In AI, yesterday’s jaw dropping valuation is today’s starting point. Eighteen months after two high school friends joined the Y Combinator startup incubator with no idea and no product, their AI data company, AfterQuery, has reached a $3.2 billion valuation, two people with direct knowledge of the matter told Forbes .

That’s more than ten times the $300 million valuation AfterQuery received five months ago. It would make the company the fastest startup in YC history to go from inception to unicorn status, according to YC partner Gustaf Alströmer.

AfterQuery declined to comment.

AfterQuery is profitable and has lined up a lead investor, one of the people said. Its 23 year-old founder Spencer Mateega posted to X in July that recurring revenue was in the “hundreds of millions,” up from $100 million in April.

San Francisco-based AfterQuery is part of a fast growing class of AI data providers that provide what frontier AI labs increasingly need more than anything else: complex reasoning data created and vetted by humans. Its work was used in Nvidia’s new series of Nemotron models, and the company has also worked with former OpenAI CTO Mira Murati’s Thinking Machines Lab and legal AI startup Legora. Unlike some of its peers, AfterQuery works closely with some Chinese AI labs as well.

The state of the market explains the speed. Frontier labs have already swallowed most of the web. Synthetic data can help, but it has limits. What’s most desirable is domain-specific data created by people who know what they’re doing: software engineers, lawyers, financial analysts.

That shift has turned formerly unglamorous data labeling into one of the quickest paths to behemoth valuation and billionaire status. In 2021, Scale AI made then 24 year-old Alexandr Wang the original data labeling billionaire and the world’s youngest self-made one until last October when Mercor’s founders surpassed him at age 22. That data labeling company is now in talks with Nvidia at a $20 billion valuation.

Founded in February 2025 by Mateega, 23, and Carlos Georgescu, 22, AfterQuery did not begin as a data company. The duo first planned to build AI agents for finance. But as they tested, they found that leading models were falling flat on nuanced white-collar workflows. This wasn’t because of architecture limits, but because they lacked the training needed to make professional judgement calls and decisions.

So Mateega and Georgescu pivoted. Instead of building apps on top of AI, , they focused on capturing expert human-generated judgment calls and step-by-step reasoning across finance, software engineering, law, and medicine.

Mateega told Forbes in April that AfterQuery’s edge over Mercor—which relies on large pools of contractors selected by an AI interviewer—are custom software systems to validate human-generated training data.

AfterQuery’s human experts generate data screened with a series of checks to ensure that it falls within a “Goldilocks” range — difficult enough to challenge state-of-the-art systems, but not stump it. The goal is to produce data that models can actually learn from.

AfterQuery also publishes its own research to prove its data actually works, a key concern for AI labs spending heavily on training material. Instead of handing data to AI labs and letting them evaluate it, AfterQuery replicates what a researcher would do: train a model on the data and measure how benchmark performance changes.

“This is another thing which our peers do not do: we have researchers internally create a post-training pipeline,” Mateega said in April. “We objectively show to the labs before they even look at our data that the data’s high quality.”