AI-Powered Sorting Facilities Are Reshaping the Recycling Industry as AMP's First Fully Automated Plant Nears Operation
AMP's Commerce City facility will process 62,000 tons of recycling annually with minimal human intervention, as AI sorting systems spread across the waste industry.
Overview
The recycling industry is undergoing a shift from labor-intensive manual sorting to AI-driven automation, and a facility nearing completion in Commerce City, Colorado, may represent the clearest example yet. Built by Waste Connections and operated by Louisville-based AMP, the plant is designed to process 62,000 tons of single-stream recycling per year with minimal human intervention, using the company’s AMP ONE sortation system.
The facility, which broke ground in mid-2025 and is expected to open this year, relies on optical-sorting AI software that tracks video from overhead cameras and uses compressed-air jets to launch individual items off conveyor belts into sorted chambers. According to Denverite, the air-based system can process up to 2,000 items per minute, compared to roughly 80 pieces per minute for a robotic arm and far fewer for a human worker.
What We Know
AMP’s partnership with Waste Connections began in 2020 with 24 AI-guided robotic systems and has since expanded to more than 50 installations across the hauler’s network, according to the company’s announcement. The Commerce City plant represents a new phase: rather than retrofitting existing facilities, it was designed from scratch as a largely autonomous operation.
The system continuously tunes itself, identifying jams, monitoring material purity, and adapting to changes in the waste stream without manual intervention. AMP operates under a pay-per-ton agreement with guaranteed performance, a business model that shifts financial risk from the facility owner to the technology provider.
AMP’s existing facility in Ohio already diverts approximately 90 percent of incoming material from landfills, compared to roughly 75 percent at a comparable human-operated plant, as reported by Denverite. Mark Ceresa, a Waste Connections vice president, told the outlet that current manual sorting is “a very low-skill, high-injury job,” framing automation as both an economic and worker-safety improvement.
An Industry in Transition
The Commerce City plant is part of a broader pattern. A February 2026 analysis by Resource Recycling described a new generation of recycling facilities as “cyber-physical factories” that integrate robotics, sensors, and software for real-time monitoring, drawing an analogy to semiconductor manufacturing. Companies including AMP Robotics, ZenRobotics, and TOMRA are deploying machine vision, hyperspectral imaging, and X-ray fluorescence systems that sort metals, plastics, and batteries faster and more accurately than manual crews.
Beyond sorting, new chemical and metallurgical processes are improving what can be recovered. UK-based HyProMag has developed hydrogen processing of magnetic scrap that consumes up to 88 percent less energy than mining and refining rare earth magnets from ore, targeting data center drives and other magnet-rich electronic waste. Another UK firm, DEScycle, uses deep eutectic solvents to extract precious metals with lower energy input and fewer emissions than traditional smelting.
The scale of the challenge is substantial. According to the Resource Recycling analysis, the transition is being driven in part by the sheer volume of electronic waste, which continues to grow faster than collection infrastructure can absorb it, and by the rising value of critical materials such as cobalt, palladium, and rare earth elements embedded in discarded devices.
What We Don’t Know
Whether fully automated facilities like Commerce City can maintain their performance advantage as they scale remains an open question. AMP’s 90 percent diversion rate in Ohio is impressive, but the Commerce City plant is significantly larger and will face a more varied waste stream. The pay-per-ton model also has not been tested widely enough to determine whether it can sustain profitability for the technology provider at scale.
Capital costs remain a barrier. Retrofitting existing plants with AI-guided systems is cheaper than building from scratch, but the patchwork of state recycling laws in the United States means demand for advanced sorting capacity varies sharply by region. Colorado’s recent expansion of recycling mandates and circular economy requirements may create favorable conditions locally, but national adoption of similar standards remains uncertain.
The recycling industry also faces a trust gap. Decades of contaminated recyclables being diverted to landfills or shipped overseas have eroded public confidence in the system. Whether AI-driven facilities can rebuild that trust will depend not just on technical performance but on transparent reporting of actual diversion and recovery rates.