AI technologies that might transform how people heat and insulate their homes are some of the 10 clean energy innovations to have reached the final of the £2 Manchester Prize.
The finalists include a non-intrusive system that uses AI to design bespoke panels that turn bricks into radiators that warm homes from the outside in, streamlining the process for older housing stock transitioning to using heat pumps.
Other finalists include AI-enabled heat-mapping drones to help councils and housing associations pinpoint social housing stock in need of insulation upgrades more efficiently, and an AI technology to identify materials in end-of-life solar panels and then recycle them into new photovoltaic cells instead of going to landfill.
The Manchester Prize, funded by the Department of Science, Innovation and Technology and delivered by Challenge Works (part of the Nesta group), “is rewarding UK-led AI breakthroughs that support the public good, including growing the economy, improving public services and helping to create a just transition to Net Zero for everyone.”
Secretary of State for Science, Innovation and Technology, Peter Kyle said:
“AI is opening up transformative new ways to tackle climate change and support the UK’s ambition to become a clean energy superpower.
“That includes using the technology to keep our homes warm, while also supporting projects which will use AI to slash carbon emissions in our cement and steel industries – sectors which account for 16% of global emissions.
“This is how we deliver our Plan for Change – harnessing innovation to solve major challenges, cut energy bills, and improve lives across Britain.”
Energy Secretary Ed Miliband said:
“Clean power is the economic opportunity of the 21st century and these projects will help households and businesses take advantage of lower bills, in a smarter and faster way than ever before.
“From specially designed radiator walls to a smart power grid that flicks on and off as we need, AI has the potential to help every home in Britain to feel the benefits of warmer homes and homegrown clean energy.”
Heating homes from the outside in
The 10 finalists of the Manchester Prize include EnergyWall by Underheat, in partnership with University of Salford. EnergyWall upgrades a building’s walls, gently warming or cooling homes from the outside, turning bricks into radiators that maintain a comfortable internal temperature all year round.
Using AI to analyse a building and off-site manufacturing, it designs and installs pipe systems into insulation panels for the walls of a building, making retrofitting buildings with heat pumps faster, cheaper, and less disruptive. This approach is ideal for social housing, helping reduce carbon emissions, cut energy bills, and tackle condensation that causes mould.
It’s a smarter, scalable way to decarbonise heating and fight fuel poverty across the UK.
Giving solar panels a second life
Green Loops by University of Wolverhampton, in partnership with ABCircular GmbH Berlin, tackles the challenge of recycling end-of-life photovoltaic (PV) cells by creating high-efficiency solar panels from recycled materials. While there are tens of millions solar panels in the UK, the specialist infrastructure to dispose and recycle of them is currently lacking.
Green Loops uses machine learning to analyse the optical properties of materials and structures of solar cells. Using highly conductive artificially engineered MXene-based metamaterials, Green Loops optimises the design of solar cells to enhance energy performance while reducing manufacturing costs.
With the growing e-waste problem from old solar panels, the technology helps reduce waste, supports a circular economy, and makes solar energy more sustainable and accessible.
A “Google Maps” for home heat loss
Rapid Thermal Performance Assessment algorithms (RaThPAs) by Kestrix uses AI and thermal drones to map heat loss across entire neighbourhoods, acting as fast, 3D energy surveys from the sky.
This helps stakeholders like utilities, councils and housing providers plan energy upgrades with fewer costly, time-consuming site visits. Like a “Google Maps of heat loss,” the system shows where buildings are leaking heat and recommends fixes.
With a team of experts in computer vision and physics, Kestrix aims to speed up home retrofits, in turn cutting emissions, saving households money, and making homes warmer and healthier at scale.
Julia King, Baroness Brown of Cambridge, chair of the Manchester Prize judging panel said:
“We are at a critical juncture in the journey to net zero, the next decade is make or break if the world is to keep global temperatures rise to well below 2C by 2050. Global emissions need to halve by 2030 compared to 1990 levels if we are to stay on track, while electricity production will need to double by 2050 to meet the demands of an electrified economy – clean energy innovation is essential. The rapid advancement of AI means we have tools like never before to achieve the goal of decarbonising the economy while supporting individuals, communities and businesses to thrive.”
Other solutions in the running include an AI application to improve the production of biogas where available material inputs can vary from day to day, AI to dramatically cut the CO2 emissions from concrete production (responsible for 8% of global emissions) and from steel production (responsible for a further 8% of global emissions), and wider emissions from industry.
Other finalists include AI technologies to help the logistics industry cut its emissions, and AI to ensure the energy grid remains balanced at all times as more of our energy comes from wind and solar.
The 10 teams behind the advanced AI solutions have each received £100,000 in seed funding, plus compute credits worth £60,000, and additional non-financial support to develop solutions capable of winning the £1 million grand prize in spring 2026. The winning solution will demonstrate not only technical innovation, but also an evidenced road map to near-term (2030) adoption, scale and impact.
In March 2025, the Manchester Prize in its inaugural year awarded £1m to Polaron – an AI technology that dramatically accelerates the development of new advanced materials from years to a matter of days. Increasing the energy capacity of electric vehicle batteries by 10%.
A list of the finalists is included below. Visit manchesterprize.org to find out more about the Manchester Prize and all 10 finalists.
Manchester Prize (Year 2) finalists
Agent Net Zero by University of Sheffield and AMRC. Agent Net Zero is an innovative AI system that helps industrial companies become more sustainable by analyzing their environmental impact in real-time. The system continuously monitors energy usage and emissions by connecting to various data sources across operations. Using advanced AI techniques, Agent Net Zero identifies environmental hotspots and automatically suggests practical improvements. This gives businesses clear, actionable insights to reduce their carbon footprint while maintaining productivity and competitiveness, essentially providing a “sustainability assistant” that works 24/7 to help companies achieve their net-zero goals.
BiofuelAi by University of Surrey. BiofuelAi brings cutting-edge AI and machine learning to the biofuel industry, optimising complex, variable processes in real time. Traditional biogas production often relies on operator intuition due to unpredictable biological systems because biofuels are made from multiple material inputs. BiofuelAi solves this with advanced predictive models that create a digital twin of each site, enabling whole-system optimisation – from daily feedstock recipes to long-term acquisition strategies. Developed by AI and sustainability experts, the platform boosts efficiency, profitability, and environmental impact, offering a scalable solution for cleaner, data-driven energy production worldwide.
Carbon Re by Carbon Re. Cement forms the foundation of our modern world but it has a sustainability problem – it is responsible for around 8% of global CO₂ emissions. Carbon Re is tackling this challenge by building AI process control software to cut emissions in cement production. Acting like self-driving for industrial plants, Carbon Re optimises industrial processes in real-time, helping manufacturers cut both costs and carbon while transitioning to low-carbon operations. A joint spin out of University College London and the University of Cambridge, Carbon Re was founded to deliver immediate climate impact for heavy industry.
Cavolo by Kale AI. Cavolo uses advanced AI to make city deliveries more efficient and eco-friendly. The system helps businesses switch from traditional delivery vans to Light Electric Vehicles (LEVs), which are more efficient in busy cities. By using AI, Cavolo optimises delivery routes in real-time, reducing traffic, energy use, and emissions. The technology helps make urban logistics faster and greener, allowing businesses to deliver goods quickly while saving time and reducing their environmental impact.
Deep.Optimiser-PhyX by Deep.Meta. Deep.Meta is tackling carbon emissions in the steel industry with an AI-powered Digital Twin – a smart digital replica of the production process that combines physics and machine learning to optimise furnace operations. By using real-time sensor data and material science, Deep.Meta more accurately predicts steel slab temperatures and improves scheduling, boosting energy efficiency and significantly cutting emissions. Unlike black-box AI, which can discourage adoption, Deep.Meta’s explainable, physics-based models offer clear reasoning, building trust with users. Founded by experts in metallurgy and machine learning, Deep.Meta is already partnering with global steelmakers and aims to scale through broader industry collaboration.
DRIVE (Deep Re-enforcement learning for Intelligent Vehicle and Energy optimisation) by Flexible Power Systems. Flexible Power Systems (FPS) helps big fleets like vans, trucks, and buses switch to electric by managing vehicles, chargers, and schedules with smart software. FPS uses advanced AI called Deep Reinforcement Learning to solve complex, fast-changing problems – like where and when to charge – more quickly and efficiently. After training in a virtual world, the AI can make smart decisions in real time. First used in EV fleets, this technology could also help with bigger energy challenges in the future.
EnergyWall by Underheat, in partnership with University of Salford. EnergyWall upgrades a building’s walls, gently warming or cooling homes from the outside, turning bricks into radiators that maintain a comfortable internal temperature all year round. Using AI to analyse a building and off-site manufacturing, it designs and installs pipe systems into insulation panels for the walls of a building, making retrofitting buildings with heat pumps faster, cheaper, and less disruptive. This approach is ideal for social housing, helping reduce carbon emissions, cut energy bills, and tackle condensation that causes mould. It’s a smarter, scalable way to decarbonise heating and fight fuel poverty across the UK.
Green Loops by University of Wolverhampton, in partnership with ABCircular GmbH Berlin. Green Loops tackles the challenge of recycling end-of-life photovoltaic (PV) cells by creating high-efficiency solar panels from recycled materials. It uses machine learning to analyse the optical properties of materials and structures of solar cells. Using highly conductive artificially engineered MXene-based metamaterials, Green Loops optimises the design of solar cells to enhance energy performance while reducing manufacturing costs. With the growing e-waste problem from old solar panels, the technology helps reduce waste, supports a circular economy, and makes solar energy more sustainable and accessible.
Grid Stability by University of Manchester. For electricity grids to function, there must be balance between the electricity going into the grid and the electricity leaving it. Grid Stability Monitor uses AI and machine learning to quickly analyse power grid stability as more low-carbon technologies like wind, solar, EVs and heat pumps connect. It replaces slow, complex simulations with rapid, AI-driven assessments, enabling real-time monitoring, faster decision-making, and more confident planning. This helps grid operators maintain reliability while scaling up clean energy solutions and cutting emissions.
Rapid Thermal Performance Assessment algorithms (RaThPAs) by Kestrix. Kestrix uses AI and thermal drones to map heat loss across entire neighbourhoods, acting as fast, 3D energy surveys from the sky. This helps stakeholders like utilities, councils and housing providers plan energy upgrades with fewer costly, time-consuming site visits. Like a “Google Maps of heat loss,” the system shows where buildings are leaking heat and recommends fixes. With a team of experts in computer vision and physics, Kestrix aims to speed up home retrofits, in turn cutting emissions, saving households money, and making homes warmer and healthier at scale.
