The fashion and retail supply chain’s rapid embrace of artificial intelligence (AI) has proved a rather binary existence—a duality that Future Snoops is all too familiar with.
On one hand, the environmental footprint of all-things-AI is mushrooming beyond the pale of the technology’s purported potential to save the world from the brink of burning. On the other hand, machine learning’s mere potential to address the planet’s climate challenge(s) is equally profound. Which is right?
Related Stories
The New York agency’s director of sustainability, Emma Grace Bailey, cited the dilemma as one of the most common (and contemporaneous) concerns heard.
“When someone asks, ‘if AI bad for the environment,’ the honest answer is that it can be,” said Emma Grace Bailey, director of sustainability at Future Snoops. “But it can also be one of the most powerful sustainability tools we have.”
That contradictory concept was the focus of the New York–based futures agency’s December installment of its bimonthly webinar series, “Sustainability: No Filter,” which Bailey described as a deliberately unsanitized look at the industry’s favorite sustainability silver bullets.
“We’ve probably all seen the headlines about data centers using huge amounts of energy and water,” Bailey said during the session. “We’ve heard about training large AI models requiring as much electricity as small towns. As AI adoption explodes, that footprint is only growing faster—and that part of the truth is real.”
The shortage market is outstripping sustainable energy’s capacity. Such demand for this disequilibrium dynamic has even led to a renewed reliance on fossil fuels, Bailey said, citing Noman Bashir, a climate impact fellow at the Massachusetts Institute of Technology.
“The demand for new data centers cannot be met in a sustainable way. The pace at which companies are building new data centers means the bulk of the electricity to power them must come from fossil fuels,” she said, quoting a Jan. 17 article by MIT News.
But the other part—what Bailey said is usually the part that gets left out of the conversation—is that AI is also helping.
“It’s optimizing supply chains, reducing waste, helping brands choose better materials and predicting demand, so companies don’t over produce—it’s even spotting climate risks before they cause major disruptions,” she said. Over 60 percent of global carbon emissions originate from supply chains, Bailey said while discussing AI’s “unparalleled ability” to gather and analyze data.
With that perspective in mind, Bailey said the ultimate environmental impact of AI is not predetermined; rather, it depends entirely on how we build it and where we apply it—and the choices we make along the way.
As for how generative AI actually impacts the physical planet, let’s start with where the machines behind “chat” are found.
Data centers—the physical backbone of the virtual revolution—require absurd amounts (as in eight times more than the typical computing workload) of electricity to both train AI models and operate them. A Goldman Sachs calculation that data center energy consumption will grow exponentially—accelerating by 175 percent by 2030—contextualized the massive strain the AI boom is placing on global resources.
The process of training a single large AI model alone can consume 1,287 megawatt-hours of electricity—enough to power 120 average U.S. homes for a year and generate approximately 552 tons of carbon dioxide, according to a 2021 study by scientists from Google and the University of California at Berkeley.
The ongoing use of a trained model to generate “answers” (text, images, etc.) is known as inference. This process—which boils down to how AI models analyze new data and produce outputs—is now responsible for 80-90 percent of all AI computing resource use; the International Energy Agency (IEA) estimated it to be 10 times more energy-intensive than the average Google search.
The 2024 global share of data centers accounted for approximately 1.5 percent of the world’s total energy consumption, according to the IEA. The U.S. itself accounts for 45 percent of the world’s data center energy use; AI-specific needs account for 4.4 percent of the country’s total electricity consumption.
Looking ahead, the IEA determined that AI data centers could account for 35 percent of Ireland’s total energy use by the end of 2026. It’s not that far-fetched; task-specific models (as opposed to general-purpose, like large language models) require energy that is split relatively evenly between data processing, model training and inference. Generating one AI image requires the energy equivalent of a full smartphone charge, MIT reported; 1,000 text prompts require 16 percent of a charge.
By the end of next year, data center electricity consumption is expected to approach 1,500 terawatt-hours—solidifying their place as the fifth most energy-intensive “country” in the world, ranking between Japan and Russia, though the projection is based on conservative estimates.
The pressure extends beyond electricity. A United Nations report warned that AI-related infrastructure could soon consume six times as much water as Denmark—a country of six million people. It’s a particularly alarming statistic, considering that a quarter of the global population still lacks reliable access to clean water. And that data centers often concentrate this water demand in specific, localized areas.
With that in mind, the model’s rapid obsolescence is all the obscener.
“We’re burning enormous amounts of energy on models with very short shelf lives,” Bailey said. “That’s part of what makes this moment so urgent.”
Despite these enormous environmental costs, Future Snoops presented AI as equally possessing an extraordinary capacity to solve complex sustainability challenges. Plus, 81 percent of executives are already using AI to advance their sustainability goals, according to a 2025 Deloitte report.
Bailey argued that focusing only on AI’s footprint obscures where the technology is already being used to reduce waste and emissions—particularly upstream, before products reach consumers.
“AI is helping brands choose better materials, reduce overproduction and optimize supply chains,” she said. “It’s also spotting climate risks before they cause major disruptions.”
In product design, AI-powered eco-design tools allow designers to model the carbon impact of materials, dyes and processes in real time. Platforms such as Fairly Made’s Ecodesign software give fashion teams fast feedback on how design decisions can affect a product’s overall performance, per Bailey. Virtual sampling is another area Future Snoops sees AI cutting waste. At the point of sale, AI is being used to address returns—nearly half of which never re-enter the market, according to Bailey. Forty-four percent of returned products “are never used again,” she said, “a huge driver of that is poor fit.”
Nike Fit uses AI and augmented reality to scan customers’ feet and provide model-specific sizing recommendations, while Levi’s is rolling out AI-powered outfitting tools that help shoppers visualize complete looks before purchasing—efforts aimed at ensuring customers buy items they’re more likely to keep.
AI is also beginning to unclog bottlenecks in resale and recycling. Patagonia’s partnership with resale platform Trove integrates secondhand inventory directly into its primary e-commerce site, placing pre-owned goods alongside new products.
Future Snoops presented these findings as evidence of AI’s greatest sustainability potential in supply chains, where more than 60 percent of global emissions originate and visibility remains limited.
“AI can map emissions across materials, suppliers and transport routes in a way humans simply can’t,” she said.
Reckitt, the consumer goods company behind brands like Lysol and Durex, used generative AI to calculate emissions across its entire product portfolio (roughly 25,000 items) in under four months, improving the resolution of its emissions data by a factor of 75. DHL uses AI to optimize last-mile delivery routes, improving fuel efficiency and reducing idling. At the same time, Ikea’s demand-sensing tools analyze roughly 200 data points per product (including weather patterns and local events) to reduce overproduction.
“This isn’t about perfection,” Bailey said. “It’s about having the insight to act at the right time, instead of reacting after the damage is already done.”
The takeaway, she emphasized, is not that AI is inherently sustainable—but that its impact is still being decided.
“The future isn’t predetermined,” Bailey said. “If we’re intentional—if we design responsibly and apply AI where it actually reduces waste—it can become one of the most powerful sustainability tools of the next decade.”
