How AI Helps Recyclers Turn Mixed Waste Into Reusable Material

As world leaders gather for COP31 in Türkiye, the event's zero-waste emphasis puts a concrete challenge in focus: extending the life of products and materials. Four members of NVIDIA's Inception startup program are using vision AI and robotics to pick valuable material off conveyor belts of used clothing, bottles and mixed recyclables, diverting thousands of tons from landfills.

The scale of the problem is large. Less than 7% of materials entering the global economy come from recycled or reused sources, while municipal solid waste generation exceeds 2 billion tons per year — a figure that could nearly double by 2050. Recycled plastic struggles to compete with virgin material on cost and quality: incoming waste arrives inconsistent, sorting leans on hard-to-find manual labor, and mixed scrap often gets downcycled into low-grade products.

AIRECO, a Taiwan-based startup, argues that AI-driven sorting can lower costs and lift quality enough to make recycled material commercially useful. Its vision AI sorts plastics more finely, pulling specific materials out of mixed scrap piles for closed-loop recycling — reprocessing used items into new versions of the same product. "The finer you sort, the more you can recycle," said Chun-Feng Liu, founder and CEO of AIRECO, adding that finely sorted material is likelier to stay in the loop, a bottle recycled back into a bottle rather than into harder-to-recycle products. The company runs inference on NVIDIA Jetson Orin NX modules across a pilot line at a recycling plant in Tainan, including driving robotic arms that pick targets off a conveyor belt, with recognition accuracy above 95% on the test line.

For its full production line planned for next year, AIRECO aims to lift annual waste processing capacity from 18,000 to 48,000 tons while cutting operating costs by 85%. At material recovery facilities, which separate household recyclables, sorting must happen in real time as belts move fast — demanding rapid analysis to identify and separate items. EverestLabs AI, a U.S. startup, deploys computer vision-guided robots on those lines, with about 150 robots currently in facilities across North America and Australia.

NVIDIA frames the effort as infrastructure for a more circular economy. "A more circular economy relies on making better use of what we already have," said Josh Parker, head of sustainability at NVIDIA, describing AI as a way to understand waste streams, identify what can be recovered and automate sorting. What to watch: whether the pilot-to-production leap delivers the promised capacity and cost figures, whether accuracy holds as mixed waste becomes more varied, and how far these systems can scale beyond today's deployed fleets.

The open questions builders should track are whether fine-grained sorting economics improve throughput enough to compete with virgin material at scale, and whether robot deployments expand beyond the current ~150 units to meaningfully shift recovery rates.