Greener Textiles Through Machine Learning

A European fashion consortium led by Novozymes and supported by H&M Group, BESTSELLER, and academic partners in Copenhagen has used machine learning to optimize enzyme-based cotton dyeing recipes, cutting water consumption by 40 percent in pilot runs. The project, funded partly by the EU Horizon program, represents one of the first production-scale collaborations between fast-fashion retailers and industrial biotech on formulation chemistry.

Traditional cotton dyeing can require thousands of liters of water per kilogram of fabric, along with salts and alkalis that complicate wastewater treatment. Enzyme-assisted processes promise milder conditions, but finding stable formulations across dye classes, fabric weights, and water hardness levels has historically required exhaustive lab trial-and-error.

The ML Pipeline

Researchers trained models on 12,000 historical bath compositions, spectroscopy readings, and colorfastness scores. The system proposes candidate enzyme cocktails and temperature curves that lab technicians validate on small swatches before scaling to pilot drums. Novozymes says the model reduced experimental iterations by more than half compared with manual design-of-experiments approaches.

Color consistency — the nightmare of sustainable dyeing — improved within Delta E thresholds acceptable to retail quality teams in 87 percent of pilot batches. The remaining outliers typically involved reactive dyes on blended fabrics, which the consortium has flagged for a second research phase.

Retailers' Calculus

H&M and BESTSELLER face binding Scope 3 emissions targets and incoming EU corporate sustainability due diligence rules. Water intensity in tier-one and tier-two suppliers is a material disclosure metric. If enzyme dyeing scales, brands could market lower-impact basics without the hand-feel penalties of some early plant-based dyes that consumers rejected in 2022–2023.

Cost remains the gating factor. Enzyme recipes currently add 8–12 percent to per-unit dyeing cost in Denmark pilots. Consortium leaders argue that EU carbon pricing and wastewater fees will narrow the gap by 2028, especially for brands shipping millions of cotton tees annually.

Scaling Barriers

Pilot drums do not equal mill floors in Bangladesh or Turkey, where much of consortium members' volume is dyed. Transferring recipes requires retraining local technicians, recalibrating for different water chemistry, and convincing mill owners to pause lines for experimentation. The consortium is funding three mill partnerships in 2026 with shared savings clauses if water reductions materialize.

Chemical incumbents are not standing still. Archroma and DyStar have launched their own bio-based lines with sales teams already embedded in major mills — meaning the ML consortium must prove performance and price, not just laboratory elegance.

What Success Would Look Like

If the consortium hits its 2027 target of 5 million garments dyed with optimized enzyme baths, it would save an estimated 180 million liters of water — modest globally, but a template for how ML shortens green chemistry R&D cycles. For consumers, the first visible impact may be hang-tag claims on selected H&M Conscious lines rather than a wholesale wardrobe transformation.