Chinese companies are developing more efficient computing technologies to keep pace with the rapid growth of artificial intelligence (AI), with advances in both processing power and energy efficiency.

At Shenzhen Pinghu Laboratory, a team led by director Wan Yuxi has developed a gallium nitride crystal film on an 8-inch silicon carbide substrate to improve power delivery to graphics processing units (GPUs).

The technology can increase device efficiency from around 85 percent to 95 percent. With China’s AI computing data centres projected to require electricity equivalent to the output of five Three Gorges hydropower stations by 2035, the improvement could cut energy losses by an amount equivalent to half a station’s electricity output.

Lanxin Computing has developed chips that can run large AI models on all-in-one computers. Demand for inference computing is rising as AI agents become more widely used, particularly in government services, healthcare and finance, leaving the company’s products in short supply.