Will AI really set a new standard for recyclers?

Artificial intelligence is all around us these days. It generates content, predicts the weather and drives cars, while trials are ongoing for AI to assist in life-saving operations. We’re seeing more than customer service applications and smart homes these days; we’re witnessing smart recycling mega plants.

Worldwide, AI adoption is rapidly growing, with projections suggesting the global AI market will reach EUR 1.5 trillion by 2030 and contribute EUR 13.2 trillion to the global economy by the same year. Recycling is one of many fields benefiting from this breakthrough technology.

Multi-billion gains

AI-enabled systems are likely to become ‘the new standard’ in recycling facilities by 2030, according to researchers at Columbia University. A recent study found modern systems can deliver a 60% increase in efficiency. That means less fuel use, fewer contaminants and more value recovered from the same waste stream.

To put it in perspective: the average worker can sort between 40–50 pieces per minute. Smart mechanical sorters typically achieve double that rate, while a dual unit may reach upwards of 150 picks per minute. Multiply those gains across an entire facility, and the business case for AI becomes hard to ignore.

The global smart sorting market is expected to be EUR 12.5 billion by the end of this year. By 2033, the figure may top EUR 25 billion.

Deep learning sorters

Technology providers know this. Stadler, Steinert and Tomra Sorting Systems have all taken a deep dive into AI. Their aim is simple: separate scrap more effectively and maximise recovery rates. The most advanced systems are powered by artificial neural networks (ANNs) – computational models inspired by the human brain.

These enable ‘deep learning’, meaning machines can adapt to different input materials. Thanks to vast data libraries and countless hours of testing, today’s sorters achieve higher accuracy, greater speed and even sorting tasks in parallel.

That means flexibility for operators. Robots don’t tire, and we can rely on ANNs to remember every lesson they’ve absorbed. If you want PET bottles and trays plucked from a busy line, AI is ready for the job.

It helps to know that AI comes in different flavours. Right now, we recognise three main categories:

• Artificial narrow intelligence (ANI), or Weak AI: these are highly specialised tools such as voice-operated speakers.
• Artificial general intelligence (AGI), or Strong AI: systems that can perform complex tasks like robotic waste sorting lines.
• Artificial super intelligence (ASI): a still-theoretical level of AI that surpasses human intelligence.

For now, recycling lives firmly in the middle camp. AI-driven robotic arms guided by deep learning are transforming sorting. The idea of a fully autonomous, self-improving recycling plant remains the stuff of science fiction.

More AI, more scrap?

And then there’s the elephant in the scrapyard. Algorithms don’t exist in a vacuum. They run on chips and servers, powered by critical minerals with finite lifespans. When they burn out, they add to the already impressive e-scrap waste stream. The faster AI grows, the bigger this pile will become.

Analysts at The Oregon Group have said recently they anticipate a surge in AI technological needs, sparking a 10-year critical mineral ‘supercycle’. This is illustrated by Amazon’s commitment to spending EUR 130 billion on data centres over the next 15 years.

Add to that the energy used to train large AI models, and the water-intensive cooling systems of data centres, and it’s clear AI presents some major sustainability bottlenecks.

One thing is certain: AI is not going away. The test is whether we harness it together to close the loop, instead of widening the circle of waste.

Many possibilities

I believe we can be hopeful. Recycling has always thrived on adaptation. Where others see waste, they see value. Today’s big challenges of recovering end-of-life batteries and black plastics could become tomorrow’s biggest revenue stream. Various test centres I have visited definitely demonstrate remarkable ability.

Algorithms can predict when servers or machine parts fail, so they can be reused before they’re scrapped. Machine learning can cut the energy use of smelters and minimise recycling plant downtime. Designers could easily use AI to create products that are easier to dismantle and recycle.

So will AI be recycling’s greatest ally? The answer won’t come from the machines. It will come from us: how boldly we choose to invest, regulate and reinvent.