

What is GPGPU?

General‑Purpose computing on Graphics Processing Units (GPGPU) uses graphics‑oriented processors to perform general‑purpose calculations that would otherwise run on CPUs. Its massive parallelism, high energy efficiency, and wide memory bandwidth make it ideal for data‑intensive workloads such as AI model training, inference, and high‑performance computing (HPC).

Current Applications

GPGPU is widely deployed in AI training and inference, HPC, security, government projects, cloud data centers, and many industry‑specific AI use cases (e.g., video analysis, content generation, medical imaging, recommendation systems, and autonomous‑driving simulations).

Market Forecast

Forecasts predict that by 2025 China’s GPGPU chip and board market will reach ¥458 billion, a more than five‑fold increase from ¥86 billion in 2019, with a compound annual growth rate of 32 %. By sector, cloud and internet data centers will account for ¥228 billion, security and government data centers ¥142 billion, AI applications ¥37 billion, and HPC ¥28 billion. By application, AI inference is projected at ¥286 billion, AI training at ¥144 billion, and HPC at ¥28 billion.

Domestic Substitution Challenges

Despite the booming demand, the global GPGPU supply chain is dominated by a single vendor, leading to high prices and limited product variety. In China’s cloud AI training market, the leading supplier holds about 90 % of the market, with one product covering 50 % of the share. This concentration, combined with export restrictions on key foreign vendors, creates a strong impetus for domestic alternatives.

Policy Support and Talent Pool

The government’s “Several Policies” provide fiscal incentives, financing, R&D support, talent development, and IP protection to encourage domestic chip development. China’s experienced chip engineers, many of whom have worked at AMD, NVIDIA, Oracle, and IBM, form a solid talent base to drive indigenous GPGPU projects.

Case Study: Domestic Effort

TianShu ZhiXin, founded in early 2018, assembled a design team with decades of experience from leading global GPU and semiconductor companies. The company aims to launch its first domestically‑designed GPGPU for AI training in 2021, followed by an inference‑focused chip. It has already partnered with server manufacturers such as Inspur and New H3C, as well as domestic CPU leader Loongson, to develop integrated AI/HPC solutions for data centers.

Overall, the analysis shows that GPGPU demand in China will continue to surge, making domestic development and supply diversification critical for national security and technological independence.