This story originally appeared in Grist and is part of the Climate Desk collaboration.

In 2023, the fast-fashion giant Shein was everywhere. Crisscrossing the globe, airplanes ferried small packages of its ultra-cheap clothing from thousands of suppliers to tens of millions of customer mailboxes in 150 countries. Influencers’ “#sheinhaul” videos advertised the company’s trendy styles on social media, garnering billions of views.

At every step, data was created, collected, and analyzed. To manage all this information, the fast fashion industry has begun embracing emerging AI technologies. Shein uses proprietary machine-learning applications — essentially, pattern-identification algorithms — to measure customer preferences in real time and predict demand, which it then services with an ultra-fast supply chain.

As AI makes the business of churning out affordable, on-trend clothing faster than ever, Shein is among the brands under increasing pressure to become more sustainable, too. The company has pledged to reduce its carbon dioxide emissions by 25 percent by 2030 and achieve net-zero emissions no later than 2050.

But climate advocates and researchers say the company’s lightning-fast manufacturing practices and online-only business model are inherently emissions-heavy — and that the use of AI software to catalyze these operations could be cranking up its emissions. Those concerns were amplified by Shein’s third annual sustainability report, released late last month, which showed the company nearly doubled its carbon dioxide emissions between 2022 and 2023.

“AI enables fast fashion to become the ultra-fast fashion industry, Shein and Temu being the fore-leaders of this,” said Sage Lenier, the executive director of Sustainable and Just Future, a climate nonprofit. “They quite literally could not exist without AI.” (Temu is a rapidly rising ecommerce titan, with a marketplace of goods that rival Shein’s in variety, price, and sales.)

In the 12 years since Shein was founded, it has become known for its uniquely prolific manufacturing, which reportedly generated over $30 billion of revenue for the company in 2023. Although estimates vary, a new Shein design may take as little as 10 days to become a garment, and up to 10,000 items are added to the site each day. The company reportedly offers as many as 600,000 items for sale at any given time with an average price tag of roughly $10. (Shein declined to confirm or deny these reported numbers.) One market analysis found that 44 percent of Gen Zers in the United States buy at least one item from Shein every month.

That scale translates into massive environmental impacts. According to the company’s sustainability report, Shein emitted 16.7 million total metric tons of carbon dioxide in 2023 — more than what four coal power plants spew out in a year. The company has also come under fire for textile waste, high levels of microplastic pollution, and exploitative labor practices. According to the report, polyester — a synthetic textile known for shedding microplastics into the environment — makes up 76 percent of its total fabrics, and only 6 percent of that polyester is recycled.

And a recent investigation found that factory workers at Shein suppliers regularly work 75-hour weeks, over a year after the company pledged to improve working conditions within its supply chain. Although Shein’s sustainability report indicates that labor conditions are improving, it also shows that in third-party audits of over 3,000 suppliers and subcontractors, 71 percent received a score of C or lower on the company’s grade scale of A to E — mediocre at best.

Machine learning plays an important role in Shein’s business model. Although Peter Pernot-Day, Shein’s head of global strategy and corporate affairs, told Business Insider last August that AI was not central to its operations, he indicated otherwise during a presentation at a retail conference at the beginning of this year.

“We are using machine-learning technologies to accurately predict demand in a way that we think is cutting edge,” he said. Pernot-Day told the audience that all of Shein’s 5,400 suppliers have access to an AI software platform that gives them updates on customer preferences, and they change what they’re producing to match it in real time.

“This means we can produce very few copies of each garment,” he said. “It means we waste very little and have very little inventory waste.” On average, the company says it stocks between 100 to 200 copies of each item — a stark contrast with more conventional fast-fashion brands, which typically produce thousands of each item per season, and try to anticipate trends months in advance. Shein calls its model “on-demand,” while a technology analyst who spoke to Vox in 2021 called it “real-time” retail.

At the conference, Pernot-Day also indicated that the technology helps the company pick up on “micro trends” that customers want to wear. “We can detect that, and we can act on that in a way that I think we’ve really pioneered,” he said. A designer who filed a recent class action lawsuit in a New York District Court alleges that the company’s AI market analysis tools are used in an “industrial-scale scheme of systematic, digital copyright infringement of the work of small designers and artists,” that scrapes designs off the internet and sends them directly to factories for production.

In an emailed statement to Grist, a Shein spokesperson reiterated Peter Pernot-Day’s assertion that technology allows the company to reduce waste and increase efficiency and suggested that the company’s increased emissions in 2023 were attributable to booming business. “We do not see growth as antithetical to sustainability,” the spokesperson said.

An analysis of Shein’s sustainability report by the Business of Fashion, a trade publication, found that last year, the company’s emissions rose at almost double the rate of its revenue — making Shein the highest-emitting company in the fashion industry. By comparison, Zara’s emissions rose half as much as its revenue. For other industry titans, such as H&M and Nike, sales grew while emissions fell from the year before.

Shein’s emissions are especially high because of its reliance on air shipping, said Sheng Lu, a professor of fashion and apparel studies at the University of Delaware. “AI has wide applications in the fashion industry. It’s not necessarily that AI is bad,” Lu said. “The problem is the essence of Shein’s particular business model.”

Other major brands ship items overseas in bulk, prefer ocean shipping for its lower cost, and have suppliers and warehouses in a large number of countries, which cuts down on the distances that items need to travel to consumers.

According to the company’s sustainability report, 38 percent of Shein’s climate footprint comes from transportation between its facilities and to customers, and another 61 percent come from other parts of its supply chain. Although the company is based in Singapore and has suppliers in a handful of countries, the majority of its garments are produced in China and are mailed out by air in individually addressed packages to customers. In July, the company sent about 900,000 of these to the US every day.

Shein’s spokesperson told Grist that the company is developing a decarbonization road map to address the footprint of its supply chain. Recently, the company has increased the amount of inventory it stores in US warehouses, allowing it to offer American customers quicker delivery times, and increased its use of cargo ships, which are more carbon-efficient than cargo planes.

“Controlling the carbon emissions in the fashion industry is a really complex process,” Lu said, adding that many brands use AI to make their operations more efficient. “It really depends on how you use AI.”

There is research that indicates using certain AI technologies could help companies become more sustainable. “It’s the missing piece,” said Shahriar Akter, an associate dean of business and law at the University of Wollongong in Australia. In May, Akter and his colleagues published a study finding that when fast-fashion suppliers used AI data management software to comply with big brands’ sustainability goals, those companies were more profitable and emitted less. A key use of this technology, Atker says, is to closely monitor environmental impacts, such as pollution and emissions. “This kind of tracking was not available before AI-based tools,” he said.

Shein told Grist it does not use machine-learning data management software to track emissions, which is one of the uses of AI included in Akter’s study. But the company’s much-touted usage of machine-learning software to predict demand and reduce waste is another of the uses of AI included in the research.

Regardless, the company has a long way to go before meeting its goals. Grist calculated that the emissions Shein reportedly saved in 2023 — with measures such as providing its suppliers with solar panels and opting for ocean shipping — amounted to about 3 percent of the company’s total carbon emissions for the year.

Lenier, from Sustainable and Just Future, believes there is no ethical use of AI in the fast-fashion industry. She said that the largely unregulated technology allows brands to intensify their harmful impacts on workers and the environment. “The folks who work in fast-fashion factories are now under an incredible amount of pressure to turn out even more, even faster,” she said.

Lenier and Lu both believe that the key to a more sustainable fashion industry is convincing customers to buy less. Lu said if companies use AI to boost their sales without changing their unsustainable practices, their climate footprints will also grow accordingly. “It’s the overall effect of being able to offer more market-popular items and encourage consumers to purchase more than in the past,” he said. “Of course, the overall carbon impact will be higher.”