AI and circularity will be key to the future of materials in the AI age

Every major leap in human progress has been catalysed by a breakthrough in materials. Copper unlocked electricity, steel built cities, silicon enabled the computing age, and lithium now powers the global energy transition.

The Energy Transitions Commission has stated that the climate transition is, at its core, a materials transition. Its analysis estimates that achieving net zero by 2050 will require around 6.5 billion tonnes of materials between now and mid-century – a scale-up challenge constrained by time and environmental pressure.

It is increasingly clear that the transition cannot rely on conventional extraction alone. We need both new, more efficient materials and circular systems that keep existing ones in play. And that means rethinking how materials are discovered, produced and reused across all major industrial systems.

Yet materials innovation has historically been slow. Discoveries have usually occurred through serendipity or decades of incremental experimentation. Plastics took more than 50 years to reach ubiquity, while lithium-ion batteries took decades more to mature.

Today, artificial intelligence (AI) offers a structural break with this past. A universe of generative and diffusion computational techniques and architectures that underpin the broader AI wave are being trained to learn the “language of matter”: finding patterns across vast datasets in physics, chemistry and materials science to propose novel materials at a scale previously impossible.

The research community has already leaned in. Estimates based on a Semantic Scholar literature research suggest that in 2014, only 264 materials science papers mentioned AI or machine learning in materials. By 2024, that number had grown to nearly 10,000.

At the same time, Big Tech has moved aggressively, with Alphabet (via DeepMind), Microsoft, Meta, NVIDIA and IBM releasing foundation models for materials and launching large-scale materials programmes. And in the last few years, venture capital activity has accelerated in tandem funding a new cohort of start-ups with hundreds of millions of dollars to attack the space.

The emerging ecosystem spans a taxonomy of innovations and approaches that are seeking to take novel material development through to commercial-scale production. This can be broken down into several categories:

We are certainly early in the wave of a transformational shift, and some expectations may arguably run ahead of commercial maturity. Yet the excitement reflects genuine potential. As futurist Roy Amara put it: "We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”

In the bull case, we could see a 'TechBio moment’ for materials. Just as computational biology is transforming drug discovery, computational materials design could remake the materials industry by turning a process once defined by serendipity into one guided by computation.

Applications span every major sector:

With AI accelerating the discovery pipeline and universe of potential compositions, the question is not whether new breakthroughs are possible but how fast they can move from prediction to production, and from production to circular integration in the real economy.

In the bear case, we could see a production bottleneck. The chemical design space is astronomically large. AI can scan millions of possibilities in hours, but identifying a stable structure is only half the battle. The promise is vast, but the path to real world impact is complex.

Properties must be predicted across scales: from electronic configurations to mechanical strength. For example, thermodynamic stability doesn’t guarantee that material can be synthesized. Real-world production depends on reaction kinetics, pressure and temperature windows, precursor availability and impurity control.

The challenge is not only what to make, but how to make it reliably, safely, economically and at scale.

Material innovation cannot be divorced from planetary constraints. Mining impacts biodiversity, while mismanaged materials such as plastics become pollutants. The real promise of AI lies in pairing discovery with circular design, such as:

The gains will only drive real world impact if these breakthroughs result in a reduction of the footprint of material extraction, not just producing more materials (and waste). Yet there's a structural challenge that technology alone cannot solve.

AI materials discovery must be paired with business model innovation such as materials-as-a-service and performance-based contracts; infrastructure investment in collection, sorting and reprocessing systems; and regulatory evolution simultaneously. Without this alignment, even the most brilliant AI-discovered materials will simply accelerate linear throughput.

An additional caveat, the processing footprint of AI itself also requires a conscious consideration. Artificial intelligence may be ethereal, but its footprint is anything but. Each new model trained, every data request answered, sits atop a mountain of physical capital from servers, batteries, chillers, copper cabling and water for cooling.

Arup highlights how embodied carbon – the emissions already associated with materials and construction processes throughout the construction process – now accounts for roughly 20% of a typical data centre’s total lifecycle emissions, rising to nearly 100% when powered entirely by renewable electricity. Circular materials considerations should be front or centre of the booming AI capital expenditure build-out.

However, when material discovery AI aligns with circular logic, it becomes not only a faster way to innovate, but a smarter way to regenerate ensuring that the molecules powering the next industrial revolution help restore, not deplete, the planet that sustains it.

Given the nascency of the category, we are arguably still in the early phases of cycle, with start-ups being built, research programmes being expanded and incumbents starting to experiment – albeit without huge specificity on end applications or commercial structures.

The first wave of discovery was largely focused on optimizations and improvements to existing known materials. The moment of credibility in biotech came when Big Pharma began to rebuild its discovery engines around computational biology.

For materials, the same inflection point will come when industrial incumbents begin co-developing new catalysts, polymers and alloys with AI-native start-ups. Joint discovery programmes, data-sharing consortia and off-take agreements for AI-designed materials would mark the transition from promise to platform.

So to draw an early conclusion, the age of AI presents a profound opportunity to solve the material-intensive challenge of the climate transition. Yet, for this wave of innovation to achieve genuine planetary impact, the breakthrough promise of AI must be paired with a circular mandate.

The true measure of success will not simply be the discovery of new materials, but their integration into a regenerative system ­– one that is supported by new business models, critical infrastructure and regulatory alignment.

By consciously linking AI's accelerating power with circular logic, we can ensure that the next techno-industrial revolution is powered by materials that work in harmony with the natural system from which they are derived and deliver a new industrial paradigm – building a regenerative not extractive future.