AfterQuery records the fastest single valuation jump in Y Combinator history, reaching over $3 billion to train agents to simulate professionals

Press reports said that the startup AfterQuery, which specializes in AI training data, closed a financing round that raised its market valuation to $3.2 billion, after only five months since announcing its first $30 million round at a $300 million valuation in April. This jump, which represents a more than ten-fold increase in less than half a year, made the company the fastest to become a billion-dollar firm in the history of the Y Combinator accelerator, according to partner at the accelerator Gustav Alströmer, exceeding the usual timeframes for rising tech companies, after Forbes was the first to reveal details of the round.

The company was founded in San Francisco by two entrepreneurs who are now 22 and 23 years old, and who graduated from the winter 2025 batch of Y Combinator about 18 months ago. In April, the company reported that its current annual revenue run-rate had reached $100 million, relying on a client base that includes major AI labs and advanced technology firms such as Nvidia, Ligura, and the Korean lab Motif Technologies, reflecting a rapid demand from foundational model developers for high-value specialized data to feed the next generation of autonomous systems.

AfterQuery belongs to a new generation of startups following in the footsteps of Scale and Mercur, relying on hiring highly knowledgeable professionals such as doctors, lawyers and other specialists to train models. However, the core difference in its business model lies in going beyond traditional answer-accuracy tests; the company does not stop at validating the correctness of responses to queries, but trains models and independent agents on how to perform complex tasks as professionals do in their fields. The company describes this methodology asencoding the patterns, decision-making and reasoning logic of the world’s top practitioners, to transfer real-world problem-solving skills into algorithmic environments.

This shift in AI training architecture has direct implications for work and technology environments in the Gulf, Egypt and the Levant, as it imposes a practical reallocation of spending priorities and locally required skills. Training models and agents is no longer limited to importing ready-made weights or hiring teams to label simple text; it now demands attracting specialists in law, medicine and engineering to document complex institutional decision pathways and train agents on them. For regional companies and institutions building automated solutions in critical sectors, the bet has moved from merely deploying conventional conversational interfaces to verifying that a software agent can carry out transactions and execute branching procedures with reliable professional accuracy.

This rapid rise in the cost of advanced training data also puts technology leaders in the region faced with a decisive operational choice: either rely on outputs from global data-providers at escalating prices to fine-tune agent performance, or build local training pipelines that invest in regional professional expertise to encode procedures, regulations and operational nuances of our markets. Enabling a smart agent to complete a local legal or financial procedure requires a understanding of the actual decision sequence within institutions, making specialized human expertise an indispensable foundation for determining the cost and quality of software systems, and shaping how well our business environments can secure a place in the value-added chains of the AI industry.