New AI models that can enable the optimisation of concrete mixes through identifying alternative materials and by calculating optimal water and binder ratios are demonstrating their real-world applicability, with huge potential benefits for the efficiency and sustainability of construction.
Concrete is an indispensable construction material – particularly for large structures, heavy industrial buildings, and various types of infrastructure. To significantly reduce the industry’s carbon footprint, it is essential that concrete be decarbonised.
AI has the potential to redefine concrete technology, enabling the development of high-performance mixes that comply with environmental standards.
In May, a research team from the Massachusetts Institute of Technology (MIT) led by postdoctoral researcher Soroush Mahjoubi published a paper in Nature’s Communications Materials, outlining a machine-learning framework it developed using large language models that evaluated and sorted candidate concrete materials based on their physical and chemical properties.
Mahjoubi said there was so much data out there on potential materials – hundreds of thousands of pages of scientific literature. He remarked that sorting through this information would take many lifetimes, during which time even more materials would likely be discovered.
Describing the sought-after physical and chemical properties of concrete, Mahjoubi said: “First, there is hydraulic reactivity.
“The reason that concrete is strong is that cement – the ‘glue’ that holds it together – hardens when exposed to water.
“So, if we replace this glue, we need to make sure the substitute reacts similarly.
“Second, there is pozzolanicity – this is when a material reacts with calcium hydroxide, a byproduct created when cement meets water, to make the concrete harder and stronger over time.
“We need to balance the hydraulic and pozzolanic materials in the mix so the concrete performs at its best.”
Calcium-based materials are incredibly versatile and can form hybrid mixtures such as those containing a silica variation.
They also exhibit enhanced durability under extreme conditions, serve as an alternative binder for 3D printing, continuously utilise novel waste products, can be modified to possess self-healing properties, and are functional in applications involving engineered cementitious composites.
Using the framework, the researchers analysed scientific literature as well as more than a million rock samples, sorting selected candidate materials into 19 types, including biomass, mining byproducts, and demolished construction materials.
Notably, they discovered that suitable materials were available globally, and even more significantly, many could be incorporated into concrete mixes merely by grinding them. This suggests the potential for reducing emissions and cost savings without extensive additional processing. Mahjoubi highlighted that some of the most intriguing materials capable of partially replacing cement were ceramics.
He said: “Old tiles, bricks, pottery – all these materials may have high reactivity.
“That’s something we’ve observed in ancient Roman concrete, where ceramics were added to help waterproof structures.”
Earlier this year, Meta, the parent company of Facebook, developed an open-source AI tool to design concrete mixes that are stronger, more sustainable, and quicker to prepare than conventional concretes. This innovation accelerates construction while minimising environmental impact.
The AI tool utilises Bayesian optimisation – powered by Meta’s BoTorch and Ax frameworks – and was developed in collaboration with a US-based Swiss building materials company and recent Holcim spin-off Amrize and the University of Illinois Urbana-Champaign.
The multi-objective Bayesian optimisation algorithms enable the model to learn and optimise concrete compositions and predict compressive strength curves for different mixtures, optimising short- and long-term strength properties and sustainability.
Meta noted that concrete was typically optimised for strength and cost, but modern build – including data centres – required concrete that was optimised for sustainability, curing speed, workability, and finishability.
Meta observed that concrete is usually optimised for strength and cost; however, modern builds – including data centres – demand concrete that is optimised for sustainability, curing speed, workability, and finishability.
It said: “Innovation in concrete formulations is difficult and slow – compared to traditional concrete, current formulas for low-carbon concrete face several challenges [including] slower curing speeds, issues with surface quality, and complications in supply chains when novel materials are involved.
“But concrete suppliers can utilise AI to develop and scale innovative concrete mixes as drop-in replacements, accelerating the discovery and integration of sustainable materials for large-scale use.”
Meta said that its collaborative partners enabled the development of the AI model and pipeline to accelerate the discovery of high-performance, low-carbon concrete mixtures that met traditional requirements alongside newer sustainability needs.
It added: “Our AI algorithms incorporate specific water-to-binder ratios and volumetric material constraints, and discover high-performing formulas with faster curing and lower global warming potential values that meet the stricter requirements.
“Within two iterations, and with minor human adjustments, the AI pipeline discovered formulas that exceeded standard
low-carbon industry formulas in terms of strength, speed, and sustainability.”
After determining a specific mix, its performance was tested in a real-world application at Meta’s Rosemount data centre by Amrize and its partner, Mortensen, the general contractor responsible for the construction of the facility.
Formal tests demonstrated that the concrete mix surpassed all technical requirements while meeting the necessary credentials for the application.
Closer to home, a research project undertaken by the Queensland University of Technology and National Precast member Everhard Industries over the last two years focused on developing 3D-printed concrete that adopted lower carbon materials such as sugarcane bagasse ash and fly ash.
AI played three major roles in the study: in predictive design, automated quality control, and intelligent performance prediction.
Everhard Industries Dr Mohammad Kangavar said combining 3D printing with AI gave precast manufacturers several practical advantages.
It said: “It enables faster prototyping without the need for mould fabrication, while the use of sugarcane bagasse ash and other alternative binders helps lower the overall carbon footprint.
“Print consistency improves through automated monitoring systems, and testing costs fall because predictive design models reduce the number of physical trials required.
“AI and 3D printing together also support more efficient use of materials, which in turn reduces waste and energy demand.” Despite the promise, adoption is not yet straightforward.
Australia currently has no dedicated standards, design guidelines or durability frameworks for 3D printed civil structures.
Approval authorities, councils and transport bodies remain cautious, particularly for underground structures where long-term performance must be proven.
Dr Kangavar said the results of the project are promising, but regulatory confidence will ultimately determine how quickly it becomes mainstream.
“By reducing testing times, improving quality assurance and supporting more sustainable concrete development, AI is laying the groundwork for a new generation of smart, efficient and lower-carbon precast solutions.”
