Snorkel AI raises $350M at $3.5B valuation as revenue surges 17X on AI training data boom

Snorkel AI has raised $350 million at a $3.5 billion valuation after its annualized revenue run-rate exploded from roughly $20 million to more than $350 million in about a year, a striking sign of how valuable high-quality training data has become as AI labs push models into harder tasks.

The San Francisco startup’s latest round was led by Insight Partners and S32, with participation from new investors March Capital, Blumberg Capital, Allegis Capital, Frontline, Standard, and Third Point Ventures. Existing backers Greylock, Lightspeed, GV, Factory, Prosperity7, Walden Catalyst, and Wells Fargo also joined the round, the company said in a news release Tuesday. The new valuation is nearly triple the $1.3 billion Snorkel reached after raising $100 million in May 2025.

Snorkel will use the fresh capital infusion to expand its agentic data factory, which supplies the specialized datasets, reinforcement-learning environments, and evaluation systems used by advanced AI labs and other customers.

“The teams pushing the frontier want a research data partner who pioneers the science of data development,” said Alex Ratner, co-founder and CEO of Snorkel AI. “That’s what Snorkel was built to be: the frontier lab for agentic data, combining human excellence with over a decade of research and technology.”

The bigger story may be what happened inside the business.

Founded in 2019 by researchers from Stanford’s AI lab, Snorkel started out selling software that helped companies build and manage training datasets. It has since shifted toward delivering finished datasets and reinforcement-learning environments directly to customers.

That shift appears to have transformed the company’s economics. Snorkel said its annualized revenue run-rate has passed $350 million, up from about $20 million a year earlier, with much of the growth coming from a data-as-a-service business launched in September 2025.

AI labs are running out of easy data

The race to build more capable AI systems is changing what labs need from data providers.

Simple labeling jobs are giving way to harder assignments involving coding, law, medicine, and other specialist fields. Training AI agents can require realistic tasks, simulated environments, detailed grading criteria, and expert judgment about whether a model actually produced the right result.

Snorkel’s answer combines human specialists with thousands of AI models and agents. Experts create scenarios, tasks, and grading rubrics, then automated systems take over much of the repetitive quality-control work.

The company draws from a network of tens of thousands of specialists and sells the resulting data products rather than billing clients for hours of human labor. Coding has emerged as one of its largest areas of demand, Reuters reported. Snorkel says its customers include frontier AI labs, hyperscalers, large enterprises and the U.S. federal government.

Ratner believes humans will remain part of that process.

“Our strong view is that 100% of the data that labs will get value out of will have some human input in the foreseeable future,” Ratner said. “But 100% of that data will have to use synthetic and automated approaches to keep up with this complexity.”

That tension is shaping one of AI’s hottest infrastructure markets. Model developers need increasingly specialized examples, yet relying entirely on human workers becomes expensive and slow at scale. The emerging model mixes expert judgment with synthetic data and automation.

Investor interest in the sector has intensified since Meta paid $14.3 billion for a 49% stake in Scale AI in June 2025. Rivals including Mercor and Surge AI have attracted capital as investors chase companies sitting between AI labs and the specialized human knowledge needed to train their models.

Snorkel plans to use the new funding to hire researchers and engineers, grow its enterprise and government businesses, support third-party model evaluations, and move into new industries and data formats. According to Reuters, the company expects to reach profitability this year.

“Data is becoming more rare, more specialized, more difficult to find,” S32 partner Andy Harrison said. “If you want to train the most frontier, complex, and capable models, now you need superior data.”

For Snorkel, that scarcity has turned what once looked like a supporting layer of the AI stack into a $3.5 billion business.