Kemira uses AI to cut years of chemistry research to six months in hunt for a fix to 'forever chemicals' in water
Finnish water treatment company Kemira and British AI startup CuspAI say they have designed new materials capable of removing some of the most persistent pollutants in global water supplies — though the candidates must still prove themselves under real-world conditions.
Text by Martti Asikainen, 25.5.2026 | Photo by Adobe Stock Photos
Kemira and CuspAI have used artificial intelligence to design more than 5,000 novel materials targeting the removal of PFAS — synthetic compounds known as “forever chemicals” — from drinking water.
The Finnish chemicals group and British AI startup announced the results on 21 May, describing a discovery process that compressed what would typically take years into six months.
The programme reached this stage in six months, a process the companies say would typically take years using conventional methods. Approximately 20 priority candidates are now advancing to further development and testing, according to the two companies.
What PFAS are and why they matter
PFAS, short for per- and polyfluoroalkyl substances, are a group of several thousand synthetic chemicals used in industrial and consumer products since the 1940s. They are called “forever chemicals” because they do not break down naturally in the environment and accumulate in the human body over time, where they have been linked to a range of health concerns including certain cancers and immune dysfunction.
Regulators on both sides of the Atlantic have tightened their approach in recent years. The United States Environmental Protection Agency announced maximum contaminant limits for certain PFAS in drinking water, measured in parts per trillion, in 2024. The EU Drinking Water Directive imposes comparable obligations on member states.
The current leading technology for removing PFAS from water is granular activated carbon — a well-established though non-selective filtration material. Kemira, which is already active in the activated carbon regeneration market, initiated the collaboration to explore whether AI-driven design could produce more targeted alternatives.
How the AI approach works
Rather than screening existing known compounds, CuspAI’s platform generated entirely new metal-organic frameworks — nano-porous crystalline materials whose structure and chemistry can be precisely tuned for specific filtration applications — from scratch. The design space for frameworks of this kind runs into the hundreds of trillions, making manual exploration impractical.
The programme delivered material designs with full property data across three specific PFAS molecules — GenX, PFBS, and PFOS — and also uncovered new functional group chemistries the companies say carry potential for broader adsorption product development.
“Kemira challenged us with finding new solutions to one of the most pressing environmental problems of our time, and in six months our partnership delivered,” said Dr Chad Edwards, CEO and Co-Founder of CuspAI. The two companies are already scoping further projects under a framework agreement, though no timeline or commercial targets were provided.
“Our teams worked closely together to ensure that every material candidate was evaluated against real industrial requirements,” said Sampo Lahtinen, Executive Vice President for Research and Innovation at Kemira. “That rigour is what gives us confidence in taking these candidates forward.”
The companies describe the project as the first commercial partnership to apply generative AI end-to-end to design new materials specifically for PFAS remediation — a claim that has not been independently verified.
From discovery to testing
Identifying candidate materials through AI simulation and delivering them as a proven water treatment product are distinct stages separated by significant technical and commercial hurdles.
The priority candidates must now demonstrate performance at sub-parts-per-billion concentrations under real-world conditions, at industrial scale, and at costs competitive with existing technology — questions that the testing phase ahead will need to answer.
CuspAI, founded in 2024, has raised more than $200 million in seed and Series A funding from investors including NVIDIA, Samsung, and Temasek. Kemira employs approximately 4,900 people and is listed on Nasdaq Helsinki.
