Generative artificial intelligence has been pitched as the dawn of a more streamlined, cost and labour-efficient, and data-driven future for fashion. Through algorithms that use data and machine learning to create content, proponents say it has the potential to reinvent everything from design and forecasting to supply chain management and imagery, allowing for large-scale digitisation of systems that have previously been scattered and incoherent.
In 2023, McKinsey analysts predicted that generative AI could boost the global fashion industry’s profits by up to $275 billion in three to five years, and it is already outpacing its anticipated size, intelligence and potential.
However, as the initial rush of excitement settles, concerns about its vast environmental impact — which critics say have been deliberately obfuscated by leading tech companies — are coming to the fore. From electricity and water-hungry data centres to the natural resources that are required for hardware and e-waste that is generated at the end of its life, fashion brands looking to understand, measure and reduce their overall carbon emissions might find an additional challenge in the adoption of AI.
The buzzy tech has moved beyond its 2024 hype cycle into real-world implementation. We spoke to retailers who are seeing results.
To protect their intellectual property (IP), fashion brands tend to work with small AI startups that specialise in the sector. These AI platforms use the brand’s own data to train models instead of public data sets that are less bespoke and can lead to copyright issues. Despite feeding billions of data points into bespoke generative AI models for brands around the world, most startups are still unable to accurately measure the environmental cost of their technology against its benefits. It is still early days for implementation, they say, and gleaning data insights from brand partners is a barrier to fully quantifying the impact.
Startups also argue that sustainability concerns should be directed at the industry titans who are fuelling AI’s carbon and resource use through ever-increasing demand for hardware and energy-intensive model training, such as OpenAI, Alphabet (parent of Google), Amazon and Microsoft. In 2023, Google’s data centre electricity consumption grew by 17 per cent, contributing to the company’s overall carbon emissions increase of 13 per cent that year and a 48 per cent increase since 2019. Between 2020 and 2023, Microsoft’s emissions grew by 30 per cent. Both companies have attributed this growth to the energy-hungry data centres that are fuelling their large-scale AI platforms.
Whether fashion companies use startups or tech titans, the question remains: can the purported sustainability benefits of generative AI outweigh the environmental impacts of using the tech?
Efficiency first, sustainability second
Fashion’s early forays into the world of AI have included creating artificially generated campaigns, product imagery for e-commerce and social media, and chatbot tools that recommend products and answer customer queries. US retailer Saks Fifth Avenue is creating a more customised shopping experience for its clients with the help of Salesforce’s AI application Agentforce. Since 2020, LVMH has been building its AI Factory to streamline production logistics and predictions, while forecasting agency WGSN launched its Fashion Buying platform in 2024 to assist industry buyers through AI predictive analytics.
Maison Meta, the generative AI studio behind AI Fashion Week, has collaborated with the likes of Norma Kamali, Moncler, Revolve, Pangaia, Tory Burch and more to create AI-generated campaigns, which can allow brands to sidestep carbon-intensive physical photoshoots. But sustainability isn’t the primary motive at play. “The main reason brands come to us is to save on costs and time,” says Cyril Foiret, founder and creative director of Maison Meta. “To be perfectly honest, [sustainability] is not something that is being talked about.”
Today, collections designed by the three winners of the inaugural AI Fashion Week will go live on Revolve. The teams behind the collections break down the creation process for Vogue Business.
Sustainability might not be top of mind for brands, but that doesn’t mean the savings aren’t happening. Look at how generative AI is being used in the design process: to protect their IP and ensure the output is bespoke, brands are using closed AI models (as opposed to open models that tap large, public data sets) that are trained using private data on their style DNA, reference imagery, sketches and mood boards, as well as information about product sell-through and inventory. Generative AI platforms like Raspberry AI, Fashable and Refabric then produce designs based on this data, testing various colours, materials and finishes, creating a tech pack for manufacturers and even product imagery to test trend and consumer demand before ordering and producing garments — all of which should combine to reduce waste.
Refabric, which is part of LVMH’s startup acceleration programme, says its instant digital sampling tech can reduce overproduction in the design phase. “For one order, some brands are creating more than 100 physical samples,” explains Begüm Doğru Öztekin, Refabric co-founder and CEO. “So it’s a very big pain point. We are accelerating the time between concept and collection, and can eliminate the physical sampling process.”
There are other explicit benefits, too. Software company Vaayu, which is adopted by companies including Asket, Veja and Vestiaire Collective, uses AI models to help brands with specific sustainability tasks. Vaayu’s Kria Impact Modelling Engine uses brand production data and machine learning to track and measure a product’s impact across 16 different metrics, identifying risk areas and producing multiple scenarios that will reduce a product’s footprint.
However, there is a risk that AI tools could also be used to accelerate production and promote overconsumption. It comes down to the ethics of the company in question, says Martin Smith, co-founder and CEO of Commonshare, a platform specialising in data collection and management for supply chain traceability, claim verification and decarbonisation. The widespread adoption of AI-powered design and production systems — which can cut lead times from 23 weeks to two weeks — might result in more brands adopting ultra-fast high volume business models.
“It goes back to the mission of the company,” says Smith. “If [a company doesn’t] care at all about sustainability, then they’re probably going to use AI in the same way [as they have been operating to date]. But if it’s a company with the goal to get to net zero, then AI is going to be a tool that allows them to get there faster.”
Generative AI, which can create content in the form of images, videos and text, is a new field for fashion and retail that promises to aid creativity and business processes — but also introduces new considerations.
Tracking generative AI’s environmental impact
Generative AI is powered by AI chips that can be trained to process, remember, learn from and store vast amounts of data. They’re capable of parallel processing, which allows them to perform multiple complex tasks at the same time. For a small agency like Maison Meta, these can be powered through a personal computer, while large companies rely on data centres that hold tens of thousands of AI chips in one location. These data centres are the primary sustainability concern for AI, not just because they’re energy intensive to run; they also require vast amounts of water. A 2023 report by the University of California Riverside and the University of Texas Arlington found that Microsoft’s GPT-3 language model in its US data centres consumed 700,000 litres of water to keep onsite servers cool while they worked 24-7 to train AI models. What’s more, servers tend to have a short lifespan of around three years, leading to significant electronic waste.
In 2023, Digiconomist, a research firm focused on the impact of digital innovations like blockchain and AI, published a report that highlighted the tech’s rapidly increasing carbon footprint. Founder Alex de Vries predicted that by 2027, AI data centres around the world could consume more electricity than a country the size of the Netherlands, Argentina or Sweden; de Vries now believes that AI has outpaced his original estimate. “It’s well on track to beat my predictions in the 2023 paper,” he says. “This is a massive development in an extremely short amount of time, and that’s extremely concerning.”
According to a 2025 survey of 2,000 cross-sector businesses adopting generative AI by consultancy firm Capgemini, 48 per cent of executives knew that its use had contributed to their rising emissions, but only 12 per cent were actually measuring this. Instead of tracking and reducing this impact, Capgemini found that 42 per cent of respondents were in fact looking at adjusting their sustainability commitments to accommodate the expansion of AI.
Fashable, a generative AI platform that works with Lojas Renner, Brazil’s largest apparel retailer, as well as footwear brand Puma, uses Microsoft’s Azure machine learning platform to build bespoke AI models for each client. On Azure, Fashable can track its carbon usage through a dashboard that shares insights into the carbon emissions of its cloud-based generative AI platform. “In the last 12 months, we used 606 kilograms of carbon,” says Fashable founder Orlando Ribas Fernandes. “That is the equivalent of 150 T-shirts [based on fashion design software company Browzwear’s carbon calculator]; but using AI, we created more than 40,000 items. With these insights, we can show brand partners what they are saving, not only in terms of time and money, but also in terms of their footprint.”
The problem arises when AI companies want to measure the exact impact reduction their clients achieve by making a product using AI versus traditional processes. Brands using AI don’t usually share information like sell-through rates on AI-produced garments or carbon emissions reductions from digital sampling, so AI companies are unable to paint an accurate picture of the true impact potential of their technology. The AI companies Vogue Business spoke to were therefore unable to provide concrete data such as Life Cycle Assessments to back up their sustainability claims, but say they’re working with external providers and brand partners to navigate this complex data collection and analysis process.
Digiconomist’s de Vries is sceptical of claims that the benefits of generative AI outweigh the costs. “You can’t properly quantify it on either side. That’s a big issue to start out with, because we’re missing a lot of things, like transparency,” he says.
De Vries believes that brands should rethink jumping on the generative AI bandwagon, and take a more considered approach to adoption. “Start thinking about the problem you’re trying to solve, and then try to figure out the best-fit solution from there, rather than trying to force AI on every little thing you can come up with,” he says. “This typically happens during hype phases, where technology becomes a solution looking for a problem.”
Sign up to receive the Vogue Business newsletter for the latest luxury news and insights, plus exclusive membership discounts.
Comments, questions or feedback? Email us at feedback@voguebusiness.com.
Generative AI hits a fashion acceleration point
