Behind the secondhand clothes we buy online lies a supply chain that feels stuck in the 19th century: warehouses full of clothing bales, graders sorting garments by hand for quality and style, paper records and back-and-forth negotiations between buyers and sellers. While global demand for pre-owned fashion is exploding, the supply chain has remained manual, slow, and full of friction. Fleek, a UK-based startup founded in 2021, has just raised a $25 million Series B to change that picture, led by Burda Principal Investments with participation from eBay, FJ Labs, H14, and existing investors Andreessen Horowitz, HV Capital, and Y Combinator.

The core of the platform is a B2B marketplace that connects wholesale used-clothing suppliers with retailers, resellers, and boutiques in more than 100 countries. The real discontinuity, however, lies in the internally developed AI tools. Fleek Sort, the proprietary model, identifies, categorizes, and grades garments from images and videos, automating sorting and grading operations that still rely heavily on human hands and eyes today. Inventory is digitized before being listed, while dynamic pricing, recommendation, and matching models help supply meet demand on a global scale.

Fleek’s thesis is straightforward: “The system is broken, the market it serves is exploding, and nobody is building the technology and infrastructure to fix it,” said Abhi Arora, co-founder and CEO. The platform’s numbers back it up: more than 2,000 wholesale suppliers and graders connected to over 50,000 buyers, and over 12 million clothing items kept in circulation. The funding will go toward growing the engineering team, scaling the platform, and further developing AI capabilities across the chain – from inventory processing and merchandising to buyer discovery.

This investment signals a structural shift: AI applied to physical goods is no longer confined to fast-fashion giants or Amazon-style automated warehouses. Even fragmented, low-margin markets like secondhand are going through a phase of technological infrastructure building. But the stakes go beyond operational efficiency. As sorting and pricing are increasingly handed to proprietary models, the control point for data shifts. Whoever trains and hosts the model – in this case, Fleek’s cloud platform – accumulates growing information about transactions, buying preferences, and lot characteristics, strengthening its position as an intermediary.

For suppliers, especially larger ones handling significant volumes, the decision to entrust cataloging and pricing to an external marketplace is not neutral. Handing over inventory data to a cloud model means giving up a degree of informational sovereignty, exposing oneself to the risk that the platform’s algorithm might end up favoring other players competitively. Though the speed and access to global demand currently make it hard to bypass such platforms, the landscape may evolve. It is not far-fetched that in the future, AI sorting solutions for secondhand goods could be run on-premises, returning data control to suppliers and offering an alternative to the centralized logic of large marketplaces. The game, in short, has only just begun.