When a shopper opens a livestream on Whatnot, they have roughly 50 milliseconds to see something that catches their eye before they scroll on. That timing tension is exactly why the company just acquired Shaped, a New York-based AI startup that builds real-time ranking systems for recommendations and search. The deal, announced Wednesday, signals that Whatnot is betting its next phase of growth on AI that can keep pace with a marketplace that changes by the minute.
Whatnot has become one of the fastest-growing social commerce platforms since launching in 2019, originally focused on collectibles and trading cards before expanding into categories like luxury goods, electronics, sneakers, and home decor. The company was valued at $4.5 billion in 2022 and has raised over $750 million from investors including Andreessen Horowitz, Sequoia Capital, and BOND. Now it is doubling down on AI as its primary lever for scaling discovery across a rapidly expanding catalog.
The Deal and What It Brings
Under the terms of the acquisition, Shaped founder and CEO Tullie Murrell will lead a newly formed applied AI research group at Whatnot. Nearly a dozen Shaped engineers and researchers will join the company to accelerate its work on discovery and personalization across the marketplace. Financial terms of the transaction were not disclosed.
Emmanuel Fuentes, VP of Data and AI at Whatnot, framed the acquisition as a natural extension of the company's existing AI work. Since the early days of Whatnot, data and applied machine learning have been at the core of the platform, from early collaborative filters to real-time inference and recommendation systems. Fuentes noted that Shaped has been operating at that same intersection for years, shipping production-ready ML and AI for marketplaces like Whatnot's.
Shaped was founded in 2021 by Murrell, Daniel Camilleri, and Shai Bruhis. The startup raised $10 million in equity funding from Madrona, Y Combinator, Liquid 2 Ventures, Uncommon Capital, and Tribe Capital. Its platform delivered sub-50 millisecond relevance across text, user, and session data, serving as ranking infrastructure for hundreds of marketplaces and content brands. For a live commerce environment where inventory shifts by the second, that kind of speed is not a luxury; it is a requirement.
Why Real-Time AI Is the Hardest Problem in Live Commerce
Live commerce presents a fundamentally different recommendation challenge than traditional e-commerce. On Amazon or Shopify, product catalogs are relatively static. A recommendation engine can run batch computations, update models overnight, and serve the same suggestions to thousands of users over hours or days. In a livestream marketplace, none of that works.
Inventory changes by the second as auctions close and new items go live. Buyer intent shifts as hosts move between products. And the platform must balance what a single buyer wants right now with what keeps the entire marketplace healthy over the long term. Shaped's platform was built specifically to handle these multi-sided dynamics.
Murrell described the challenge succinctly: in a live ecosystem, the standard recommendation playbook does not apply. You have to balance immediate buyer relevance with the long-term health of the marketplace. Building an AI engine that can manage those dynamics at enormous scale is a massive technical challenge.
The numbers explain why Whatnot is investing now. Cross-category buying on the platform has grown 170% year-over-year, meaning buyers are increasingly exploring categories far beyond what they originally came for. In just the first half of 2026, Whatnot launched over 45 new product categories. That kind of expansion strains any recommendation system built for a narrower catalog. Shaped's infrastructure is intended to sharpen those discovery systems as Whatnot scales.
What the Acquisition Signals for the Broader Market
This deal is part of a broader pattern in e-commerce and marketplace technology: companies are buying AI talent and infrastructure rather than building it from scratch. The acqui-hire has evolved into something more strategic. Whatnot is not just absorbing Shaped's team; it is absorbing a production-tested recommendation platform that already works at marketplace scale.
For founders building in the AI and commerce intersection, the message is clear. The companies that win in live commerce will be the ones that solve real-time personalization first. Inventory-aware, intent-aware, and latency-constrained recommendation systems represent a moat that is expensive to build and hard to replicate. Shaped spent years and millions of dollars building that moat. Whatnot just acquired it in a single transaction.
The acquisition also underscores a shift in how live commerce platforms think about competition. Whatnot competes not just with other live shopping apps like TalkShopLive and NTWRK, but with every marketplace that offers instant discovery. TikTok Shop, Amazon Live, and Instagram Shopping are all racing to build the same capabilities. By buying Shaped rather than building internally, Whatnot effectively compresses years of R&D into a single quarter.
For Whatnot's millions of buyers, the change will be invisible but immediate. Recommendations will get sharper. The platform will surface products from categories a buyer has never explored before. And the 50-millisecond window that determines whether someone stays or scrolls will get narrower and more precise. That is the edge that Shaped's AI was built to deliver.
