Whatnot Buys AI Firm Shaped to Speed Up Live Shopping Picks
Livestream giant Whatnot has absorbed recommendation startup Shaped to cut its suggestion lag from minutes to seconds, and India's live-commerce race should take note.
The News
Whatnot, the American livestream shopping platform valued at more than $11 billion, has bought Shaped, a machine learning startup that builds real-time recommendation and search engines. The deal was announced on 15 July 2026. Terms were not disclosed.
Shaped's founder and chief executive Tullie Murrell, a former Meta engineer, joins Whatnot with roughly a dozen engineers and AI researchers. Shaped had built personalised discovery systems that blend customer data with large language models and conventional machine learning, and previously counted the shopping network QVC and rental marketplace Outdoorsy among its clients.
For Whatnot, the logic is speed. The company processes more than 500,000 hours of live video every week, with millions of real-time interactions as auctions open and inventory changes by the second. Its current recommendation systems can take minutes to react. The goal, executives say, is to collapse that towards genuine real-time suggestions.
"By combining Shaped's technology with Whatnot's existing systems, we can make recommendations faster, more responsive, and more personalised," said Emmanuel Fuentes, Whatnot's vice-president of data and AI.
Whatnot, founded in 2019, raised $225 million in a Series F round that lifted its valuation past $11 billion. Sellers have now crossed one billion cumulative orders, the platform added 20 million buyers over the past year, and it launched more than 35 new product categories in 2025 and a further 45 in the first half of 2026.
Why It Matters
This is a classic acqui-hire: a fast-scaling consumer platform absorbing a small, specialised AI team rather than building the capability in-house. The prize being fought over is attention. Live commerce works only when the right product surfaces in front of the right viewer during the few seconds an auction is live. A recommendation that lands a minute late is worthless. That makes latency, not just accuracy, the competitive frontier, and explains why Whatnot bought an entire team to shave seconds off its response time.
The last comparable wave was in traditional e-commerce a decade ago, when Amazon and Flipkart turned recommendation engines into their core moat. The difference now is the clock. Batch systems that refreshed suggestions overnight are useless in a format where the inventory is a person holding up a jersey for ninety seconds.
Indian Angle
For India, this is a preview of a battle already under way at home. A widely cited Bain and Company report with Peak XV estimated India's social commerce opportunity at $16 to $20 billion in gross merchandise value by 2025, with a longer runway towards $60 to $70 billion by 2030. Meesho, Flipkart's Shopsy and a wave of vernacular-first startups are all chasing the same live-selling behaviour that Whatnot has scaled in the West.
The deal signals where the Indian contest will be won: not on catalogue size, but on the quality and speed of the recommendation layer. Indian platforms operate across dozens of languages and price-sensitive tiers, which makes real-time personalisation both harder and more valuable. It also underscores how much of that intelligence is built by small teams, many staffed by Indian-origin engineers, that larger platforms later absorb.
There is a regulatory shadow too. As ONDC pushes to democratise digital commerce and the government scrutinises how platforms rank sellers, the recommendation engine becomes a policy question. Who a live-shopping algorithm chooses to spotlight is, increasingly, a competition matter Indian regulators will not ignore.
FAQ
How much did Whatnot pay for Shaped?
The companies did not disclose the deal terms. Shaped's founder Tullie Murrell and around a dozen engineers and AI researchers are joining Whatnot, which suggests the transaction was structured primarily as a talent-and-technology acquisition rather than a headline-grabbing cash purchase.
What does Shaped's technology actually do?
Shaped builds real-time recommendation and search systems. It combines a customer's behavioural data with large language models and machine learning to personalise what products or content a user sees. Its earlier clients included the shopping network QVC and rental marketplace Outdoorsy.
Why is real-time speed so important for live shopping?
In a live auction, inventory changes second by second and a product is on screen for moments. A suggestion that arrives minutes later, as Whatnot's older systems could take, misses the sale entirely. Cutting recommendation latency towards real time directly lifts conversion during those fleeting live windows.
What does this mean for Indian live-commerce startups?
It raises the bar. Indian players such as Meesho and Flipkart's Shopsy compete in a market Bain and Peak XV size at $16 to $20 billion by 2025. The Whatnot deal signals that winning will depend less on catalogue breadth and more on fast, multilingual, personalised recommendation engines.
Where can I read the original announcement?
The acquisition was reported by TechCrunch, which detailed the deal, the team joining Whatnot and the platform's latest growth figures. A link to the full original coverage appears at the end of this article.
This story was reported by TechCrunch. Read the full original coverage at TechCrunch.