Humanoid Robots Are Hard to Recycle and Carry Rare-Earth Magnets

TL;DR

A September 2026 trade-press article lays out, in detail, what it takes to disassemble and recycle a humanoid robot. Every specific detail in that account — how many parts a robot has, how much hazardous magnet material it contains, what makes its batteries and stored data risky to dispose of — comes from one person: consultant Robert Belt, who says he currently works with recycling firm Re-Teck LLC. Re-Teck would stand to gain business if humanoid robot recycling becomes seen as a specialised, necessary service, so this is not a neutral account. A peer-reviewed academic paper exists on the same subject, written by researchers with no apparent tie to Re-Teck, but it could not be accessed to check whether it supports or contradicts Belt’s figures. No robot manufacturer, safety regulator or standards body has published its own account of what decommissioning a humanoid robot involves.

What happened

In September 2026, The Robot Report published an article by Robert Belt describing what happens when a humanoid robot breaks down and has to be taken apart. Belt writes that a single robot contains 10,000 to 15,000 individual components spread across roughly 200 to 500 major sub-components, split among four systems: motors and gearing that let it move (20-40 electric motors with precision speed reducers), a metal or carbon-fibre frame (30-50 major structural elements held together by 1,000-3,000 fasteners), a network of position sensors and cameras (40-80 position encoders and 50-200 sensors including lidar, IMUs and cameras), and up to 80 memory and storage chips.

Belt also states each robot contains 3.5 to 4 kilograms of neodymium magnets — neodymium is a rare-earth metal used to make powerful magnets — which he says is more than is found in the magnets of an entire electric-vehicle skateboard chassis, the flat platform that carries an EV’s battery pack and drivetrain. Neither figure is checked against a manufacturer specification or an independent measurement in the dossier; it is Belt’s account alone.

The article further states that a decommissioned robot’s lithium battery packs can catch fire, release toxic gas or explode if mishandled, and that onboard memory can retain navigation maps, biometric data and facial-recognition recordings that create a data-security risk unless wiped before disposal. Belt is quoted describing the process itself: “Recycling a humanoid is less like scrapping a car and more like surgery – and right now it is still done by hand.”

Belt proposes what he calls design-for-recycling principles — standardised, quick-release parts that would let robots be pulled apart faster than by hand — as a way to make recycling scalable. The article does not name any manufacturer that has adopted or committed to that approach.

All of this rests on one named technical source. The article’s own biography line states that Belt currently works with Re-Teck LLC, a commercial recycling and reverse-supply-chain firm, to recycle and repurpose humanoid robots globally.

What this means (and what it does not)

Right now, the public conversation about humanoid robot recycling hazards rests on one voice, and that voice is not neutral: Belt’s consulting work is tied to Re-Teck, a company that would benefit from robot teardown being seen as complex and specialist-dependent. That does not mean his figures are wrong; nothing in the available record disproves them.

It also does not mean the underlying topic is invented. A peer-reviewed paper on humanoid robot end-of-life waste, titled “The Era of Humanoid Robots: Addressing Emerging End-of-Life Waste Challenges,” was published in March 2026 in the American Chemical Society journal Environmental Science & Technology by Ziyu Wang, Zhi Chen, Sarah Sajedi, Sixu Deng and Chunjiang An of Concordia University and McMaster University — researchers with no apparent connection to Re-Teck. Its existence shows academic and regulatory interest in this subject is real and predates any recycling-industry pitch. But its actual findings could not be read for this reporting: attempts to reach the publisher’s page and the PubMed listing were blocked. So it is unknown whether that paper supports, contradicts, or simply differs in detail from the component counts, magnet weight or hazard categories attributed to Belt.

What none of this establishes is that humanoid robots have a confirmed, measured hazard profile independent of one consultant’s account, or that his proposed fix has been evaluated by anyone who builds the robots.

What we still do not know

• The Wang et al. paper’s substantive findings remain unread — only its title, authors and publication record were confirmed — so whether it corroborates or contradicts Belt’s specific figures is unknown.
• No humanoid robot manufacturer — Tesla, Figure, Boston Dynamics, Agility Robotics, Unitree and 1X were checked — has published any end-of-life or decommissioning policy.
• No regulator or standards body has published a decommissioning standard specific to humanoid robots.
• No source establishes what decommissioning a robot costs, how long it takes, or how that compares with repairing one instead.
• The exact nature of Belt’s relationship with Re-Teck — employee, contractor, equity holder or informal collaborator — is not specified anywhere in the available record.
• Whether any humanoid robots have actually reached end-of-life and been decommissioned yet, or whether this remains a hypothetical future problem, is not stated.