Recycling robot maker Danu Robotics has spent six years trying to fix one of the least glamorous jobs in the circular economy: pulling the wrong items off a fast-moving conveyor of household recycling. The Edinburgh start-up’s machine, called H.E.R.O., grips items with a pincer claw rather than the suction cups most rivals use, and it relies on AI software that Danu keeps improving after the robot is installed.
The company has raised $5 million in late-seed funding, according to a TechCrunch profile published on 9 October 2026. It reports $500,000 of signed contracts, letters of interest from two large customers and more than 200 potential customers in its sales pipeline. Founder Amy Ma estimates that one robot can add $485,000 a year in revenue to a sorting site, against a $160,000 purchase price and $24,000 a year in maintenance.
This article explains how the H.E.R.O. recycling robot works, who is behind it and what its numbers mean. We work out the payback on Danu’s own figures, compare its funding and speed with better-known rivals, and look at the rules in the UK and EU that could decide how fast sorting sites automate.
Table of contents
- Why the World Needs a Better Recycling Robot
- Who Is Behind the Danu Robotics Recycling Robot
- How the H.E.R.O. Recycling Robot Works
- The Recycling Robot Business Case
- Recycling Robot Speed: Picks per Minute versus Picks per Day
- The Competition: Funding in the Recycling Robot Market
- Rules That Could Push Demand for Every Recycling Robot
- A Portable Recycling Robot: From Sorting Lines to Shopping Centres
- Recycling Robot Risks and Open Questions
- The Bottom Line
- References and Further Reading
Why the World Needs a Better Recycling Robot
Most people who rinse a yoghurt pot never see what happens next. At a materials recovery facility (MRF), mixed recycling arrives on conveyor belts, and much of the final quality control is still done by people standing beside the line.
Hand sorting is still the norm
“Despite being a multi-billion-pound global industry, the majority of waste sorting was still being done manually – dirty, dangerous, and inefficient work with high costs and low output quality,” Ma told the UK’s Startups 100 index, which ranked Danu 66th in January 2026. The work is hazardous, so staff turnover is high and labour costs keep rising.
That is the gap a recycling robot is meant to fill. A recycling robot does not get tired on a night shift, does not need protective breaks from dust and smell, and can log every item it removes. The difficulty is making one that works reliably on a messy, unpredictable stream of real waste.
Contamination costs money
Contamination is the expensive part. Startups 100 reported that contamination rates at sites relying on manual sorting “can be as much as 50%”, which can lead to fines. Contaminated loads mean “recyclables lose value, require costly secondary sorting, or end up in landfill or incineration,” Ma said.
Items that should not be there include plastic bags, foil, coloured paper and food-soiled containers, according to Danu’s partners at the University of Edinburgh. Taking them out, by hand or by recycling robot, is what turns a mixed stream into a bale buyers will pay for.
The scale of the waste problem
Ma frames the opportunity with a global figure: “we generate over three billion tonnes of waste each year, yet less than 10% is recycled.” The World Bank’s latest What a Waste 3.0 report puts municipal solid waste at 2.56 billion tonnes in 2022, rising to 3.86 billion tonnes by 2050 on current trends, so her number appears to use a broader definition of waste.
Either way, the direction is clear. Waste is growing faster than many cities’ systems can handle, and the sorting step in between collection and reprocessing is where automation can raise both volume and quality.
Who Is Behind the Danu Robotics Recycling Robot
Danu is a small company with a long development history. Its founder came to recycling from software rather than from the waste industry.
From a London bank to Edinburgh
Ma, whose full name is Xiaoyan Ma, worked as a software developer at a London bank before founding Danu. Her moment of realisation came when she saw the building’s carefully separated recycling bins being emptied into the general rubbish. “Even though they have tons of money, there’s all these different recycling bins that just end up in the same place,” she told TechCrunch.
She went on to study high-performance computing at the University of Edinburgh’s EPCC supercomputing centre, completing her MSc in 2021. “I saw the technology and thought, we can do so much better,” she said.
Six years in the making
| When | Milestone |
|---|---|
| 2020 | Danu Robotics founded in Edinburgh |
| 2021 | Ma completes her MSc at EPCC, which later helps train Danu’s AI on its Cirrus supercomputer |
| Early trials | Prototype trial at Glasgow City Council’s recycling centre |
| September 2023 | Investment from Sustainable Ventures and a place on its accelerator |
| January 2026 | Ranked 66th in the Startups 100 index |
| October 2026 | $5 million late-seed round; TechCrunch Startup Battlefield 200 |
Along the way Danu went through SOSV’s HAX hardware accelerator and Morgan Stanley’s Inclusive Ventures Lab, and worked with Glasgow and Dundee city councils. Its office is at the National Robotarium at Heriot-Watt University. The company will appear in the Startup Battlefield 200 at TechCrunch Disrupt in San Francisco from 13 to 15 October.
How the H.E.R.O. Recycling Robot Works
H.E.R.O. is Danu’s flagship product: an AI-guided recycling robot for sorting dry mixed recyclables in industrial facilities. Three design choices set it apart.
A recycling robot with a pincer claw instead of suction
The best-known recycling robots, including those from Glacier and Recycleye, use suction arms, TechCrunch reports. A suction cup grabs an item by sealing against its surface and pulling a vacuum. H.E.R.O. instead picks items with a mechanical pincer claw that closes around them.
Each approach has trade-offs. Suction is fast and forgiving of small positioning errors on flat, smooth items. A claw does not need a sealed surface, which can help with crumpled, wet or irregular objects. Danu has not published head-to-head accuracy data against suction rivals, so the advantage is a design argument for now rather than a measured one.
Two arms in under a metre of conveyor
Rather than one large arm, Danu built what Ma calls “a team-based, multi-arm system” that works “more like a group of human pickers – but faster, more precise, and at a fraction of the cost.” Many operators who had tried robots, she says, had decommissioned them “because they weren’t fit for purpose.”
Danu’s website says its two-arm recycling robot needs less than one metre of conveyor space, plus safety guards if it sits next to workers, and can be installed in less than a day. It is designed to retrofit into existing manual sorting lines, so a site does not have to rebuild the plant to try it. High-wear parts such as grippers can be swapped in about 20 minutes.
Computer vision that keeps learning
The recycling robot finds its targets with computer vision. Cameras watch the belt, and machine learning software outlines each object and decides whether it should be removed. Before any picking hardware was built, Danu had to assemble a database of waste images to train the system, and EPCC’s Cirrus supercomputer was used to process the data and train the software.
TechCrunch identifies this as the bigger difference from competitors: Danu uses AI to “continuously improve the robot’s software” after deployment. The company says its AI “brain” will improve as it adds datasets and inputs from new vision sensors.
Counting picks at the bin
Danu also makes a point about measurement. A pick is recorded only “when the object drops into the bin”, not when the arm attempts it. “We believe in measuring genuine Picks Per Day, not sprint based Picks Per Minute,” the company says, arguing that uptime matters more than peak speed.
The Recycling Robot Business Case
Danu’s recycling robot pitch to sorting sites is financial. Recycling centres earn money from the materials they recover, so a machine that pulls out more saleable material pays for itself through higher revenue, not just lower labour costs.
Danu’s own recycling robot numbers
| Figure | Value | Source |
|---|---|---|
| Purchase price per robot | $160,000 | TechCrunch |
| Maintenance per year | $24,000 | TechCrunch |
| Extra revenue per site per year | $485,000 (Ma’s estimate) | TechCrunch |
| Picking speed | 40 picks per minute | University of Edinburgh |
| Trained human pickers | 10 to 20 picks per minute | University of Edinburgh |
| Contamination after sorting | Below 1% | University of Edinburgh |
| Operating cost savings | 30% to 100% | University of Edinburgh |
| Signed contracts | $500,000 | TechCrunch |
| Sales pipeline | 200+ customers | TechCrunch |
The speed, contamination and cost-saving figures describe Danu’s prototype, as reported by the University of Edinburgh in 2024. They are company claims, not independent test results.
Payback on Danu’s figures
Using Danu’s numbers, the payback is very fast. The recycling robot costs $160,000, and the site gains $485,000 a year minus $24,000 maintenance, or $461,000. Dividing the first by the second gives a payback of about 0.35 years, or roughly four months.
Revenue is not profit, though, and the extra revenue depends on commodity prices, belt speed and what the site already recovers. The chart shows how the payback stretches if the real revenue gain is lower than Ma’s estimate.
Months to pay back $160,000, by extra revenue per year (after $24,000 maintenance)
Even at a quarter of Danu’s estimate, the recycling robot would pay for itself in under two years. Over five years at the full estimate, a site would gain $2.425 million in revenue against $280,000 of purchase and maintenance costs. That is why the figure is the core of the pitch, and why buyers will want to test it on their own lines before signing.
Buy the hardware, license the software
Danu’s website sums up its model as “buy what won’t change, license what does.” The robot hardware, built for dry mixed recyclables, is sold. The software and AI are licensed, so customers stay on the latest version as the system learns.
That gives Danu recurring revenue and gives customers a reason to stay. It also means a site’s recycling robot is only as good as the vendor’s continuing support, which matters for a small company competing with far bigger rivals.
Recycling Robot Speed: Picks per Minute versus Picks per Day
Speed is the easiest recycling robot number to compare, and the one Danu itself says to treat with care.
40 picks a minute against 10 to 20
The University of Edinburgh reported that Danu’s prototype worked at 40 picks per minute, against an average of 10 to 20 for trained human operators. That is two to four times human speed. Over an eight-hour shift, 40 picks a minute is 19,200 picks, against 4,800 to 9,600 for a person.
Items picked per eight-hour shift at the stated rates
Bars are scaled to the largest figure, 21,600. Glacier says its robot can sort more than 30 material types at 45 items per minute, according to ESG Today. On raw speed, then, Danu is in the same range as its rivals, not ahead of them.
Why uptime matters more than sprint speed
A recycling robot that picks 40 items a minute for half a shift and then jams is worth less than one that picks 30 all day. That is the logic behind Danu’s preference for “Picks Per Day”, its 20-minute gripper swaps and its rule of counting only items that land in the bin.
For buyers, the useful questions are availability over a month, mean time between stoppages and how the robot copes with the worst material on the line, not the best minute on a demo.
The Competition: Funding in the Recycling Robot Market
Danu is entering a market where better-funded companies have been shipping recycling robots for years. TechCrunch names Glacier and Recycleye as rivals, and the largest player in AI sorting is AMP.
AMP, Glacier and Recycleye
| Company | Base | Latest reported round | Notes |
|---|---|---|---|
| AMP | Louisville, Colorado | $91M Series D, December 2024 | Founded 2014; 400+ AI systems deployed; three full-scale facilities |
| Recycleye | London | $17M Series A, February 2023 | Led by DCVC; AI vision and picking robots for MRFs |
| Glacier | San Francisco | $16M Series A, April 2025 | Founded 2019; backed by Amazon’s Climate Pledge Fund; 45 items per minute |
| Danu Robotics | Edinburgh | $5M late seed, October 2026 | Founded 2020; pincer claw; two-arm retrofit unit |
AMP says its AI has identified 150 billion items and guided the sorting of more than 2.5 million tons of recyclables, according to The Robot Report. Recycleye’s round was led by deep-tech investor DCVC, Recycling Product News reported.
A $5 million round against a $91 million one
Latest reported funding round, US$ millions
AMP’s latest round alone is 18 times Danu’s. Glacier and Recycleye each raised more than three times as much at Series A. Danu’s case is that it can win on cost and fit rather than scale, with a cheaper unit that slots into existing lines.
The size of the prize
TechCrunch puts the market for recycling sorting across Europe and North America at $20 billion. At $160,000 a robot, one unit for each of Danu’s 200 pipeline customers would be $32 million of hardware sales, about 0.16% of that market. The $500,000 already signed is roughly three robots’ worth at list price, although the contracts’ mix of hardware, licences and service is not public.
Those numbers show both the opportunity and the gap. Danu does not need a large share of the market to build a solid business, but turning a pipeline into deliveries is the hard part for any hardware start-up.
Rules That Could Push Demand for Every Recycling Robot
Regulation is one of the strongest arguments for automating sorting. Rules on both sides of the Channel are pushing for cleaner, separated recycling streams.
England’s Simpler Recycling
In England, Simpler Recycling rules changed on 31 March 2025. All businesses, charities and public sector organisations must now separate their waste for recycling, with micro-firms of fewer than 10 full-time equivalent employees given until 31 March 2027. More separated material means more material arriving at sorting sites, and more pressure on its quality.
EU packaging rules from August 2026
In the EU, the Packaging and Packaging Waste Regulation, Regulation (EU) 2025/40, has been in force since 11 February 2025 and has applied since 12 August 2026. It aims to make all packaging on the EU market recyclable in an economically viable way by 2030 and to increase the use of recycled plastics.
Rules like these reward sorting sites that can produce clean, high-value material. That is the output a recycling robot is built to deliver.
UK packaging EPR
The UK’s extended producer responsibility scheme for packaging makes producers pay for managing the packaging they put on the market. As more of the cost of recycling shifts to producers, the efficiency of sorting becomes a bigger concern across the supply chain.
A Portable Recycling Robot: From Sorting Lines to Shopping Centres
With its first machines shipping, Danu’s next goal is a recycling robot that is “sturdier, smaller, and more versatile”, TechCrunch reports.
A smaller, portable recycling robot
“Once it’s smaller, it can be used as a stand-alone recycling solution for events, for shopping centers, hospitals, or airports to recycle their waste at the source,” Ma said, “and that’s the way we can really change how we’re managing our packaging waste.”
Sorting at the source would catch contamination before it reaches a central plant, when separation is cheapest. It would also open a market of venues that produce large volumes of packaging but have no reason to run an industrial sorting line.
What has to be true
A portable recycling robot faces different problems from an industrial one. It must be safe around the public, quiet, easy to empty and cheap enough for a venue’s facilities budget. It would need to cope with a messier stream than a plant’s pre-sorted belt, and someone has to collect and sell what it sorts.
None of that is impossible, and the computer vision and learning software would carry over. But it is a second product with its own engineering and sales challenges, and the main business still has to scale first.
Recycling Robot Risks and Open Questions
Danu’s story is a promising one, and many of its key numbers come from the company. Buyers and investors will want to test them.
Execution risk
Hardware start-ups need cash to build, ship and support machines, and $5 million does not go far in manufacturing. Danu must turn letters of interest and a 200-company pipeline into installed robots while competing with rivals that have raised far more. Reliability in the field, spare parts and service response will decide renewals.
Claims to verify
The $485,000 revenue estimate, the 30% to 100% operating savings and the below-1% contamination figure all need checking against a site’s own data before it buys a recycling robot. A fair trial runs the robot on a buyer’s own line for weeks, measures availability and picks per day, and compares the bale quality with the same line before installation.
Teams thinking about how robots and people share work on a line may find our report on Destro AI’s approach to robots and human workers useful, and our guide to intelligent automation covers how to build a business case before buying. For the waste that AI itself creates, see our coverage of AI data center e-waste.
The Bottom Line
Danu Robotics is a small company with a clear idea: a cheaper, retrofit recycling robot with a pincer claw and software that keeps learning, sold to sites that make money from every clean item they recover. Its own numbers suggest a payback of months rather than years, and even much more cautious assumptions leave the case intact.
The challenge is scale. Danu is up against AMP, Glacier and Recycleye, which have raised far more and shipped far more machines. Whether its better recycling robot wins will depend less on its claw than on how reliably the machines run, day after day, on the dirtiest belts in the industry.
References and Further Reading
Danu Robotics’ fight to build a better recycling robot (TechCrunch)
66. Danu Robotics (Startups 100, 2026)
Xiaoyan Ma, Danu Robotics (Edinburgh Innovations)
Revolutionising recycling with AI (EPCC)
Danu Robotics, Startup Battlefield (TechCrunch)
Burges Salmon advises Danu Robotics on Sustainable Ventures investment
AMP raises $91M (The Robot Report)
Glacier raises $16 million (ESG Today)
Recycleye’s $17 million Series A (Recycling Product News)
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
