In an industrial park in Zhangjiagang, a small city on China’s east coast, a large humming and hissing machine feeds on piles of used clothes and sorts them.
The novelty? It uses artificial intelligence to sort them by composition at high speed, offering a glimpse into how AI could play a role in reducing the impact of synthetic textile waste.
The Fastsort-Textile machine, named one of Time magazine’s Best Inventions of 2025, was created by DataBeyond, a Chinese AI recycling company founded in 2018.
“We can make full use of textile waste and reduce the amount that is incinerated which will be a great help to recycling resources,” DataBeyond CEO Mo Zhuoya said.
Synthetic textiles are derived from fossil fuels and are a low-cost, popular option for fashion production. Altogether they account for around 70% of global textile production, according to a report from Amsterdam-based nonprofit Circle Economy, which analyzes ways to reduce textile waste.
Textile waste is a major global pollutant, with China as the leading contributor. China led global textile exports at $142 billion, more than double that of the European Union, according to the World Trade Organization’s 2025 Key Insights and Trends report.
Fastsort-Textile is being used only in one location in China: Shanhesheng Environmental Technology Ltd., a textile recycling facility in Zhangjiagang that installed the machine in 2025.
The equipment uses an AI scanner to read the composition of such textiles and sorts them by fibers, after which they can be recycled.
Fastsort-Textile sorts through 100 kilograms (220 pounds) of clothes in two to three minutes , compared to around four hours for one worker to do the same thing. The machine can process two tons per hour, while two people would need two days and at reduced accuracy, according to analysis by Shanhesheng.
The AI scanner measuring 5-by-2 meters (16-by-6.5 feet) works with a series of conveyer belts. Workers load stacks of textiles onto belts that move them through the scanner, which emits a sharp hiss while reading the textiles’ composition. A live video feed displays the reading on the scanner’s side.
It takes less than one second to accurately read one item’s material composition, which is set according to customers’ desired benchmarks.
