From Recycling by Weight to Recycling by Value

From Recycling by Weight to Recycling by Value

The easiest thing to lose in electronic waste is not metal — it is reuse value that no one has identified yet. The EU generates around 5 million tonnes of electrical and electronic waste each year, growing far faster than recycling capacity. Computers sit at the high-value end of that stream: a single machine holds critical raw materials worth EUR 15–45, while a working CPU, RAM module, GPU or power supply can be worth 10–50 times its raw-material content on the secondary market. Today’s recycling lines, however, are engineered for what they do well: safe, high-throughput material recovery. After limited manual sorting, most parts move on by weight or material class. Operators cannot see component condition, live market prices and extraction cost at once — so a CPU with real resale value follows the same route as an obsolete board. The problem is not that the line cannot recycle; it is that one capability is missing: judging, before a part is shredded, whether it is worth more intact than melted down. CIRCULAI, our pilot under ROB4GREEN Challenge #3 (Parts Harvesting), is built to add exactly that capability — and to prove it under real production conditions at SUEZ Feyzin.

Before a component is shredded, can the line ask: is it worth more intact than melted down?

CIRCULAI adds a decision layer to the recycling line: see each component, estimate its value, decide whether to harvest, extract it robotically, and track the outcome. AI vision and the Digital Product Passport identify candidate high-value parts; an AI decision engine weighs resale value against extraction cost before any robot moves. 

CIRCULAI is not about making robots dismantle more — it is about dismantling only what is worth it, giving every robot action an economic reason.

CIRCULAI’s first step is not disassembly — it is understanding the machine in front of it. An RGB-D camera and barcode/QR readers inspect each incoming computer, and the AI Agent cross-references what it sees against Bill-of-Materials data in the Digital Product Passport (DPP). Where legacy equipment has no manufacturer passport, the system builds a synthetic “bridge DPP” from vision and database scans, so today’s unlabelled waste can enter a data-driven process while staying compatible with the EU’s emerging ESPR framework. Seeing a part is half the job; valuing it is the other half. The Value Harvesting Orchestrator (VHO) — CIRCULAI’s value-recovery decision engine — combines secondary-market prices, component condition, and estimated extraction time and cost, and answers the question that matters on a production line: is this part worth extracting? Economically viable components join the selective-harvesting list; everything else continues into conventional material recovery, so no robot time is spent on low-value operations. RePlan, the ROB4GREEN planning block, turns the value-ranked list into an executable dismantling sequence, and collaborative robots (Aubo/Universal Robots, running on ROS 2) carry it out. Every detection, decision, extraction and destination is written back to the DPP, turning the passport from a passive record into a closed data loop — and giving operators an auditable trail of where every part went, ready for coming DPP and ESPR obligations. Validation uses real, mixed-brand computers at SUEZ’s WEEE facility in Feyzin, taking the system from TRL 4 to TRL 7. The pilot targets — to be verified, not yet achieved — include over 90% component identification accuracy, over 5% value-tiered recovery improvement versus the manual baseline, 100% DPP completeness on processed computers, at least 50% combined energy and material savings, and at least 10% GHG reduction.

From processing by weight to a traceable, ROI-based decision on every candidate part — proving on SUEZ’s live line what is worth harvesting, and why.