Chinese chip start-up Dishan Technology has achieved a breakthrough in designing a 2-nanometre artificial intelligence chip, according to local media reports.

Shanghai-based Dishan, which focuses on the development of high-performance computing chips and sensor chips, was now in the crucial prototype verification stage for its first 2nm AI graphics processing unit (GPU), the Shanghai Morning Post reported recently.

The GPU was unveiled by the company last July. The company then said it had completed the basic design process on the chip, calling it a milestone in China’s quest to achieve self-sufficiency in high-end computing chips. It employed a hybrid FinFET/GAA process and a chiplet-based heterogeneous architecture, which was expected to improve energy efficiency by 40 per cent compared with its predecessor, according to the company.

To make the new chip compatible with Nvidia’s Compute Unified Device Architecture (CUDA) ecosystem, Dishan would improve its supporting toolchains such as CUDA-compatible compilers, local media reported, citing a company representative.

The progress marks China’s latest effort to become self-reliant in high-end AI chips amid a prolonged tech war with the US. US chip giant Nvidia, meanwhile, was seeking to return to China with the H200, its second-most powerful AI processor to date.

However, Dishan’s chip had yet to enter the tapeout stage, and was expected to take another year or two before reaching volume production and commercial deployment, local media reported.