When robots recycle: How AI and automation are redefining post-use recycling
Every year, millions of tons of electronic waste are produced worldwide, most of which is only recycled in a rudimentary way. According to Global E-waste Monitor 2024, the figure was around 62 million tonnes in 2022, of which only around 22 percent were documented as having been properly collected and recycled. Conventional shredding and sorting processes still dominate practice, but have their limits when it comes to high-quality recovery and genuine recyclability. Overall, the process is cost-intensive and prone to error. But this is exactly where opportunity arises: With AI and robotics, post-use recycling can become industrial reality. Current research projects and prototypes show how that can be achieved.
From manual to intelligent disassembly
What used to be laborious manual work in traditional recycling companies will in future be left to adaptive robots: dismantling complex electronic devices. At the end of a product’s life, a machine automatically recognizes circuit boards, housings, cables and connectors, and plans a suitable disassembly strategy to separate assemblies according to type.
The concept is no longer pure theory: the Fraunhofer project iDEAR (Intelligent Disassembly of Electronics for Remanufacturing and Recycling) is developing an AI-supported robot system that analyzes and non-destructively disassembles electronic modules. In February 2025, the team presented a demonstrator that, among other things, automatically removes the motherboard of a PC from its housing. 3D optical sensor technology, spectral data, AI recognition of components and connectors, and a digital disassembly twin serve as the basis.
Challenges in recycling e-waste
The complexity of electronic products places high demands on automated systems. A smartphone differs from a router not only in design but also in the material mix and fastening techniques. While previous recycling processes fail because of this diversity, artificial intelligence enables the decisive progress.
Models recognize components and connecting elements such as screws, clips or adhesives, assess their condition and translate them into robot actions. Multisensor technology is central to this: 3D cameras and spectral sensors record geometry, surfaces and material properties. This data flows into the digital twin, which serves as the basis for creating the optimum disassembly sequence. Systems like these can use self-learning. The database grows with each disassembled device, allowing robots to continuously adapt to new product types. The manual disassembly process thus becomes a reproducible, data-driven process.
Global projects and initiatives
It is not only the Fraunhofer Institute that is driving forward post-use recycling. Various projects and applications are emerging worldwide at the same time:
- ReconCycle (EU project): The project developed a modular, AI-supported robot cell that uses soft robotics and computer vision to automatically dismantle electronic waste. A particularly innovative feature is that the system automatically adapts to different device types.
- Recirculate (Finland): The project is developing a system to dismantle EV batteries in two phases from pack to module, and module to cell. The robot in the system identifies screws, analyzes wire orientations, and uses machine learning models to make decisions.
- Carnegie Mellon University (USA): Researchers at Carnegie Mellon University in the US have developed a robot that can disassemble a flat-screen television.
- AWS and Molg (USA): AI robots from the startup Molg are being used at AWS Reverse Logistics in Pennsylvania to test the targeted disassembly of decommissioned data center hardware in order to increase reuse and recycling rates.
- Apple (USA): The recycling robot Daisy can dismantle and recycle used iPhones.
All of these approaches point in the same direction: away from mass shredding and toward selective, intelligent post-use processes. Instead of shredding entire devices, valuable modules are selectively extracted. This is a key prerequisite for high-quality secondary raw materials.
Economic and regulatory implications
Automated disassembly is capital-intensive, but pays off in the long term thanks to higher reuse and recovery rates, and more stable quality. Although real-world examples are already reporting positive effects, reliable cost-benefit data is still rare.
The Ecodesign for Sustainable Products Regulation (ESPR) set the regulatory framework in July 2024: Future regulations may define requirements for individual product groups in terms of repairability, disassembly, recyclability and the Digital Product Passport. From February 2027, the EU Battery Regulation will require portable batteries to be removable and replaceable, making it a driver for disassembly-friendly design, especially in consumer electronics.
Anyone designing devices today should consider the accessibility of connecting elements, material combinations and data transparency from the outset. A combination of product design that takes recycling into account, AI-supported robotics and recycling logistics could be an important key in the circular economy.
Hurdles on the path to industrial implementation
Before such systems can become the industrial standard, they still face a number of challenges. The heterogeneity of old devices remains a key problem. Many devices are glued, welded or contain mixed materials that are difficult to separate automatically. In economic terms, precision robotics requires high initial investment and sufficient volume to leverage economies of scale. In addition, consistent data standards are lacking for parts lists, connection types and material data, which form the basis for AI models and digital disassembly twins to function on a broad scale.
However, AI and robotics make post-use recycling predictable, precise and economical in the long term. Projects such as Fraunhofer’s iDEAR show that non-destructive disassembly is no longer a promise for the future, but a technological reality.