Artificial intelligence semiconductor development is progressing rapidly in China, with the capabilities of companies like Moore Threads advancing to the point that the country could meet 80% of its demand with local supply, by one estimate.

At a developer conference in Beijing last month, Moore Threads showed off its new Huagang architecture for graphics processing units, along with new GPUs including Huashan for AI and Lushan for graphics. Huashan offers performance close to Nvidia’s mainstay products, while Lushan is best-in-class globally, according to the company, which plans to begin mass production of both this year.

“The new products meet the needs of domestic developers,” CEO Zhang Jianzhong said. “There will be no more need to wait for advanced products from overseas.”

Zhang, a former vice president at Nvidia, established Moore Threads in 2020. The company debuted on the Shanghai Stock Exchange’s tech-focused STAR Market in early December and has shot into the spotlight, with a market capitalization that has exceeded RMB 300 billion (USD 42 billion).

Zhang’s goal is to help end dependence on Nvidia, which accounted for 66% of the Chinese AI semiconductor market in 2024, according to US research firm Bernstein. The figure represents a high share in a field that is key to competitiveness for both companies and countries.

Export restrictions by the US government have made it harder to access Nvidia’s top-of-the-line products. President Donald Trump has said that Washington will allow the sale of some advanced Nvidia chips to China with an added 25% fee paid to the government, but China is considering rules to limit imports.

Under the proposed five-year plan drafted by the Chinese government in October 2025, Beijing aims to achieve semiconductor self-sufficiency under a “new national system,” bringing together the public and private sectors along with financial institutions.

At the core of this plan are China’s “four little dragons” of GPUs: Moore Threads, MetaX, Biren Technology, and Suiyuan Technology, also known as Enflame.

MetaX, like Moore Threads, held an IPO on the STAR Market late last year, while Biren debuted in Hong Kong this month, and Enflame is reportedly considering a listing. All are putting they money they raise toward chip R&D.

Big established tech companies are also ramping up efforts. Baidu unit Kunlunxin plans to launch five AI chips by 2030. HiSilicon, a subsidiary of Huawei, is working on semiconductor development, and Cambricon Technologies, founded by former professors at the Chinese Academy of Sciences, is focusing on AI chips as well.

On the production side are companies like Semiconductor Manufacturing International Corporation (SMIC), which reportedly is now able to provide a stable supply of seven-nanometer semiconductors for AI applications. SMIC is said to be manufacturing chips for other Chinese companies, and it has been expanding production capacity.

Nvidia’s share of China’s AI semiconductor market is forecast to plunge to 8% this year, according to Chinese media citing a recent Bernstein report. Chinese players like Huawei and Cambricon are expected to fill the gap, reaching an aggregate market share of around 80%.

AI chips are a part of the battle for tech dominance between the US and China, and could be discussed at an expected meeting in April between Trump and Chinese President Xi Jinping.

Chinese AI chips “have evolved from ‘usable’ to ‘easy to use,'” said Zheng Weimin, a professor in the computer science department at Tsinghua University.

The country’s supply chain remains a work in progress, however. The challenge will be transitioning a development structure that had been led by Nvidia to domestic production, according to Zheng.

This article first appeared on Nikkei Asia. It has been republished here as part of 36Kr’s ongoing partnership with Nikkei.